[ARM] Add larger than legal ICmp costs

[dmgreen]

Summary

A v8i32 compare will produce a v8i1 predicate, but during codegen the v8i32 will be split into two v4i32, potentially requiring two v4i1 predicates to be merged into a single v8i1. Because this merging of two v4i1's into a v8i1 is very expensive, we need to make the cost of the compare equally high.

This patch adds the cost of that to ARMTTIImpl::getCmpSelInstrCost. Because we don't know whether the user of the predicate can be split, and the cost model is mostly pre-instruction, we may be pessimistic, but that should only be for larger and legal types. This does add a min/max detection to the costmodel where it can be detected, to keep those in line with the cost of simple min/max instructions. Otherwise for the most part, costs that were already expensive have become more expensive.

[SjoerdMeijer]

Looks reasonable.

llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp

863

Nit: *Sel

896

Nit: if (IID)

937

Nit: do we need in comments what these false/true booleans are?

llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp

if (CostKind == TTI::TCK_CodeSize && ISD == ISD::SELECT &&
     // i1 values may need rematerialising by using mov immediates and/or
     // flag setting instructions.
     if (ValTy->isIntegerTy(1))
       ++Cost;
 
     return Cost;
   }
 
+  // If this is a vector min/max/abs, use the cost of that intrinsic directly
+  // instead. Hopefully when min/max intrinsics are more prevalent this code
+  // will not be needed.
+  const Instruction* Sel = I;
+  if ((Opcode == Instruction::ICmp || Opcode == Instruction::FCmp) && Sel &&
+      Sel->hasOneUse())
+    Sel = cast<Instruction>(Sel->user_back());
+  if (Sel && ValTy->isVectorTy()) {
+    const Value *LHS, *RHS;
+    SelectPatternFlavor SPF = matchSelectPattern(Sel, LHS, RHS).Flavor;
+    unsigned IID = 0;
+    switch(SPF) {
+    case SPF_ABS:
+      IID = Intrinsic::abs;
+      break;
+    case SPF_SMIN:
+      IID = Intrinsic::smin;
+      break;
+    case SPF_SMAX:
+      IID = Intrinsic::smax;
+      break;
+    case SPF_UMIN:
+      IID = Intrinsic::umin;
+      break;
+    case SPF_UMAX:
+      IID = Intrinsic::umax;
+      break;
+    case SPF_FMINNUM:
+      IID = Intrinsic::minnum;
+      break;
+    case SPF_FMAXNUM:
+      IID = Intrinsic::maxnum;
+      break;
+    default:
+      break;
+    }
+    if (IID != 0) {
+      // The ICmp is free, the select gets the cost of the min/max/etc
+      if (Sel != I)
+        return 0;
+      IntrinsicCostAttributes CostAttrs(IID, ValTy, {ValTy, ValTy});
+      return getIntrinsicInstrCost(CostAttrs, CostKind);
+    }
+  }
+
   // On NEON a vector select gets lowered to vbsl.
   if (ST->hasNEON() && ValTy->isVectorTy() && ISD == ISD::SELECT && CondTy) {
     // Lowering of some vector selects is currently far from perfect.
     static const TypeConversionCostTblEntry NEONVectorSelectTbl[] = {
       { ISD::SELECT, MVT::v4i1, MVT::v4i64, 4*4 + 1*2 + 1 },
       { ISD::SELECT, MVT::v8i1, MVT::v8i64, 50 },
       { ISD::SELECT, MVT::v16i1, MVT::v16i64, 100 }
     };
 
     EVT SelCondTy = TLI->getValueType(DL, CondTy);
     EVT SelValTy = TLI->getValueType(DL, ValTy);
     if (SelCondTy.isSimple() && SelValTy.isSimple()) {
       if (const auto *Entry = ConvertCostTableLookup(NEONVectorSelectTbl, ISD,
                                                      SelCondTy.getSimpleVT(),
                                                      SelValTy.getSimpleVT()))
         return Entry->Cost;
     }
 
     std::pair<int, MVT> LT = TLI->getTypeLegalizationCost(DL, ValTy);
     return LT.first;
   }
 
+  if (ST->hasMVEIntegerOps() && ValTy->isVectorTy() &&
+      (Opcode == Instruction::ICmp || Opcode == Instruction::FCmp) &&
+      cast<FixedVectorType>(ValTy)->getNumElements() > 1) {
+    FixedVectorType *VecValTy = cast<FixedVectorType>(ValTy);
+    FixedVectorType *VecCondTy = dyn_cast_or_null<FixedVectorType>(CondTy);
+    if (!VecCondTy)
+      VecCondTy = cast<FixedVectorType>(CmpInst::makeCmpResultType(VecValTy));
+
+    // If we don't have mve.fp any fp operations will need to be scalarized.
+    if (Opcode == Instruction::FCmp && !ST->hasMVEFloatOps()) {
+      return BaseT::getScalarizationOverhead(VecValTy, false, true) +
+             BaseT::getScalarizationOverhead(VecCondTy, true, false) +
+             VecValTy->getNumElements() *
+                 getCmpSelInstrCost(Opcode, ValTy->getScalarType(),
+                                    CondTy->getScalarType(), VecPred, CostKind,
+                                    I);
+    }
+
+    std::pair<unsigned, MVT> LT = TLI->getTypeLegalizationCost(DL, ValTy);
+    int BaseCost = ST->getMVEVectorCostFactor(CostKind);
+    // There are two types - the input that specifies the type of the compare
+    // and the output vXi1 type. Because we don't know how the output will be
+    // split, we may need an expensive shuffle to get two in sync. This has the
+    // effect of making larger than legal compares (v8i32 for example)
+    // expensive.
+    if (LT.second.getVectorNumElements() > 2) {
+      if (LT.first > 1)
+        return LT.first * BaseCost +
+               BaseT::getScalarizationOverhead(VecCondTy, true, false);
+      return BaseCost;
+    }
+  }
+
   // Default to cheap (throughput/size of 1 instruction) but adjust throughput
   // for "multiple beats" potentially needed by MVE instructions.
   int BaseCost = 1;
   if (ST->hasMVEIntegerOps() && ValTy->isVectorTy())
     BaseCost = ST->getMVEVectorCostFactor(CostKind);
 
   return BaseCost *
          BaseT::getCmpSelInstrCost(Opcode, ValTy, CondTy, VecPred, CostKind, I);

llvm/test/Analysis/CostModel/ARM/arith-overflow.ll

 ;
 ; MVE-RECIP-LABEL: 'sadd'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %I64 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.sadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 242 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.sadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 866 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.sadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I32 = call { i32, i1 } @llvm.sadd.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.sadd.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 306 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.sadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1122 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.sadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 498 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.sadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1890 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.sadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I16 = call { i16, i1 } @llvm.sadd.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 298 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.sadd.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1106 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.sadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4260 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.sadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1874 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.sadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7332 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.sadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I8 = call { i8, i1 } @llvm.sadd.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1098 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.sadd.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4244 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.sadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16680 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.sadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7316 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.sadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 28968 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.sadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'sadd'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.sadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.sadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.sadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I32 = call { i32, i1 } @llvm.sadd.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-SIZE-LABEL: 'sadd'
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.sadd.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.sadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.sadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 150 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.sadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I32 = call { i32, i1 } @llvm.sadd.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.sadd.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.sadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.sadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 203 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.sadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 787 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.sadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I16 = call { i16, i1 } @llvm.sadd.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.sadd.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.sadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.sadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 779 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.sadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3092 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.sadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I8 = call { i8, i1 } @llvm.sadd.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.sadd.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.sadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.sadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3084 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.sadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12310 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.sadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = call {i64, i1} @llvm.sadd.with.overflow.i64(i64 undef, i64 undef)
   %V2I64 = call {<2 x i64>, <2 x i1>} @llvm.sadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
   %V4I64 = call {<4 x i64>, <4 x i1>} @llvm.sadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
   %V8I64 = call {<8 x i64>, <8 x i1>} @llvm.sadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 
   %I32 = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-RECIP-LABEL: 'uadd'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.uadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.uadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 288 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.uadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = call { i32, i1 } @llvm.uadd.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.uadd.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.uadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.uadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 72 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.uadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 272 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.uadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call { i16, i1 } @llvm.uadd.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.uadd.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.uadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.uadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 264 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.uadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1040 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.uadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call { i8, i1 } @llvm.uadd.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.uadd.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.uadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.uadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1032 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.uadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4112 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.uadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'uadd'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.uadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.uadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.uadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = call { i32, i1 } @llvm.uadd.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-SIZE-LABEL: 'uadd'
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.uadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.uadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 145 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.uadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = call { i32, i1 } @llvm.uadd.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.uadd.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.uadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.uadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.uadd.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 264 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.uadd.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call { i16, i1 } @llvm.uadd.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.uadd.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.uadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.uadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 260 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.uadd.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1032 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.uadd.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call { i8, i1 } @llvm.uadd.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.uadd.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.uadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.uadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1028 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.uadd.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4104 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.uadd.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 undef, i64 undef)
   %V2I64 = call {<2 x i64>, <2 x i1>} @llvm.uadd.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
   %V4I64 = call {<4 x i64>, <4 x i1>} @llvm.uadd.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
   %V8I64 = call {<8 x i64>, <8 x i1>} @llvm.uadd.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 
   %I32 = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-RECIP-LABEL: 'ssub'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %I64 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.ssub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 242 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.ssub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 866 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.ssub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I32 = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 90 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.ssub.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 306 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.ssub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1122 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.ssub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 498 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.ssub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1890 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.ssub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I16 = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 298 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.ssub.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1106 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.ssub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4260 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.ssub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1874 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.ssub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7332 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.ssub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I8 = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1098 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.ssub.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4244 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.ssub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16680 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.ssub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7316 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.ssub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 28968 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.ssub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'ssub'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.ssub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.ssub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.ssub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I32 = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-SIZE-LABEL: 'ssub'
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.ssub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.ssub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 150 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.ssub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I32 = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.ssub.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.ssub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.ssub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 203 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.ssub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 787 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.ssub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I16 = call { i16, i1 } @llvm.ssub.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.ssub.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.ssub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.ssub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 779 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.ssub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3092 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.ssub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %I8 = call { i8, i1 } @llvm.ssub.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.ssub.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.ssub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.ssub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3084 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.ssub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 12310 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.ssub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = call {i64, i1} @llvm.ssub.with.overflow.i64(i64 undef, i64 undef)
   %V2I64 = call {<2 x i64>, <2 x i1>} @llvm.ssub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
   %V4I64 = call {<4 x i64>, <4 x i1>} @llvm.ssub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
   %V8I64 = call {<8 x i64>, <8 x i1>} @llvm.ssub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 
   %I32 = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-RECIP-LABEL: 'usub'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.usub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.usub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 288 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.usub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.usub.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.usub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.usub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 72 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.usub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 272 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.usub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.usub.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.usub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.usub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 264 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.usub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1040 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.usub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.usub.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.usub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.usub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1032 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.usub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4112 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.usub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'usub'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.usub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.usub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.usub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-SIZE-LABEL: 'usub'
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I64 = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.usub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 41 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.usub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 145 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.usub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I32 = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.usub.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.usub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.usub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.usub.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 264 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.usub.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I16 = call { i16, i1 } @llvm.usub.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.usub.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.usub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.usub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 260 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.usub.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1032 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.usub.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %I8 = call { i8, i1 } @llvm.usub.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.usub.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.usub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.usub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1028 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.usub.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4104 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.usub.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = call {i64, i1} @llvm.usub.with.overflow.i64(i64 undef, i64 undef)
   %V2I64 = call {<2 x i64>, <2 x i1>} @llvm.usub.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
   %V4I64 = call {<4 x i64>, <4 x i1>} @llvm.usub.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
   %V8I64 = call {<8 x i64>, <8 x i1>} @llvm.usub.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 
   %I32 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-RECIP-LABEL: 'smul'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.smul.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.smul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 288 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.smul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 768 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.smul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I32 = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.smul.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 316 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.smul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1208 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.smul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 380 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.smul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1464 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.smul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = call { i16, i1 } @llvm.smul.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.smul.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 164 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.smul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 488 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.smul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 420 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.smul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1512 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.smul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = call { i8, i1 } @llvm.smul.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.smul.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 228 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.smul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 616 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.smul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1252 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.smul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4712 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.smul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'smul'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.smul.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.smul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.smul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.smul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I32 = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-SIZE-LABEL: 'smul'
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 13 for instruction: %I64 = call { i64, i1 } @llvm.smul.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.smul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 101 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.smul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 325 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.smul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %I32 = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.smul.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 153 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.smul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 557 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.smul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 218 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.smul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 816 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.smul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I16 = call { i16, i1 } @llvm.smul.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.smul.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 75 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.smul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 147 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.smul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 332 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.smul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1174 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.smul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I8 = call { i8, i1 } @llvm.smul.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 71 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.smul.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 139 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.smul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 275 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.smul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1164 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.smul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4374 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.smul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = call {i64, i1} @llvm.smul.with.overflow.i64(i64 undef, i64 undef)
   %V2I64 = call {<2 x i64>, <2 x i1>} @llvm.smul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
   %V4I64 = call {<4 x i64>, <4 x i1>} @llvm.smul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
   %V8I64 = call {<8 x i64>, <8 x i1>} @llvm.smul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 
   %I32 = call {i32, i1} @llvm.smul.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-RECIP-LABEL: 'umul'
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %I64 = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 108 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.umul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 248 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.umul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 624 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.umul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.umul.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 312 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.umul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1200 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.umul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 376 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.umul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1456 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.umul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call { i16, i1 } @llvm.umul.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.umul.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.umul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 480 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.umul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 416 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.umul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1504 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.umul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 92 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.umul.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 224 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.umul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 608 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.umul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 1248 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.umul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 4704 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.umul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-RECIP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; V8M-SIZE-LABEL: 'umul'
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %I64 = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 undef, i64 undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.umul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 29 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.umul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 53 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.umul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; V8M-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 undef, i32 undef)

 ;
 ; MVE-SIZE-LABEL: 'umul'
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 11 for instruction: %I64 = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 undef, i64 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 25 for instruction: %V2I64 = call { <2 x i64>, <2 x i1> } @llvm.umul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 61 for instruction: %V4I64 = call { <4 x i64>, <4 x i1> } @llvm.umul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 181 for instruction: %V8I64 = call { <8 x i64>, <8 x i1> } @llvm.umul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %I32 = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 undef, i32 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V4I32 = call { <4 x i32>, <4 x i1> } @llvm.umul.with.overflow.v4i32(<4 x i32> undef, <4 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 151 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.umul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 553 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.umul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 216 for instruction: %V8I32 = call { <8 x i32>, <8 x i1> } @llvm.umul.with.overflow.v8i32(<8 x i32> undef, <8 x i32> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 812 for instruction: %V16I32 = call { <16 x i32>, <16 x i1> } @llvm.umul.with.overflow.v16i32(<16 x i32> undef, <16 x i32> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I16 = call { i16, i1 } @llvm.umul.with.overflow.i16(i16 undef, i16 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %V8I16 = call { <8 x i16>, <8 x i1> } @llvm.umul.with.overflow.v8i16(<8 x i16> undef, <8 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 73 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.umul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 143 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.umul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 330 for instruction: %V16I16 = call { <16 x i16>, <16 x i1> } @llvm.umul.with.overflow.v16i16(<16 x i16> undef, <16 x i16> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1170 for instruction: %V32I16 = call { <32 x i16>, <32 x i1> } @llvm.umul.with.overflow.v32i16(<32 x i16> undef, <32 x i16> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %I8 = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 undef, i8 undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 70 for instruction: %V16I8 = call { <16 x i8>, <16 x i1> } @llvm.umul.with.overflow.v16i8(<16 x i8> undef, <16 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 137 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.umul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
-; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 271 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.umul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1162 for instruction: %V32I8 = call { <32 x i8>, <32 x i1> } @llvm.umul.with.overflow.v32i8(<32 x i8> undef, <32 x i8> undef)
+; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4370 for instruction: %V64I8 = call { <64 x i8>, <64 x i1> } @llvm.umul.with.overflow.v64i8(<64 x i8> undef, <64 x i8> undef)
 ; MVE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
 ;
   %I64 = call {i64, i1} @llvm.umul.with.overflow.i64(i64 undef, i64 undef)
   %V2I64 = call {<2 x i64>, <2 x i1>} @llvm.umul.with.overflow.v2i64(<2 x i64> undef, <2 x i64> undef)
   %V4I64 = call {<4 x i64>, <4 x i1>} @llvm.umul.with.overflow.v4i64(<4 x i64> undef, <4 x i64> undef)
   %V8I64 = call {<8 x i64>, <8 x i1>} @llvm.umul.with.overflow.v8i64(<8 x i64> undef, <8 x i64> undef)
 
   %I32 = call {i32, i1} @llvm.umul.with.overflow.i32(i32 undef, i32 undef)

llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll

;
   %s = call i32 @llvm.ctlz.i32(i32 %a, i1 true)
   %v = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %va, i1 true)
   ret void
 }
 
 define void @fshl(i32 %a, i32 %b, i32 %c, <16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc) {
 ; THRU-LABEL: 'fshl'
 ; THRU-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
-; THRU-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 832 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; LATE-LABEL: 'fshl'
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SIZE-LABEL: 'fshl'
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 546 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 805 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SIZE_LATE-LABEL: 'fshl'
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 564 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 826 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
   %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
   ret void
 }
 
 define void @maskedgather(<16 x float*> %va, <16 x i1> %vb, <16 x float> %vc) {

 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   call void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float> %va, <16 x float*> %vb, i32 1, <16 x i1> %vc)
   ret void
 }
 
 define void @reduce_fmax(<16 x float> %va) {
 ; THRU-LABEL: 'reduce_fmax'
-; THRU-NEXT:  Cost Model: Found an estimated cost of 632 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 696 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
 ; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; LATE-LABEL: 'reduce_fmax'
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
 ; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SIZE-LABEL: 'reduce_fmax'
-; SIZE-NEXT:  Cost Model: Found an estimated cost of 620 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 685 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
 ; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; SIZE_LATE-LABEL: 'reduce_fmax'
-; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 628 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 694 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
 ; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
   ret void
 }
 
 define void @memcpy(i8* %a, i8* %b, i32 %c) {
 ; THRU-LABEL: 'memcpy'

llvm/test/Analysis/CostModel/ARM/mve-cmp.ll

 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
-; RUN: opt -cost-model -analyze -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve < %s | FileCheck %s --check-prefixes=CHECK
-; RUN: opt -cost-model -analyze -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve.fp < %s | FileCheck %s --check-prefixes=CHECK
+; RUN: opt -cost-model -analyze -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve < %s | FileCheck %s --check-prefixes=CHECK,CHECK-MVE
+; RUN: opt -cost-model -analyze -mtriple=thumbv8.1m.main-none-eabi -mattr=+mve.fp < %s | FileCheck %s --check-prefixes=CHECK,CHECK-MVEFP
 
 define void @icmp() {
 ; CHECK-LABEL: 'icmp'
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v2i8 = icmp slt <2 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i8 = icmp slt <4 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i8 = icmp slt <8 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16i8 = icmp slt <16 x i8> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v32i8 = icmp slt <32 x i8> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1028 for instruction: %v32i8 = icmp slt <32 x i8> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v2i16 = icmp slt <2 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i16 = icmp slt <4 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8i16 = icmp slt <8 x i16> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16i16 = icmp slt <16 x i16> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 260 for instruction: %v16i16 = icmp slt <16 x i16> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v2i32 = icmp slt <2 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4i32 = icmp slt <4 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8i32 = icmp slt <8 x i32> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16i32 = icmp slt <16 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %v8i32 = icmp slt <8 x i32> undef, undef
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 264 for instruction: %v16i32 = icmp slt <16 x i32> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v2i64 = icmp slt <2 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %v4i64 = icmp slt <4 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 144 for instruction: %v8i64 = icmp slt <8 x i64> undef, undef
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2i8 = icmp slt <2 x i8> undef, undef
   %v4i8 = icmp slt <4 x i8> undef, undef
   %v8i8 = icmp slt <8 x i8> undef, undef
;
   %v2i64 = icmp slt <2 x i64> undef, undef
   %v4i64 = icmp slt <4 x i64> undef, undef
   %v8i64 = icmp slt <8 x i64> undef, undef
 
   ret void
 }
 
 define void @fcmp() {
-; CHECK-LABEL: 'fcmp'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = fcmp olt <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = fcmp olt <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = fcmp olt <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v16f16 = fcmp olt <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f32 = fcmp olt <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f32 = fcmp olt <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v8f32 = fcmp olt <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16f32 = fcmp olt <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v2f64 = fcmp olt <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %v4f64 = fcmp olt <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 144 for instruction: %v8f64 = fcmp olt <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; CHECK-MVE-LABEL: 'fcmp'
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %v2f16 = fcmp olt <2 x half> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %v4f16 = fcmp olt <4 x half> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 136 for instruction: %v8f16 = fcmp olt <8 x half> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v16f16 = fcmp olt <16 x half> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %v2f32 = fcmp olt <2 x float> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %v4f32 = fcmp olt <4 x float> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 136 for instruction: %v8f32 = fcmp olt <8 x float> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v16f32 = fcmp olt <16 x float> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %v2f64 = fcmp olt <2 x double> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %v4f64 = fcmp olt <4 x double> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 136 for instruction: %v8f64 = fcmp olt <8 x double> undef, undef
+; CHECK-MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; CHECK-MVEFP-LABEL: 'fcmp'
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = fcmp olt <2 x half> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = fcmp olt <4 x half> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = fcmp olt <8 x half> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 260 for instruction: %v16f16 = fcmp olt <16 x half> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f32 = fcmp olt <2 x float> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f32 = fcmp olt <4 x float> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 68 for instruction: %v8f32 = fcmp olt <8 x float> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 264 for instruction: %v16f32 = fcmp olt <16 x float> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v2f64 = fcmp olt <2 x double> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %v4f64 = fcmp olt <4 x double> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 144 for instruction: %v8f64 = fcmp olt <8 x double> undef, undef
+; CHECK-MVEFP-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
   %v2f16 = fcmp olt <2 x half> undef, undef
   %v4f16 = fcmp olt <4 x half> undef, undef
   %v8f16 = fcmp olt <8 x half> undef, undef
   %v16f16 = fcmp olt <16 x half> undef, undef
 
   %v2f32 = fcmp olt <2 x float> undef, undef
   %v4f32 = fcmp olt <4 x float> undef, undef

llvm/test/Analysis/CostModel/ARM/reduce-smax.ll

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V16 = call i32 @llvm.vector.reduce.smax.v16i32(<16 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 237 for instruction: %V32 = call i32 @llvm.vector.reduce.smax.v32i32(<32 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i32'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i32 @llvm.vector.reduce.smax.v2i32(<2 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V8 = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 632 for instruction: %V16 = call i32 @llvm.vector.reduce.smax.v16i32(<16 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2184 for instruction: %V32 = call i32 @llvm.vector.reduce.smax.v32i32(<32 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 696 for instruction: %V16 = call i32 @llvm.vector.reduce.smax.v16i32(<16 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2504 for instruction: %V32 = call i32 @llvm.vector.reduce.smax.v32i32(<32 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i32 @llvm.vector.reduce.smax.v2i32(<2 x i32> undef)
   %V4  = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> undef)
   %V8  = call i32 @llvm.vector.reduce.smax.v8i32(<8 x i32> undef)
   %V16 = call i32 @llvm.vector.reduce.smax.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.vector.reduce.smax.v32i32(<32 x i32> undef)
   ret i32 undef

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 503 for instruction: %V64 = call i16 @llvm.vector.reduce.smax.v64i16(<64 x i16> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i16'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i16 @llvm.vector.reduce.smax.v2i16(<2 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i16 @llvm.vector.reduce.smax.v4i16(<4 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i16 @llvm.vector.reduce.smax.v8i16(<8 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 1176 for instruction: %V16 = call i16 @llvm.vector.reduce.smax.v16i16(<16 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2720 for instruction: %V32 = call i16 @llvm.vector.reduce.smax.v32i16(<32 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 8880 for instruction: %V64 = call i16 @llvm.vector.reduce.smax.v64i16(<64 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2976 for instruction: %V32 = call i16 @llvm.vector.reduce.smax.v32i16(<32 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 10160 for instruction: %V64 = call i16 @llvm.vector.reduce.smax.v64i16(<64 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i16 @llvm.vector.reduce.smax.v2i16(<2 x i16> undef)
   %V4  = call i16 @llvm.vector.reduce.smax.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.vector.reduce.smax.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.vector.reduce.smax.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.vector.reduce.smax.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.vector.reduce.smax.v64i16(<64 x i16> undef)

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i8'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i8 @llvm.vector.reduce.smax.v2i8(<2 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i8 @llvm.vector.reduce.smax.v8i8(<8 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 4128 for instruction: %V16 = call i8 @llvm.vector.reduce.smax.v16i8(<16 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 5668 for instruction: %V32 = call i8 @llvm.vector.reduce.smax.v32i8(<32 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 11820 for instruction: %V64 = call i8 @llvm.vector.reduce.smax.v64i8(<64 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 36412 for instruction: %V128 = call i8 @llvm.vector.reduce.smax.v128i8(<128 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 12844 for instruction: %V64 = call i8 @llvm.vector.reduce.smax.v64i8(<64 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 41532 for instruction: %V128 = call i8 @llvm.vector.reduce.smax.v128i8(<128 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2   = call i8 @llvm.vector.reduce.smax.v2i8(<2 x i8> undef)
   %V4   = call i8 @llvm.vector.reduce.smax.v4i8(<4 x i8> undef)
   %V8   = call i8 @llvm.vector.reduce.smax.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.vector.reduce.smax.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.vector.reduce.smax.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.vector.reduce.smax.v64i8(<64 x i8> undef)

llvm/test/Analysis/CostModel/ARM/reduce-smin.ll

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 237 for instruction: %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i32'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i32 @llvm.vector.reduce.smin.v2i32(<2 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V8 = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 632 for instruction: %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2184 for instruction: %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 696 for instruction: %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2504 for instruction: %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i32 @llvm.vector.reduce.smin.v2i32(<2 x i32> undef)
   %V4  = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> undef)
   %V8  = call i32 @llvm.vector.reduce.smin.v8i32(<8 x i32> undef)
   %V16 = call i32 @llvm.vector.reduce.smin.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.vector.reduce.smin.v32i32(<32 x i32> undef)
   ret i32 undef

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 503 for instruction: %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i16'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i16 @llvm.vector.reduce.smin.v2i16(<2 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 1176 for instruction: %V16 = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2720 for instruction: %V32 = call i16 @llvm.vector.reduce.smin.v32i16(<32 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 8880 for instruction: %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2976 for instruction: %V32 = call i16 @llvm.vector.reduce.smin.v32i16(<32 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 10160 for instruction: %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i16 @llvm.vector.reduce.smin.v2i16(<2 x i16> undef)
   %V4  = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.vector.reduce.smin.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.vector.reduce.smin.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.vector.reduce.smin.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.vector.reduce.smin.v64i16(<64 x i16> undef)

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i8'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i8 @llvm.vector.reduce.smin.v2i8(<2 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i8 @llvm.vector.reduce.smin.v8i8(<8 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 4128 for instruction: %V16 = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 5668 for instruction: %V32 = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 11820 for instruction: %V64 = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 36412 for instruction: %V128 = call i8 @llvm.vector.reduce.smin.v128i8(<128 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 12844 for instruction: %V64 = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 41532 for instruction: %V128 = call i8 @llvm.vector.reduce.smin.v128i8(<128 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2   = call i8 @llvm.vector.reduce.smin.v2i8(<2 x i8> undef)
   %V4   = call i8 @llvm.vector.reduce.smin.v4i8(<4 x i8> undef)
   %V8   = call i8 @llvm.vector.reduce.smin.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.vector.reduce.smin.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.vector.reduce.smin.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.vector.reduce.smin.v64i8(<64 x i8> undef)

llvm/test/Analysis/CostModel/ARM/reduce-umax.ll

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V16 = call i32 @llvm.vector.reduce.umax.v16i32(<16 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 237 for instruction: %V32 = call i32 @llvm.vector.reduce.umax.v32i32(<32 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i32'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i32 @llvm.vector.reduce.umax.v2i32(<2 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V8 = call i32 @llvm.vector.reduce.umax.v8i32(<8 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 632 for instruction: %V16 = call i32 @llvm.vector.reduce.umax.v16i32(<16 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2184 for instruction: %V32 = call i32 @llvm.vector.reduce.umax.v32i32(<32 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 696 for instruction: %V16 = call i32 @llvm.vector.reduce.umax.v16i32(<16 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2504 for instruction: %V32 = call i32 @llvm.vector.reduce.umax.v32i32(<32 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i32 @llvm.vector.reduce.umax.v2i32(<2 x i32> undef)
   %V4  = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> undef)
   %V8  = call i32 @llvm.vector.reduce.umax.v8i32(<8 x i32> undef)
   %V16 = call i32 @llvm.vector.reduce.umax.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.vector.reduce.umax.v32i32(<32 x i32> undef)
   ret i32 undef

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 503 for instruction: %V64 = call i16 @llvm.vector.reduce.umax.v64i16(<64 x i16> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i16'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i16 @llvm.vector.reduce.umax.v2i16(<2 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i16 @llvm.vector.reduce.umax.v4i16(<4 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i16 @llvm.vector.reduce.umax.v8i16(<8 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 1176 for instruction: %V16 = call i16 @llvm.vector.reduce.umax.v16i16(<16 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2720 for instruction: %V32 = call i16 @llvm.vector.reduce.umax.v32i16(<32 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 8880 for instruction: %V64 = call i16 @llvm.vector.reduce.umax.v64i16(<64 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2976 for instruction: %V32 = call i16 @llvm.vector.reduce.umax.v32i16(<32 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 10160 for instruction: %V64 = call i16 @llvm.vector.reduce.umax.v64i16(<64 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i16 @llvm.vector.reduce.umax.v2i16(<2 x i16> undef)
   %V4  = call i16 @llvm.vector.reduce.umax.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.vector.reduce.umax.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.vector.reduce.umax.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.vector.reduce.umax.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.vector.reduce.umax.v64i16(<64 x i16> undef)

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i8'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i8 @llvm.vector.reduce.umax.v2i8(<2 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i8 @llvm.vector.reduce.umax.v8i8(<8 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 4128 for instruction: %V16 = call i8 @llvm.vector.reduce.umax.v16i8(<16 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 5668 for instruction: %V32 = call i8 @llvm.vector.reduce.umax.v32i8(<32 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 11820 for instruction: %V64 = call i8 @llvm.vector.reduce.umax.v64i8(<64 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 36412 for instruction: %V128 = call i8 @llvm.vector.reduce.umax.v128i8(<128 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 12844 for instruction: %V64 = call i8 @llvm.vector.reduce.umax.v64i8(<64 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 41532 for instruction: %V128 = call i8 @llvm.vector.reduce.umax.v128i8(<128 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2   = call i8 @llvm.vector.reduce.umax.v2i8(<2 x i8> undef)
   %V4   = call i8 @llvm.vector.reduce.umax.v4i8(<4 x i8> undef)
   %V8   = call i8 @llvm.vector.reduce.umax.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.vector.reduce.umax.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.vector.reduce.umax.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.vector.reduce.umax.v64i8(<64 x i8> undef)

llvm/test/Analysis/CostModel/ARM/reduce-umin.ll

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 133 for instruction: %V16 = call i32 @llvm.vector.reduce.umin.v16i32(<16 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 237 for instruction: %V32 = call i32 @llvm.vector.reduce.umin.v32i32(<32 x i32> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i32'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i32 @llvm.vector.reduce.umin.v2i32(<2 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i32 @llvm.vector.reduce.umin.v4i32(<4 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 240 for instruction: %V8 = call i32 @llvm.vector.reduce.umin.v8i32(<8 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 632 for instruction: %V16 = call i32 @llvm.vector.reduce.umin.v16i32(<16 x i32> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2184 for instruction: %V32 = call i32 @llvm.vector.reduce.umin.v32i32(<32 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 696 for instruction: %V16 = call i32 @llvm.vector.reduce.umin.v16i32(<16 x i32> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2504 for instruction: %V32 = call i32 @llvm.vector.reduce.umin.v32i32(<32 x i32> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i32 @llvm.vector.reduce.umin.v2i32(<2 x i32> undef)
   %V4  = call i32 @llvm.vector.reduce.umin.v4i32(<4 x i32> undef)
   %V8  = call i32 @llvm.vector.reduce.umin.v8i32(<8 x i32> undef)
   %V16 = call i32 @llvm.vector.reduce.umin.v16i32(<16 x i32> undef)
   %V32 = call i32 @llvm.vector.reduce.umin.v32i32(<32 x i32> undef)
   ret i32 undef

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 503 for instruction: %V64 = call i16 @llvm.vector.reduce.umin.v64i16(<64 x i16> undef)
 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i16'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i16 @llvm.vector.reduce.umin.v2i16(<2 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i16 @llvm.vector.reduce.umin.v4i16(<4 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i16 @llvm.vector.reduce.umin.v8i16(<8 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 1176 for instruction: %V16 = call i16 @llvm.vector.reduce.umin.v16i16(<16 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 2720 for instruction: %V32 = call i16 @llvm.vector.reduce.umin.v32i16(<32 x i16> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 8880 for instruction: %V64 = call i16 @llvm.vector.reduce.umin.v64i16(<64 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 2976 for instruction: %V32 = call i16 @llvm.vector.reduce.umin.v32i16(<32 x i16> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 10160 for instruction: %V64 = call i16 @llvm.vector.reduce.umin.v64i16(<64 x i16> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2  = call i16 @llvm.vector.reduce.umin.v2i16(<2 x i16> undef)
   %V4  = call i16 @llvm.vector.reduce.umin.v4i16(<4 x i16> undef)
   %V8  = call i16 @llvm.vector.reduce.umin.v8i16(<8 x i16> undef)
   %V16 = call i16 @llvm.vector.reduce.umin.v16i16(<16 x i16> undef)
   %V32 = call i16 @llvm.vector.reduce.umin.v32i16(<32 x i16> undef)
   %V64 = call i16 @llvm.vector.reduce.umin.v64i16(<64 x i16> undef)

 ; NEON-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
 ; MVE-LABEL: 'reduce_i8'
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V2 = call i8 @llvm.vector.reduce.umin.v2i8(<2 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4 = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 788 for instruction: %V8 = call i8 @llvm.vector.reduce.umin.v8i8(<8 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 4128 for instruction: %V16 = call i8 @llvm.vector.reduce.umin.v16i8(<16 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 5668 for instruction: %V32 = call i8 @llvm.vector.reduce.umin.v32i8(<32 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 11820 for instruction: %V64 = call i8 @llvm.vector.reduce.umin.v64i8(<64 x i8> undef)
-; MVE-NEXT:  Cost Model: Found an estimated cost of 36412 for instruction: %V128 = call i8 @llvm.vector.reduce.umin.v128i8(<128 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 12844 for instruction: %V64 = call i8 @llvm.vector.reduce.umin.v64i8(<64 x i8> undef)
+; MVE-NEXT:  Cost Model: Found an estimated cost of 41532 for instruction: %V128 = call i8 @llvm.vector.reduce.umin.v128i8(<128 x i8> undef)
 ; MVE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
 ;
   %V2   = call i8 @llvm.vector.reduce.umin.v2i8(<2 x i8> undef)
   %V4   = call i8 @llvm.vector.reduce.umin.v4i8(<4 x i8> undef)
   %V8   = call i8 @llvm.vector.reduce.umin.v8i8(<8 x i8> undef)
   %V16  = call i8 @llvm.vector.reduce.umin.v16i8(<16 x i8> undef)
   %V32  = call i8 @llvm.vector.reduce.umin.v32i8(<32 x i8> undef)
   %V64  = call i8 @llvm.vector.reduce.umin.v64i8(<64 x i8> undef)

llvm/test/CodeGen/ARM/vselect_imax.ll

 define void @func_blend10(%T0_10* %loadaddr, %T0_10* %loadaddr2,
                            %T1_10* %blend, %T0_10* %storeaddr) {
   %v0 = load %T0_10, %T0_10* %loadaddr
   %v1 = load %T0_10, %T0_10* %loadaddr2
   %c = icmp slt %T0_10 %v0, %v1
 ; CHECK: vmin.s16
 ; CHECK: vmin.s16
 ; COST: func_blend10
-; COST: cost of 2 {{.*}} icmp
-; COST: cost of 2 {{.*}} select
+; COST: cost of 0 {{.*}} icmp
+; COST: cost of 4 {{.*}} select
   %r = select %T1_10 %c, %T0_10 %v0, %T0_10 %v1
   store %T0_10 %r, %T0_10* %storeaddr
   ret void
 }
 %T0_14 = type <8 x i32>
 %T1_14 = type <8 x i1>
 ; CHECK-LABEL: func_blend14:
 define void @func_blend14(%T0_14* %loadaddr, %T0_14* %loadaddr2,
                            %T1_14* %blend, %T0_14* %storeaddr) {
   %v0 = load %T0_14, %T0_14* %loadaddr
   %v1 = load %T0_14, %T0_14* %loadaddr2
   %c = icmp slt %T0_14 %v0, %v1
 ; CHECK: vmin.s32
 ; CHECK: vmin.s32
 ; COST: func_blend14
-; COST: cost of 2 {{.*}} icmp
-; COST: cost of 2 {{.*}} select
+; COST: cost of 0 {{.*}} icmp
+; COST: cost of 4 {{.*}} select
   %r = select %T1_14 %c, %T0_14 %v0, %T0_14 %v1
   store %T0_14 %r, %T0_14* %storeaddr
   ret void
 }
 %T0_15 = type <16 x i32>
 %T1_15 = type <16 x i1>
 ; CHECK-LABEL: func_blend15:
 define void @func_blend15(%T0_15* %loadaddr, %T0_15* %loadaddr2,
                            %T1_15* %blend, %T0_15* %storeaddr) {
 ; CHECK: vmin.s32
 ; CHECK: vmin.s32
   %v0 = load %T0_15, %T0_15* %loadaddr
   %v1 = load %T0_15, %T0_15* %loadaddr2
   %c = icmp slt %T0_15 %v0, %v1
 ; COST: func_blend15
-; COST: cost of 4 {{.*}} icmp
-; COST: cost of 4 {{.*}} select
+; COST: cost of 0 {{.*}} icmp
+; COST: cost of 8 {{.*}} select
   %r = select %T1_15 %c, %T0_15 %v0, %T0_15 %v1
   store %T0_15 %r, %T0_15* %storeaddr
   ret void
 }
 
 ; We adjusted the cost model of the following selects. When we improve code
 ; lowering we also need to adjust the cost.
 %T0_18 = type <4 x i64>

 ; CHECK-NEXT:    vst1.64 {d16, d17}, [r3:128]!
 ; CHECK-NEXT:    vst1.64 {d18, d19}, [r3:128]
 ; CHECK-NEXT:    pop {r4, r5, r6, r7, r11, lr}
 ; CHECK-NEXT:    mov pc, lr
   %v0 = load %T0_18, %T0_18* %loadaddr
   %v1 = load %T0_18, %T0_18* %loadaddr2
   %c = icmp slt %T0_18 %v0, %v1
 ; COST: func_blend18
-; COST: cost of 2 {{.*}} icmp
-; COST: cost of 19 {{.*}} select
+; COST: cost of 0 {{.*}} icmp
+; COST: cost of 21 {{.*}} select
   %r = select %T1_18 %c, %T0_18 %v0, %T0_18 %v1
   store %T0_18 %r, %T0_18* %storeaddr
   ret void
 }
 %T0_19 = type <8 x i64>
 %T1_19 = type <8 x i1>
 define void @func_blend19(%T0_19* %loadaddr, %T0_19* %loadaddr2,
                            %T1_19* %blend, %T0_19* %storeaddr) {

 ; CHECK-NEXT:    vst1.64 {d16, d17}, [r0:128]
 ; CHECK-NEXT:    vst1.64 {d20, d21}, [r3:128]
 ; CHECK-NEXT:    pop {r4, r5, r6, r7, r11, lr}
 ; CHECK-NEXT:    mov pc, lr
   %v0 = load %T0_19, %T0_19* %loadaddr
   %v1 = load %T0_19, %T0_19* %loadaddr2
   %c = icmp slt %T0_19 %v0, %v1
 ; COST: func_blend19
-; COST: cost of 4 {{.*}} icmp
-; COST: cost of 50 {{.*}} select
+; COST: cost of 0 {{.*}} icmp
+; COST: cost of 54 {{.*}} select
   %r = select %T1_19 %c, %T0_19 %v0, %T0_19 %v1
   store %T0_19 %r, %T0_19* %storeaddr
   ret void
 }
 %T0_20 = type <16 x i64>
 %T1_20 = type <16 x i1>
 define void @func_blend20(%T0_20* %loadaddr, %T0_20* %loadaddr2,
                            %T1_20* %blend, %T0_20* %storeaddr) {

 ; CHECK-NEXT:    vpop {d8, d9, d10, d11}
 ; CHECK-NEXT:    add sp, sp, #4
 ; CHECK-NEXT:    pop {r4, r5, r6, r7, r8, r9, r10, r11, lr}
 ; CHECK-NEXT:    mov pc, lr
   %v0 = load %T0_20, %T0_20* %loadaddr
   %v1 = load %T0_20, %T0_20* %loadaddr2
   %c = icmp slt %T0_20 %v0, %v1
 ; COST: func_blend20
-; COST: cost of 8 {{.*}} icmp
-; COST: cost of 100 {{.*}} select
+; COST: cost of 0 {{.*}} icmp
+; COST: cost of 108 {{.*}} select
   %r = select %T1_20 %c, %T0_20 %v0, %T0_20 %v1
   store %T0_20 %r, %T0_20* %storeaddr
   ret void
 }

llvm/test/Transforms/LoopVectorize/ARM/mve-icmpcost.ll

 ; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction:   %inc = add nuw nsw i32 %i.016, 1
 ; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction:   %exitcond.not = icmp eq i32 %inc, %n
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
 ; CHECK: LV: Vector loop of width 4 costs: 2.
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %i.016 = phi i32 [ 0, %for.body.lr.ph ], [ %inc, %for.inc ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %arrayidx = getelementptr inbounds i16, i16* %s, i32 %i.016
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %1 = load i16, i16* %arrayidx, align 2
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %conv = sext i16 %1 to i32
-; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction:   %cmp2 = icmp sgt i32 %conv, %conv1
+; CHECK: LV: Found an estimated cost of 68 for VF 8 For instruction:   %cmp2 = icmp sgt i32 %conv, %conv1
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   br i1 %cmp2, label %if.then, label %for.inc
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %conv6 = add i16 %1, %0
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %arrayidx7 = getelementptr inbounds i16, i16* %d, i32 %i.016
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   store i16 %conv6, i16* %arrayidx7, align 2
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   br label %for.inc
 ; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction:   %inc = add nuw nsw i32 %i.016, 1
 ; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction:   %exitcond.not = icmp eq i32 %inc, %n
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   br i1 %exitcond.not, label %for.cond.cleanup.loopexit, label %for.body
-; CHECK: LV: Vector loop of width 8 costs: 1.
-; CHECK: LV: Selecting VF: 8.
+; CHECK: LV: Vector loop of width 8 costs: 9.
+; CHECK: LV: Selecting VF: 4.
 define void @expensive_icmp(i16* noalias nocapture %d, i16* nocapture readonly %s, i32 %n, i16 zeroext %m) #0 {
 entry:
   %cmp15 = icmp sgt i32 %n, 0
   br i1 %cmp15, label %for.body.lr.ph, label %for.cond.cleanup
 
 for.body.lr.ph:                                   ; preds = %entry
   %conv1 = zext i16 %m to i32
   %0 = trunc i32 %n to i16

 ; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
 ; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction:   %0 = load i8, i8* %pSrcA.addr.011, align 1
 ; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction:   %conv1 = sext i8 %0 to i32
 ; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
 ; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction:   %1 = load i8, i8* %pSrcB.addr.09, align 1
 ; CHECK: LV: Found an estimated cost of 4 for VF 2 For instruction:   %conv3 = sext i8 %1 to i32
 ; CHECK: LV: Found an estimated cost of 14 for VF 2 For instruction:   %mul = mul nsw i32 %conv3, %conv1
 ; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction:   %shr = ashr i32 %mul, 7
-; CHECK: LV: Found an estimated cost of 12 for VF 2 For instruction:   %2 = icmp slt i32 %shr, 127
-; CHECK: LV: Found an estimated cost of 12 for VF 2 For instruction:   %spec.select.i = select i1 %2, i32 %shr, i32 127
+; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 24 for VF 2 For instruction:   %spec.select.i = select i1 %2, i32 %shr, i32 127
 ; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   %conv4 = trunc i32 %spec.select.i to i8
 ; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
 ; CHECK: LV: Found an estimated cost of 10 for VF 2 For instruction:   store i8 %conv4, i8* %pDst.addr.010, align 1
 ; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction:   %dec = add i32 %blkCnt.012, -1
 ; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction:   %cmp.not = icmp eq i32 %dec, 0
 ; CHECK: LV: Found an estimated cost of 0 for VF 2 For instruction:   br i1 %cmp.not, label %while.end.loopexit, label %while.body
 ; CHECK: LV: Vector loop of width 2 costs: 44.
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction:   %0 = load i8, i8* %pSrcA.addr.011, align 1
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %conv1 = sext i8 %0 to i32
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction:   %1 = load i8, i8* %pSrcB.addr.09, align 1
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %conv3 = sext i8 %1 to i32
 ; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction:   %mul = mul nsw i32 %conv3, %conv1
 ; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction:   %shr = ashr i32 %mul, 7
-; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction:   %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %2 = icmp slt i32 %shr, 127
 ; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction:   %spec.select.i = select i1 %2, i32 %shr, i32 127
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %conv4 = trunc i32 %spec.select.i to i8
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 4 For instruction:   store i8 %conv4, i8* %pDst.addr.010, align 1
 ; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction:   %dec = add i32 %blkCnt.012, -1
 ; CHECK: LV: Found an estimated cost of 1 for VF 4 For instruction:   %cmp.not = icmp eq i32 %dec, 0
 ; CHECK: LV: Found an estimated cost of 0 for VF 4 For instruction:   br i1 %cmp.not, label %while.end.loopexit, label %while.body
-; CHECK: LV: Vector loop of width 4 costs: 4.
+; CHECK: LV: Vector loop of width 4 costs: 3.
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %0 = load i8, i8* %pSrcA.addr.011, align 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %conv1 = sext i8 %0 to i32
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %1 = load i8, i8* %pSrcB.addr.09, align 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %conv3 = sext i8 %1 to i32
 ; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction:   %mul = mul nsw i32 %conv3, %conv1
 ; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction:   %shr = ashr i32 %mul, 7
-; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction:   %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %2 = icmp slt i32 %shr, 127
 ; CHECK: LV: Found an estimated cost of 4 for VF 8 For instruction:   %spec.select.i = select i1 %2, i32 %shr, i32 127
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   %conv4 = trunc i32 %spec.select.i to i8
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 8 For instruction:   store i8 %conv4, i8* %pDst.addr.010, align 1
 ; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction:   %dec = add i32 %blkCnt.012, -1
 ; CHECK: LV: Found an estimated cost of 1 for VF 8 For instruction:   %cmp.not = icmp eq i32 %dec, 0
 ; CHECK: LV: Found an estimated cost of 0 for VF 8 For instruction:   br i1 %cmp.not, label %while.end.loopexit, label %while.body
 ; CHECK: LV: Vector loop of width 8 costs: 3.
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %blkCnt.012 = phi i32 [ %dec, %while.body ], [ %blockSize, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %pSrcA.addr.011 = phi i8* [ %incdec.ptr, %while.body ], [ %pSrcA, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %pDst.addr.010 = phi i8* [ %incdec.ptr5, %while.body ], [ %pDst, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %pSrcB.addr.09 = phi i8* [ %incdec.ptr2, %while.body ], [ %pSrcB, %while.body.preheader ]
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %incdec.ptr = getelementptr inbounds i8, i8* %pSrcA.addr.011, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction:   %0 = load i8, i8* %pSrcA.addr.011, align 1
 ; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction:   %conv1 = sext i8 %0 to i32
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %incdec.ptr2 = getelementptr inbounds i8, i8* %pSrcB.addr.09, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction:   %1 = load i8, i8* %pSrcB.addr.09, align 1
 ; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction:   %conv3 = sext i8 %1 to i32
 ; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction:   %mul = mul nsw i32 %conv3, %conv1
 ; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction:   %shr = ashr i32 %mul, 7
-; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction:   %2 = icmp slt i32 %shr, 127
+; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %2 = icmp slt i32 %shr, 127
 ; CHECK: LV: Found an estimated cost of 8 for VF 16 For instruction:   %spec.select.i = select i1 %2, i32 %shr, i32 127
 ; CHECK: LV: Found an estimated cost of 6 for VF 16 For instruction:   %conv4 = trunc i32 %spec.select.i to i8
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   %incdec.ptr5 = getelementptr inbounds i8, i8* %pDst.addr.010, i32 1
 ; CHECK: LV: Found an estimated cost of 2 for VF 16 For instruction:   store i8 %conv4, i8* %pDst.addr.010, align 1
 ; CHECK: LV: Found an estimated cost of 1 for VF 16 For instruction:   %dec = add i32 %blkCnt.012, -1
 ; CHECK: LV: Found an estimated cost of 1 for VF 16 For instruction:   %cmp.not = icmp eq i32 %dec, 0
 ; CHECK: LV: Found an estimated cost of 0 for VF 16 For instruction:   br i1 %cmp.not, label %while.end.loopexit, label %while.body
 ; CHECK: LV: Vector loop of width 16 costs: 3.