[llvm] [AArch64] Refactor redundant PTEST optimisations (NFC) (PR #87802)

Momchil Velikov via llvm-commits llvm-commits at lists.llvm.org
Mon Jun 3 07:57:34 PDT 2024


https://github.com/momchil-velikov updated https://github.com/llvm/llvm-project/pull/87802

>From ee468ec26ba3ae15182623687d26c888d35b6a3f Mon Sep 17 00:00:00 2001
From: Momchil Velikov <momchil.velikov at arm.com>
Date: Wed, 27 Dec 2023 14:47:06 +0000
Subject: [PATCH 1/2] [LoopVectorize][AArch64][SVE] Generate wide active lane
 masks

This patch makes the LoopVectorize generate lane masks longer than
the VF to allow the target to better utilise the instruction set.
The vectorizer emit one or more wide `llvm.get.active.lane.mask.*`
calls plus several `llvm.vector.extract.*` calls to yield the required
number of VF-wide masks.

The motivating exammple is a vectorised loop with unroll factor 2 that
can use the SVE2.1 `whilelo` instruction with predicate pair result, or
a SVE `whilelo` instruction with smaller element size plus
`punpklo`/`punpkhi`.

How wide is the lane mask that the vectoriser emits is controlled
by a TargetTransformInfo hook `getMaxPredicateLength`.The default
impementation (return the same length as the VF) keeps the
change non-functional for targets that can't or are not prepared
to handle wider lane masks.
---
 .../llvm/Analysis/TargetTransformInfo.h       |   10 +
 .../llvm/Analysis/TargetTransformInfoImpl.h   |    2 +
 llvm/include/llvm/CodeGen/BasicTTIImpl.h      |    2 +
 llvm/lib/Analysis/TargetTransformInfo.cpp     |    4 +
 .../AArch64/AArch64TargetTransformInfo.cpp    |    9 +
 .../AArch64/AArch64TargetTransformInfo.h      |    2 +
 .../Vectorize/LoopVectorizationPlanner.h      |    8 +
 .../Transforms/Vectorize/LoopVectorize.cpp    |    7 +-
 llvm/lib/Transforms/Vectorize/VPlan.cpp       |    7 +-
 llvm/lib/Transforms/Vectorize/VPlan.h         |   50 +-
 .../Transforms/Vectorize/VPlanPatternMatch.h  |   34 +-
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp |  107 +-
 .../Transforms/Vectorize/VPlanTransforms.cpp  |   18 +-
 llvm/lib/Transforms/Vectorize/VPlanValue.h    |    1 +
 .../CodeGen/AArch64/sve-wide-lane-mask.ll     | 1069 ++++++++++
 .../LoopVectorize/AArch64/masked-call.ll      |  359 ++--
 .../LoopVectorize/AArch64/pr73894.ll          |    4 +-
 .../AArch64/scalable-strict-fadd.ll           | 1788 +++++++++--------
 .../AArch64/sve-tail-folding-unroll.ll        |  322 +--
 .../AArch64/sve-wide-lane-mask.ll             |  656 ++++++
 .../AArch64/uniform-args-call-variants.ll     |  102 +-
 .../ARM/tail-folding-prefer-flag.ll           |    6 +-
 .../strict-fadd-interleave-only.ll            |    4 +-
 23 files changed, 3235 insertions(+), 1336 deletions(-)
 create mode 100644 llvm/test/CodeGen/AArch64/sve-wide-lane-mask.ll
 create mode 100644 llvm/test/Transforms/LoopVectorize/AArch64/sve-wide-lane-mask.ll

diff --git a/llvm/include/llvm/Analysis/TargetTransformInfo.h b/llvm/include/llvm/Analysis/TargetTransformInfo.h
index cefce93f9e25c..79c386384d5c4 100644
--- a/llvm/include/llvm/Analysis/TargetTransformInfo.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfo.h
@@ -1244,6 +1244,8 @@ class TargetTransformInfo {
   /// and the number of execution units in the CPU.
   unsigned getMaxInterleaveFactor(ElementCount VF) const;
 
+  ElementCount getMaxPredicateLength(ElementCount VF) const;
+
   /// Collect properties of V used in cost analysis, e.g. OP_PowerOf2.
   static OperandValueInfo getOperandInfo(const Value *V);
 
@@ -2002,6 +2004,9 @@ class TargetTransformInfo::Concept {
   virtual bool shouldPrefetchAddressSpace(unsigned AS) const = 0;
 
   virtual unsigned getMaxInterleaveFactor(ElementCount VF) = 0;
+
+  virtual ElementCount getMaxPredicateLength(ElementCount VF) const = 0;
+
   virtual InstructionCost getArithmeticInstrCost(
       unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
       OperandValueInfo Opd1Info, OperandValueInfo Opd2Info,
@@ -2627,6 +2632,11 @@ class TargetTransformInfo::Model final : public TargetTransformInfo::Concept {
   unsigned getMaxInterleaveFactor(ElementCount VF) override {
     return Impl.getMaxInterleaveFactor(VF);
   }
+
+  ElementCount getMaxPredicateLength(ElementCount VF) const override {
+    return Impl.getMaxPredicateLength(VF);
+  }
+
   unsigned getEstimatedNumberOfCaseClusters(const SwitchInst &SI,
                                             unsigned &JTSize,
                                             ProfileSummaryInfo *PSI,
diff --git a/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h b/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
index 9a57331d281db..44812222bb57e 100644
--- a/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
+++ b/llvm/include/llvm/Analysis/TargetTransformInfoImpl.h
@@ -537,6 +537,8 @@ class TargetTransformInfoImplBase {
 
   unsigned getMaxInterleaveFactor(ElementCount VF) const { return 1; }
 
+  ElementCount getMaxPredicateLength(ElementCount VF) const { return VF; }
+
   InstructionCost getArithmeticInstrCost(
       unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
       TTI::OperandValueInfo Opd1Info, TTI::OperandValueInfo Opd2Info,
diff --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
index 2091432d4fe27..c2d2db250ef21 100644
--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -890,6 +890,8 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
 
   unsigned getMaxInterleaveFactor(ElementCount VF) { return 1; }
 
+  ElementCount getMaxPredicateLength(ElementCount VF) const { return VF; }
+
   InstructionCost getArithmeticInstrCost(
       unsigned Opcode, Type *Ty, TTI::TargetCostKind CostKind,
       TTI::OperandValueInfo Opd1Info = {TTI::OK_AnyValue, TTI::OP_None},
diff --git a/llvm/lib/Analysis/TargetTransformInfo.cpp b/llvm/lib/Analysis/TargetTransformInfo.cpp
index 82b6d7e7c4833..383bc2fcb2a99 100644
--- a/llvm/lib/Analysis/TargetTransformInfo.cpp
+++ b/llvm/lib/Analysis/TargetTransformInfo.cpp
@@ -821,6 +821,10 @@ unsigned TargetTransformInfo::getMaxInterleaveFactor(ElementCount VF) const {
   return TTIImpl->getMaxInterleaveFactor(VF);
 }
 
+ElementCount TargetTransformInfo::getMaxPredicateLength(ElementCount VF) const {
+  return TTIImpl->getMaxPredicateLength(VF);
+}
+
 TargetTransformInfo::OperandValueInfo
 TargetTransformInfo::getOperandInfo(const Value *V) {
   OperandValueKind OpInfo = OK_AnyValue;
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index f49c73dc79519..ff5036538589c 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -3383,6 +3383,15 @@ unsigned AArch64TTIImpl::getMaxInterleaveFactor(ElementCount VF) {
   return ST->getMaxInterleaveFactor();
 }
 
+ElementCount AArch64TTIImpl::getMaxPredicateLength(ElementCount VF) const {
+  // Do not create masks bigger than `<vscale x 16 x i1>`.
+  unsigned N = ST->hasSVE() ? 16 : 0;
+  // Do not create masks that are more than twice the VF.
+  N = std::min(N, 2 * VF.getKnownMinValue());
+  return VF.isScalable() ? ElementCount::getScalable(N)
+                         : ElementCount::getFixed(N);
+}
+
 // For Falkor, we want to avoid having too many strided loads in a loop since
 // that can exhaust the HW prefetcher resources.  We adjust the unroller
 // MaxCount preference below to attempt to ensure unrolling doesn't create too
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
index 2f44aaa3e26ab..6f2af4eaf40d6 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.h
@@ -157,6 +157,8 @@ class AArch64TTIImpl : public BasicTTIImplBase<AArch64TTIImpl> {
 
   unsigned getMaxInterleaveFactor(ElementCount VF);
 
+  ElementCount getMaxPredicateLength(ElementCount VF) const;
+
   bool prefersVectorizedAddressing() const;
 
   InstructionCost getMaskedMemoryOpCost(unsigned Opcode, Type *Src,
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
index c03c278fcebe7..f07bdea90c0f7 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
@@ -202,6 +202,14 @@ class VPBuilder {
   VPValue *createICmp(CmpInst::Predicate Pred, VPValue *A, VPValue *B,
                       DebugLoc DL = {}, const Twine &Name = "");
 
+  VPValue *createGetActiveLaneMask(VPValue *IV, VPValue *TC, DebugLoc DL,
+                                   const Twine &Name = "") {
+    auto *ALM = new VPActiveLaneMaskRecipe(IV, TC, DL, Name);
+    if (BB)
+      BB->insert(ALM, InsertPt);
+    return ALM;
+  }
+
   //===--------------------------------------------------------------------===//
   // RAII helpers.
   //===--------------------------------------------------------------------===//
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index a9ee9f62197e1..cb8786fe32b44 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -588,6 +588,10 @@ class InnerLoopVectorizer {
   /// count of the original loop for both main loop and epilogue vectorization.
   void setTripCount(Value *TC) { TripCount = TC; }
 
+  ElementCount getMaxPredicateLength(ElementCount VF) const {
+    return TTI->getMaxPredicateLength(VF);
+  }
+
 protected:
   friend class LoopVectorizationPlanner;
 
@@ -7509,7 +7513,8 @@ LoopVectorizationPlanner::executePlan(
   LLVM_DEBUG(BestVPlan.dump());
 
   // Perform the actual loop transformation.
-  VPTransformState State(BestVF, BestUF, LI, DT, ILV.Builder, &ILV, &BestVPlan,
+  VPTransformState State(BestVF, BestUF, TTI.getMaxPredicateLength(BestVF), LI,
+                         DT, ILV.Builder, &ILV, &BestVPlan,
                          OrigLoop->getHeader()->getContext());
 
   // 0. Generate SCEV-dependent code into the preheader, including TripCount,
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp
index f17be451e6846..dfe25f54130f2 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp
@@ -215,12 +215,13 @@ VPBasicBlock::iterator VPBasicBlock::getFirstNonPhi() {
   return It;
 }
 
-VPTransformState::VPTransformState(ElementCount VF, unsigned UF, LoopInfo *LI,
+VPTransformState::VPTransformState(ElementCount VF, unsigned UF,
+                                   ElementCount MaxPred, LoopInfo *LI,
                                    DominatorTree *DT, IRBuilderBase &Builder,
                                    InnerLoopVectorizer *ILV, VPlan *Plan,
                                    LLVMContext &Ctx)
-    : VF(VF), UF(UF), CFG(DT), LI(LI), Builder(Builder), ILV(ILV), Plan(Plan),
-      LVer(nullptr),
+    : VF(VF), UF(UF), MaxPred(MaxPred), CFG(DT), LI(LI), Builder(Builder),
+      ILV(ILV), Plan(Plan), LVer(nullptr),
       TypeAnalysis(Plan->getCanonicalIV()->getScalarType(), Ctx) {}
 
 Value *VPTransformState::get(VPValue *Def, const VPIteration &Instance) {
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index d6b63a5d43c46..06f77433af6b8 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -235,13 +235,14 @@ struct VPIteration {
 /// VPTransformState holds information passed down when "executing" a VPlan,
 /// needed for generating the output IR.
 struct VPTransformState {
-  VPTransformState(ElementCount VF, unsigned UF, LoopInfo *LI,
-                   DominatorTree *DT, IRBuilderBase &Builder,
+  VPTransformState(ElementCount VF, unsigned UF, ElementCount MaxPred,
+                   LoopInfo *LI, DominatorTree *DT, IRBuilderBase &Builder,
                    InnerLoopVectorizer *ILV, VPlan *Plan, LLVMContext &Ctx);
 
   /// The chosen Vectorization and Unroll Factors of the loop being vectorized.
   ElementCount VF;
   unsigned UF;
+  ElementCount MaxPred;
 
   /// Hold the indices to generate specific scalar instructions. Null indicates
   /// that all instances are to be generated, using either scalar or vector
@@ -1174,7 +1175,6 @@ class VPInstruction : public VPRecipeWithIRFlags {
     Not,
     SLPLoad,
     SLPStore,
-    ActiveLaneMask,
     ExplicitVectorLength,
     CalculateTripCountMinusVF,
     // Increment the canonical IV separately for each unrolled part.
@@ -1329,6 +1329,50 @@ class VPInstruction : public VPRecipeWithIRFlags {
   }
 };
 
+class VPActiveLaneMaskRecipe : public VPRecipeWithIRFlags {
+  const std::string Name;
+
+public:
+  VPActiveLaneMaskRecipe(VPValue *IV, VPValue *TC, DebugLoc DL = {},
+                         const Twine &Name = "")
+      : VPRecipeWithIRFlags(VPDef::VPActiveLaneMaskSC,
+                            std::initializer_list<VPValue *>{IV, TC}, DL),
+        Name(Name.str()) {}
+
+  VP_CLASSOF_IMPL(VPDef::VPActiveLaneMaskSC)
+
+  VPActiveLaneMaskRecipe *clone() override {
+    SmallVector<VPValue *, 2> Operands(operands());
+    assert(Operands.size() == 2 && "by construction");
+    auto *New = new VPActiveLaneMaskRecipe(Operands[0], Operands[1],
+                                           getDebugLoc(), Name);
+    New->transferFlags(*this);
+    return New;
+  }
+
+  void execute(VPTransformState &State) override;
+
+#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+  /// Print the recipe.
+  void print(raw_ostream &O, const Twine &Indent,
+             VPSlotTracker &SlotTracker) const override;
+#endif
+
+  bool onlyFirstLaneUsed(const VPValue *Op) const override {
+    assert(is_contained(operands(), Op) &&
+           "Op must be an operand of the recipe");
+
+    return true;
+  }
+
+  bool onlyFirstPartUsed(const VPValue *Op) const override {
+    assert(is_contained(operands(), Op) &&
+           "Op must be an operand of the recipe");
+
+    return false;
+  }
+};
+
 /// VPWidenRecipe is a recipe for producing a copy of vector type its
 /// ingredient. This recipe covers most of the traditional vectorization cases
 /// where each ingredient transforms into a vectorized version of itself.
diff --git a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
index 0587468807435..5cd18ea7ba2c7 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
@@ -221,11 +221,6 @@ m_BranchOnCond(const Op0_t &Op0) {
   return m_VPInstruction<VPInstruction::BranchOnCond>(Op0);
 }
 
-template <typename Op0_t, typename Op1_t>
-inline BinaryVPInstruction_match<Op0_t, Op1_t, VPInstruction::ActiveLaneMask>
-m_ActiveLaneMask(const Op0_t &Op0, const Op1_t &Op1) {
-  return m_VPInstruction<VPInstruction::ActiveLaneMask>(Op0, Op1);
-}
 
 template <typename Op0_t, typename Op1_t>
 inline BinaryVPInstruction_match<Op0_t, Op1_t, VPInstruction::BranchOnCount>
@@ -296,6 +291,35 @@ inline BinaryVPInstruction_match<Op0_t, Op1_t, VPInstruction::LogicalAnd>
 m_LogicalAnd(const Op0_t &Op0, const Op1_t &Op1) {
   return m_VPInstruction<VPInstruction::LogicalAnd, Op0_t, Op1_t>(Op0, Op1);
 }
+
+template <typename Op0_t, typename Op1_t>
+struct VPActiveLaneMask_match {
+  Op0_t Op0;
+  Op1_t Op1;
+
+  VPActiveLaneMask_match(Op0_t Op0, Op1_t Op1) : Op0(Op0), Op1(Op1) {}
+
+  bool match(const VPValue *V) {
+    auto *DefR = V->getDefiningRecipe();
+    return DefR && match(DefR);
+  }
+
+  bool match(const VPRecipeBase *R) {
+    auto *DefR = dyn_cast<VPActiveLaneMaskRecipe>(R);
+    if (!DefR)
+      return false;
+    assert(DefR->getNumOperands() == 2 &&
+           "recipe with matched opcode does not have 2 operands");
+    return Op0.match(DefR->getOperand(0)) && Op1.match(DefR->getOperand(1));
+  }
+};
+
+template <typename Op0_t, typename Op1_t>
+inline VPActiveLaneMask_match<Op0_t, Op1_t>
+m_ActiveLaneMask(const Op0_t &Op0, const Op1_t &Op1) {
+  return {Op0, Op1};
+}
+
 } // namespace VPlanPatternMatch
 } // namespace llvm
 
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 5eb99ffd1e10e..a648fb99dd2d1 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -351,24 +351,7 @@ Value *VPInstruction::generatePerPart(VPTransformState &State, unsigned Part) {
     Value *Op2 = State.get(getOperand(2), Part);
     return Builder.CreateSelect(Cond, Op1, Op2, Name);
   }
-  case VPInstruction::ActiveLaneMask: {
-    // Get first lane of vector induction variable.
-    Value *VIVElem0 = State.get(getOperand(0), VPIteration(Part, 0));
-    // Get the original loop tripcount.
-    Value *ScalarTC = State.get(getOperand(1), VPIteration(Part, 0));
 
-    // If this part of the active lane mask is scalar, generate the CMP directly
-    // to avoid unnecessary extracts.
-    if (State.VF.isScalar())
-      return Builder.CreateCmp(CmpInst::Predicate::ICMP_ULT, VIVElem0, ScalarTC,
-                               Name);
-
-    auto *Int1Ty = Type::getInt1Ty(Builder.getContext());
-    auto *PredTy = VectorType::get(Int1Ty, State.VF);
-    return Builder.CreateIntrinsic(Intrinsic::get_active_lane_mask,
-                                   {PredTy, ScalarTC->getType()},
-                                   {VIVElem0, ScalarTC}, nullptr, Name);
-  }
   case VPInstruction::FirstOrderRecurrenceSplice: {
     // Generate code to combine the previous and current values in vector v3.
     //
@@ -636,7 +619,6 @@ bool VPInstruction::onlyFirstLaneUsed(const VPValue *Op) const {
   case VPInstruction::PtrAdd:
     // TODO: Cover additional opcodes.
     return vputils::onlyFirstLaneUsed(this);
-  case VPInstruction::ActiveLaneMask:
   case VPInstruction::ExplicitVectorLength:
   case VPInstruction::CalculateTripCountMinusVF:
   case VPInstruction::CanonicalIVIncrementForPart:
@@ -671,9 +653,6 @@ void VPInstruction::print(raw_ostream &O, const Twine &Indent,
   case VPInstruction::SLPStore:
     O << "combined store";
     break;
-  case VPInstruction::ActiveLaneMask:
-    O << "active lane mask";
-    break;
   case VPInstruction::ExplicitVectorLength:
     O << "EXPLICIT-VECTOR-LENGTH";
     break;
@@ -713,8 +692,94 @@ void VPInstruction::print(raw_ostream &O, const Twine &Indent,
     DL.print(O);
   }
 }
+
+void VPActiveLaneMaskRecipe::print(raw_ostream &O, const Twine &Indent,
+                                   VPSlotTracker &SlotTracker) const {
+  O << Indent << "EMIT ";
+
+  printAsOperand(O, SlotTracker);
+  O << " = active lane mask";
+  printFlags(O);
+  printOperands(O, SlotTracker);
+
+  if (auto DL = getDebugLoc()) {
+    O << ", !dbg ";
+    DL.print(O);
+  }
+}
+
 #endif
 
+void VPActiveLaneMaskRecipe::execute(VPTransformState &State) {
+  assert(!State.Instance && "VPInstruction executing an Instance");
+
+  IRBuilderBase &Builder = State.Builder;
+  Builder.SetCurrentDebugLocation(getDebugLoc());
+
+  // If this the active lane mask is scalar, generate the CMP directly
+  // to avoid unnecessary extracts.
+  if (State.VF.isScalar()) {
+    for (int Part = State.UF - 1; Part >= 0; --Part) {
+      // Get first lane of vector induction variable.
+      Value *VIVElem0 = State.get(getOperand(0), VPIteration(Part, 0));
+      // Get the original loop tripcount.
+      Value *ScalarTC = State.get(getOperand(1), VPIteration(0, 0));
+
+      Value *V = Builder.CreateCmp(CmpInst::Predicate::ICMP_ULT, VIVElem0,
+                                   ScalarTC, Name);
+      State.set(this, V, Part);
+    }
+    return;
+  }
+
+  auto *Int1Ty = Type::getInt1Ty(Builder.getContext());
+  auto *PredTy = VectorType::get(Int1Ty, State.VF);
+
+  unsigned MaxPred = std::min(State.MaxPred.getKnownMinValue(),
+                              State.UF * State.VF.getKnownMinValue());
+  if (State.UF <= 1 || MaxPred <= State.VF.getKnownMinValue() ||
+      MaxPred % State.VF.getKnownMinValue() != 0) {
+    for (int Part = State.UF - 1; Part >= 0; --Part) {
+      // Get first lane of vector induction variable.
+      Value *VIVElem0 = State.get(getOperand(0), VPIteration(Part, 0));
+      // Get the original loop tripcount.
+      Value *ScalarTC = State.get(getOperand(1), VPIteration(0, 0));
+      Value *V = Builder.CreateIntrinsic(Intrinsic::get_active_lane_mask,
+                                         {PredTy, ScalarTC->getType()},
+                                         {VIVElem0, ScalarTC}, nullptr, Name);
+      State.set(this, V, Part);
+    }
+    return;
+  }
+
+  // Generate long active lane masks covering all the unrolled iterations.
+  unsigned PartsPerMask = MaxPred / State.VF.getKnownMinValue();
+  auto *LongPredTy = VectorType::get(Int1Ty, MaxPred, State.VF.isScalable());
+  SmallVector<Value *> LongMask(State.UF / PartsPerMask, nullptr);
+  for (int Part = State.UF - PartsPerMask; Part >= 0; Part -= PartsPerMask) {
+    // Get first lane of vector induction variable.
+    Value *VIVElem0 = State.get(getOperand(0), VPIteration(Part, 0));
+    // Get the original loop tripcount.
+    Value *ScalarTC = State.get(getOperand(1), VPIteration(0, 0));
+    Value *V = Builder.CreateIntrinsic(Intrinsic::get_active_lane_mask,
+                                       {LongPredTy, ScalarTC->getType()},
+                                       {VIVElem0, ScalarTC}, nullptr, Name);
+    LongMask[Part / PartsPerMask] = V;
+  }
+
+  for (int Part = State.UF - 1; Part >= 0; --Part) {
+    Value *ALM = LongMask[Part / PartsPerMask];
+    const unsigned I = Part % PartsPerMask;
+    Value *V = Builder.CreateIntrinsic(
+        Intrinsic::vector_extract, {PredTy, ALM->getType()},
+        {ALM, ConstantInt::get(Type::getInt64Ty(Builder.getContext()),
+                               I * State.VF.getKnownMinValue())},
+        nullptr, Name);
+
+    State.set(this, V, Part);
+  }
+}
+
 void VPWidenCallRecipe::execute(VPTransformState &State) {
   assert(State.VF.isVector() && "not widening");
   Function *CalledScalarFn = getCalledScalarFunction();
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 422579ea8b84f..061956f0e30ec 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1186,9 +1186,8 @@ static VPActiveLaneMaskPHIRecipe *addVPLaneMaskPhiAndUpdateExitBranch(
       "index.part.next");
 
   // Create the active lane mask instruction in the VPlan preheader.
-  auto *EntryALM =
-      Builder.createNaryOp(VPInstruction::ActiveLaneMask, {EntryIncrement, TC},
-                           DL, "active.lane.mask.entry");
+  auto *EntryALM = Builder.createGetActiveLaneMask(EntryIncrement, TC, DL,
+                                                   "active.lane.mask.entry");
 
   // Now create the ActiveLaneMaskPhi recipe in the main loop using the
   // preheader ActiveLaneMask instruction.
@@ -1202,9 +1201,8 @@ static VPActiveLaneMaskPHIRecipe *addVPLaneMaskPhiAndUpdateExitBranch(
   auto *InLoopIncrement =
       Builder.createOverflowingOp(VPInstruction::CanonicalIVIncrementForPart,
                                   {IncrementValue}, {false, false}, DL);
-  auto *ALM = Builder.createNaryOp(VPInstruction::ActiveLaneMask,
-                                   {InLoopIncrement, TripCount}, DL,
-                                   "active.lane.mask.next");
+  auto *ALM = Builder.createGetActiveLaneMask(InLoopIncrement, TripCount, DL,
+                                              "active.lane.mask.next");
   LaneMaskPhi->addOperand(ALM);
 
   // Replace the original terminator with BranchOnCond. We have to invert the
@@ -1278,15 +1276,15 @@ void VPlanTransforms::addActiveLaneMask(
          "Must have widened canonical IV when tail folding!");
   auto *WideCanonicalIV =
       cast<VPWidenCanonicalIVRecipe>(*FoundWidenCanonicalIVUser);
-  VPSingleDefRecipe *LaneMask;
+  VPValue *LaneMask;
   if (UseActiveLaneMaskForControlFlow) {
     LaneMask = addVPLaneMaskPhiAndUpdateExitBranch(
         Plan, DataAndControlFlowWithoutRuntimeCheck);
   } else {
     VPBuilder B = VPBuilder::getToInsertAfter(WideCanonicalIV);
-    LaneMask = B.createNaryOp(VPInstruction::ActiveLaneMask,
-                              {WideCanonicalIV, Plan.getTripCount()}, nullptr,
-                              "active.lane.mask");
+    LaneMask = B.createGetActiveLaneMask(WideCanonicalIV, Plan.getTripCount(),
+                                         WideCanonicalIV->getDebugLoc(),
+                                         "active.lane.mask");
   }
 
   // Walk users of WideCanonicalIV and replace all compares of the form
diff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h
index 8d945f6f2b8ea..b90070579e5e8 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h
@@ -358,6 +358,7 @@ class VPDef {
     VPWidenSC,
     VPWidenSelectSC,
     VPBlendSC,
+    VPActiveLaneMaskSC,
     // START: Phi-like recipes. Need to be kept together.
     VPWidenPHISC,
     VPPredInstPHISC,
diff --git a/llvm/test/CodeGen/AArch64/sve-wide-lane-mask.ll b/llvm/test/CodeGen/AArch64/sve-wide-lane-mask.ll
new file mode 100644
index 0000000000000..866185cac6a0a
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve-wide-lane-mask.ll
@@ -0,0 +1,1069 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mattr=+sve    < %s | FileCheck %s -check-prefix CHECK-SVE
+; RUN: llc -mattr=+sve2p1 < %s | FileCheck %s -check-prefix CHECK-SVE2p1
+
+target triple = "aarch64-unknown-linux"
+
+; Byte elements, UF=0
+define void @f_b_0(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_b_0:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB0_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    mov w9, w2
+; CHECK-SVE-NEXT:    rdvl x10, #1
+; CHECK-SVE-NEXT:    mov x8, xzr
+; CHECK-SVE-NEXT:    whilelo p0.b, xzr, x9
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB0_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1b { z0.b }, p0/z, [x1, x8]
+; CHECK-SVE-NEXT:    mul z0.b, z0.b, #3
+; CHECK-SVE-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; CHECK-SVE-NEXT:    add x8, x8, x10
+; CHECK-SVE-NEXT:    whilelo p0.b, x8, x9
+; CHECK-SVE-NEXT:    b.mi .LBB0_2
+; CHECK-SVE-NEXT:  .LBB0_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_b_0:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB0_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    mov w9, w2
+; CHECK-SVE2p1-NEXT:    mov x8, xzr
+; CHECK-SVE2p1-NEXT:    whilelo p0.b, xzr, x9
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB0_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1b { z0.b }, p0/z, [x1, x8]
+; CHECK-SVE2p1-NEXT:    mul z0.b, z0.b, #3
+; CHECK-SVE2p1-NEXT:    st1b { z0.b }, p0, [x0, x8]
+; CHECK-SVE2p1-NEXT:    addvl x8, x8, #1
+; CHECK-SVE2p1-NEXT:    whilelo p0.b, x8, x9
+; CHECK-SVE2p1-NEXT:    b.mi .LBB0_2
+; CHECK-SVE2p1-NEXT:  .LBB0_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp7 = icmp sgt i32 %n, 0
+  br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 4
+  %active.lane.mask.entry = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 %wide.trip.count)
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next, %vector.body ]
+  %active.lane.mask = phi <vscale x 16 x i1> [ %active.lane.mask.entry, %for.body.preheader ], [ %active.lane.mask.next, %vector.body ]
+  %2 = getelementptr inbounds i8, ptr %src, i64 %index
+  %wide.masked.load = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr %2, i32 1, <vscale x 16 x i1> %active.lane.mask, <vscale x 16 x i8> poison)
+  %3 = mul <vscale x 16 x i8> %wide.masked.load, shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+  %4 = getelementptr inbounds i8, ptr %dst, i64 %index
+  tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> %3, ptr %4, i32 1, <vscale x 16 x i1> %active.lane.mask)
+  %index.next = add i64 %index, %1
+  %active.lane.mask.next = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %5 = extractelement <vscale x 16 x i1> %active.lane.mask.next, i64 0
+  br i1 %5, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+; Halfword elements, UF=0
+define void @f_h_0(ptr noalias %dst, ptr %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_h_0:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB1_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    mov w9, w2
+; CHECK-SVE-NEXT:    cnth x10
+; CHECK-SVE-NEXT:    mov x8, xzr
+; CHECK-SVE-NEXT:    whilelo p0.h, xzr, x9
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB1_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1h { z0.h }, p0/z, [x1, x8, lsl #1]
+; CHECK-SVE-NEXT:    mul z0.h, z0.h, #3
+; CHECK-SVE-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; CHECK-SVE-NEXT:    add x8, x8, x10
+; CHECK-SVE-NEXT:    whilelo p0.h, x8, x9
+; CHECK-SVE-NEXT:    b.mi .LBB1_2
+; CHECK-SVE-NEXT:  .LBB1_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_h_0:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB1_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    mov w9, w2
+; CHECK-SVE2p1-NEXT:    mov x8, xzr
+; CHECK-SVE2p1-NEXT:    whilelo p0.h, xzr, x9
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB1_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1h { z0.h }, p0/z, [x1, x8, lsl #1]
+; CHECK-SVE2p1-NEXT:    mul z0.h, z0.h, #3
+; CHECK-SVE2p1-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; CHECK-SVE2p1-NEXT:    inch x8
+; CHECK-SVE2p1-NEXT:    whilelo p0.h, x8, x9
+; CHECK-SVE2p1-NEXT:    b.mi .LBB1_2
+; CHECK-SVE2p1-NEXT:  .LBB1_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp7 = icmp sgt i32 %n, 0
+  br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 3
+  %active.lane.mask.entry = tail call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 %wide.trip.count)
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next, %vector.body ]
+  %active.lane.mask = phi <vscale x 8 x i1> [ %active.lane.mask.entry, %for.body.preheader ], [ %active.lane.mask.next, %vector.body ]
+  %2 = getelementptr inbounds i16, ptr %src, i64 %index
+  %wide.masked.load = tail call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr %2, i32 2, <vscale x 8 x i1> %active.lane.mask, <vscale x 8 x i16> poison)
+  %3 = mul <vscale x 8 x i16> %wide.masked.load, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 3, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %4 = getelementptr inbounds i16, ptr %dst, i64 %index
+  tail call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> %3, ptr %4, i32 2, <vscale x 8 x i1> %active.lane.mask)
+  %index.next = add i64 %index, %1
+  %active.lane.mask.next = tail call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %5 = extractelement <vscale x 8 x i1> %active.lane.mask.next, i64 0
+  br i1 %5, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+; Double-precision float elements, UF=0
+define void @f_d_0(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_d_0:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB2_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    mov w9, w2
+; CHECK-SVE-NEXT:    fmov z0.d, #3.00000000
+; CHECK-SVE-NEXT:    cntd x10
+; CHECK-SVE-NEXT:    mov x8, xzr
+; CHECK-SVE-NEXT:    whilelo p0.d, xzr, x9
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB2_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1d { z1.d }, p0/z, [x1, x8, lsl #3]
+; CHECK-SVE-NEXT:    fmul z1.d, z1.d, z0.d
+; CHECK-SVE-NEXT:    st1d { z1.d }, p0, [x0, x8, lsl #3]
+; CHECK-SVE-NEXT:    add x8, x8, x10
+; CHECK-SVE-NEXT:    whilelo p0.d, x8, x9
+; CHECK-SVE-NEXT:    b.mi .LBB2_2
+; CHECK-SVE-NEXT:  .LBB2_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_d_0:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB2_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    mov w9, w2
+; CHECK-SVE2p1-NEXT:    fmov z0.d, #3.00000000
+; CHECK-SVE2p1-NEXT:    mov x8, xzr
+; CHECK-SVE2p1-NEXT:    whilelo p0.d, xzr, x9
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB2_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1d { z1.d }, p0/z, [x1, x8, lsl #3]
+; CHECK-SVE2p1-NEXT:    fmul z1.d, z1.d, z0.d
+; CHECK-SVE2p1-NEXT:    st1d { z1.d }, p0, [x0, x8, lsl #3]
+; CHECK-SVE2p1-NEXT:    incd x8
+; CHECK-SVE2p1-NEXT:    whilelo p0.d, x8, x9
+; CHECK-SVE2p1-NEXT:    b.mi .LBB2_2
+; CHECK-SVE2p1-NEXT:  .LBB2_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp6 = icmp sgt i32 %n, 0
+  br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 1
+  %active.lane.mask.entry = tail call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 %wide.trip.count)
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next, %vector.body ]
+  %active.lane.mask = phi <vscale x 2 x i1> [ %active.lane.mask.entry, %for.body.preheader ], [ %active.lane.mask.next, %vector.body ]
+  %2 = getelementptr inbounds double, ptr %src, i64 %index
+  %wide.masked.load = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr %2, i32 8, <vscale x 2 x i1> %active.lane.mask, <vscale x 2 x double> poison)
+  %3 = fmul <vscale x 2 x double> %wide.masked.load, shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+  %4 = getelementptr inbounds double, ptr %dst, i64 %index
+  tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> %3, ptr %4, i32 8, <vscale x 2 x i1> %active.lane.mask)
+  %index.next = add i64 %index, %1
+  %active.lane.mask.next = tail call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %5 = extractelement <vscale x 2 x i1> %active.lane.mask.next, i64 0
+  br i1 %5, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+; Byte elements, UF=2
+define void @f_b_2(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_b_2:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB3_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    rdvl x11, #1
+; CHECK-SVE-NEXT:    mov w8, w2
+; CHECK-SVE-NEXT:    mov x9, xzr
+; CHECK-SVE-NEXT:    add x10, x0, x11
+; CHECK-SVE-NEXT:    whilelo p0.b, x11, x8
+; CHECK-SVE-NEXT:    add x11, x1, x11
+; CHECK-SVE-NEXT:    rdvl x12, #2
+; CHECK-SVE-NEXT:    rdvl x13, #3
+; CHECK-SVE-NEXT:    whilelo p1.b, xzr, x8
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB3_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1b { z0.b }, p1/z, [x1, x9]
+; CHECK-SVE-NEXT:    ld1b { z1.b }, p0/z, [x11, x9]
+; CHECK-SVE-NEXT:    add x14, x13, x9
+; CHECK-SVE-NEXT:    mul z0.b, z0.b, #3
+; CHECK-SVE-NEXT:    mul z1.b, z1.b, #3
+; CHECK-SVE-NEXT:    st1b { z0.b }, p1, [x0, x9]
+; CHECK-SVE-NEXT:    st1b { z1.b }, p0, [x10, x9]
+; CHECK-SVE-NEXT:    add x9, x12, x9
+; CHECK-SVE-NEXT:    whilelo p0.b, x14, x8
+; CHECK-SVE-NEXT:    whilelo p1.b, x9, x8
+; CHECK-SVE-NEXT:    b.mi .LBB3_2
+; CHECK-SVE-NEXT:  .LBB3_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_b_2:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB3_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    rdvl x10, #1
+; CHECK-SVE2p1-NEXT:    mov w8, w2
+; CHECK-SVE2p1-NEXT:    mov x9, xzr
+; CHECK-SVE2p1-NEXT:    addvl x11, x1, #1
+; CHECK-SVE2p1-NEXT:    whilelo p1.b, xzr, x8
+; CHECK-SVE2p1-NEXT:    whilelo p0.b, x10, x8
+; CHECK-SVE2p1-NEXT:    addvl x10, x0, #1
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB3_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1b { z0.b }, p1/z, [x1, x9]
+; CHECK-SVE2p1-NEXT:    ld1b { z1.b }, p0/z, [x11, x9]
+; CHECK-SVE2p1-NEXT:    addvl x12, x9, #3
+; CHECK-SVE2p1-NEXT:    mul z0.b, z0.b, #3
+; CHECK-SVE2p1-NEXT:    mul z1.b, z1.b, #3
+; CHECK-SVE2p1-NEXT:    st1b { z0.b }, p1, [x0, x9]
+; CHECK-SVE2p1-NEXT:    st1b { z1.b }, p0, [x10, x9]
+; CHECK-SVE2p1-NEXT:    addvl x9, x9, #2
+; CHECK-SVE2p1-NEXT:    whilelo p0.b, x12, x8
+; CHECK-SVE2p1-NEXT:    whilelo p1.b, x9, x8
+; CHECK-SVE2p1-NEXT:    b.mi .LBB3_2
+; CHECK-SVE2p1-NEXT:  .LBB3_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp7 = icmp sgt i32 %n, 0
+  br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 5
+  %2 = tail call i64 @llvm.vscale.i64()
+  %3 = shl nuw nsw i64 %2, 4
+  %active.lane.mask.entry = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %3, i64 %wide.trip.count)
+  %active.lane.mask.entry10 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 %wide.trip.count)
+  %4 = tail call i64 @llvm.vscale.i64()
+  %5 = shl nuw nsw i64 %4, 4
+  %6 = tail call i64 @llvm.vscale.i64()
+  %7 = shl nuw nsw i64 %6, 4
+  %8 = tail call i64 @llvm.vscale.i64()
+  %9 = shl nuw nsw i64 %8, 4
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next13, %vector.body ]
+  %active.lane.mask = phi <vscale x 16 x i1> [ %active.lane.mask.entry10, %for.body.preheader ], [ %active.lane.mask.next14, %vector.body ]
+  %active.lane.mask11 = phi <vscale x 16 x i1> [ %active.lane.mask.entry, %for.body.preheader ], [ %active.lane.mask.next, %vector.body ]
+  %10 = getelementptr inbounds i8, ptr %src, i64 %index
+  %11 = getelementptr inbounds i8, ptr %10, i64 %5
+  %wide.masked.load = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr %10, i32 1, <vscale x 16 x i1> %active.lane.mask, <vscale x 16 x i8> poison)
+  %wide.masked.load12 = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull %11, i32 1, <vscale x 16 x i1> %active.lane.mask11, <vscale x 16 x i8> poison)
+  %12 = mul <vscale x 16 x i8> %wide.masked.load, shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+  %13 = mul <vscale x 16 x i8> %wide.masked.load12, shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+  %14 = getelementptr inbounds i8, ptr %dst, i64 %index
+  %15 = getelementptr inbounds i8, ptr %14, i64 %7
+  tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> %12, ptr %14, i32 1, <vscale x 16 x i1> %active.lane.mask)
+  tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> %13, ptr %15, i32 1, <vscale x 16 x i1> %active.lane.mask11)
+  %index.next = add i64 %index, %1
+  %index.next13 = add i64 %index, %1
+  %16 = add i64 %index.next, %9
+  %active.lane.mask.next = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %16, i64 %wide.trip.count)
+  %active.lane.mask.next14 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %17 = extractelement <vscale x 16 x i1> %active.lane.mask.next14, i64 0
+  br i1 %17, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+; Halfword elements, UF=2
+define void @f_h_2(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_h_2:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB4_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    mov w9, w2
+; CHECK-SVE-NEXT:    rdvl x10, #1
+; CHECK-SVE-NEXT:    mov x8, xzr
+; CHECK-SVE-NEXT:    add x11, x0, x10
+; CHECK-SVE-NEXT:    whilelo p1.b, xzr, x9
+; CHECK-SVE-NEXT:    add x12, x1, x10
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB4_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1h { z0.h }, p1/z, [x1, x8, lsl #1]
+; CHECK-SVE-NEXT:    ld1h { z1.h }, p0/z, [x12, x8, lsl #1]
+; CHECK-SVE-NEXT:    mul z0.h, z0.h, #3
+; CHECK-SVE-NEXT:    mul z1.h, z1.h, #3
+; CHECK-SVE-NEXT:    st1h { z0.h }, p1, [x0, x8, lsl #1]
+; CHECK-SVE-NEXT:    st1h { z1.h }, p0, [x11, x8, lsl #1]
+; CHECK-SVE-NEXT:    add x8, x10, x8
+; CHECK-SVE-NEXT:    whilelo p1.b, x8, x9
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    mov z0.h, p1/z, #1 // =0x1
+; CHECK-SVE-NEXT:    fmov w13, s0
+; CHECK-SVE-NEXT:    tbnz w13, #0, .LBB4_2
+; CHECK-SVE-NEXT:  .LBB4_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_h_2:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB4_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    mov w9, w2
+; CHECK-SVE2p1-NEXT:    addvl x10, x0, #1
+; CHECK-SVE2p1-NEXT:    mov x8, xzr
+; CHECK-SVE2p1-NEXT:    whilelo { p0.h, p1.h }, xzr, x9
+; CHECK-SVE2p1-NEXT:    addvl x11, x1, #1
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB4_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1h { z0.h }, p0/z, [x1, x8, lsl #1]
+; CHECK-SVE2p1-NEXT:    ld1h { z1.h }, p1/z, [x11, x8, lsl #1]
+; CHECK-SVE2p1-NEXT:    mul z0.h, z0.h, #3
+; CHECK-SVE2p1-NEXT:    mul z1.h, z1.h, #3
+; CHECK-SVE2p1-NEXT:    st1h { z0.h }, p0, [x0, x8, lsl #1]
+; CHECK-SVE2p1-NEXT:    st1h { z1.h }, p1, [x10, x8, lsl #1]
+; CHECK-SVE2p1-NEXT:    addvl x8, x8, #1
+; CHECK-SVE2p1-NEXT:    whilelo { p0.h, p1.h }, x8, x9
+; CHECK-SVE2p1-NEXT:    mov z0.h, p0/z, #1 // =0x1
+; CHECK-SVE2p1-NEXT:    fmov w12, s0
+; CHECK-SVE2p1-NEXT:    tbnz w12, #0, .LBB4_2
+; CHECK-SVE2p1-NEXT:  .LBB4_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp7 = icmp sgt i32 %n, 0
+  br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 4
+  %active.lane.mask.entry = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 %wide.trip.count)
+  %active.lane.mask.entry10 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.entry, i64 8)
+  %active.lane.mask.entry11 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.entry, i64 0)
+  %2 = tail call i64 @llvm.vscale.i64()
+  %3 = shl nuw nsw i64 %2, 3
+  %4 = tail call i64 @llvm.vscale.i64()
+  %5 = shl nuw nsw i64 %4, 3
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next14, %vector.body ]
+  %active.lane.mask = phi <vscale x 8 x i1> [ %active.lane.mask.entry11, %for.body.preheader ], [ %active.lane.mask.next16, %vector.body ]
+  %active.lane.mask12 = phi <vscale x 8 x i1> [ %active.lane.mask.entry10, %for.body.preheader ], [ %active.lane.mask.next15, %vector.body ]
+  %6 = getelementptr inbounds i16, ptr %src, i64 %index
+  %7 = getelementptr inbounds i16, ptr %6, i64 %3
+  %wide.masked.load = tail call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr %6, i32 2, <vscale x 8 x i1> %active.lane.mask, <vscale x 8 x i16> poison)
+  %wide.masked.load13 = tail call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr nonnull %7, i32 2, <vscale x 8 x i1> %active.lane.mask12, <vscale x 8 x i16> poison)
+  %8 = mul <vscale x 8 x i16> %wide.masked.load, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 3, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %9 = mul <vscale x 8 x i16> %wide.masked.load13, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 3, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %10 = getelementptr inbounds i16, ptr %dst, i64 %index
+  %11 = getelementptr inbounds i16, ptr %10, i64 %5
+  tail call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> %8, ptr %10, i32 2, <vscale x 8 x i1> %active.lane.mask)
+  tail call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> %9, ptr %11, i32 2, <vscale x 8 x i1> %active.lane.mask12)
+  %index.next = add i64 %index, %1
+  %index.next14 = add i64 %index, %1
+  %active.lane.mask.next = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %active.lane.mask.next15 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.next, i64 8)
+  %active.lane.mask.next16 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.next, i64 0)
+  %12 = extractelement <vscale x 8 x i1> %active.lane.mask.next16, i64 0
+  br i1 %12, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+; Double-precision float elements, UF=2
+define void @f3(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f3:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB5_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    mov w9, w2
+; CHECK-SVE-NEXT:    rdvl x11, #1
+; CHECK-SVE-NEXT:    fmov z0.d, #3.00000000
+; CHECK-SVE-NEXT:    mov x8, xzr
+; CHECK-SVE-NEXT:    add x10, x0, x11
+; CHECK-SVE-NEXT:    whilelo p1.s, xzr, x9
+; CHECK-SVE-NEXT:    add x11, x1, x11
+; CHECK-SVE-NEXT:    cntw x12
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB5_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1d { z1.d }, p1/z, [x1, x8, lsl #3]
+; CHECK-SVE-NEXT:    ld1d { z2.d }, p0/z, [x11, x8, lsl #3]
+; CHECK-SVE-NEXT:    fmul z1.d, z1.d, z0.d
+; CHECK-SVE-NEXT:    fmul z2.d, z2.d, z0.d
+; CHECK-SVE-NEXT:    st1d { z1.d }, p1, [x0, x8, lsl #3]
+; CHECK-SVE-NEXT:    st1d { z2.d }, p0, [x10, x8, lsl #3]
+; CHECK-SVE-NEXT:    add x8, x12, x8
+; CHECK-SVE-NEXT:    whilelo p1.s, x8, x9
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    mov z1.d, p1/z, #1 // =0x1
+; CHECK-SVE-NEXT:    fmov x13, d1
+; CHECK-SVE-NEXT:    tbnz w13, #0, .LBB5_2
+; CHECK-SVE-NEXT:  .LBB5_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f3:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB5_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    mov w9, w2
+; CHECK-SVE2p1-NEXT:    addvl x10, x0, #1
+; CHECK-SVE2p1-NEXT:    fmov z0.d, #3.00000000
+; CHECK-SVE2p1-NEXT:    mov x8, xzr
+; CHECK-SVE2p1-NEXT:    whilelo { p0.d, p1.d }, xzr, x9
+; CHECK-SVE2p1-NEXT:    addvl x11, x1, #1
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB5_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1d { z1.d }, p0/z, [x1, x8, lsl #3]
+; CHECK-SVE2p1-NEXT:    ld1d { z2.d }, p1/z, [x11, x8, lsl #3]
+; CHECK-SVE2p1-NEXT:    fmul z1.d, z1.d, z0.d
+; CHECK-SVE2p1-NEXT:    fmul z2.d, z2.d, z0.d
+; CHECK-SVE2p1-NEXT:    st1d { z1.d }, p0, [x0, x8, lsl #3]
+; CHECK-SVE2p1-NEXT:    st1d { z2.d }, p1, [x10, x8, lsl #3]
+; CHECK-SVE2p1-NEXT:    incw x8
+; CHECK-SVE2p1-NEXT:    whilelo { p0.d, p1.d }, x8, x9
+; CHECK-SVE2p1-NEXT:    mov z1.d, p0/z, #1 // =0x1
+; CHECK-SVE2p1-NEXT:    fmov x12, d1
+; CHECK-SVE2p1-NEXT:    tbnz w12, #0, .LBB5_2
+; CHECK-SVE2p1-NEXT:  .LBB5_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp6 = icmp sgt i32 %n, 0
+  br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 2
+  %active.lane.mask.entry = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 %wide.trip.count)
+  %active.lane.mask.entry1 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.entry, i64 2)
+  %active.lane.mask.entry2 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.entry, i64 0)
+  %2 = tail call i64 @llvm.vscale.i64()
+  %3 = shl nuw nsw i64 %2, 1
+  %4 = tail call i64 @llvm.vscale.i64()
+  %5 = shl nuw nsw i64 %4, 1
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next5, %vector.body ]
+  %active.lane.mask = phi <vscale x 2 x i1> [ %active.lane.mask.entry2, %for.body.preheader ], [ %active.lane.mask.next7, %vector.body ]
+  %active.lane.mask3 = phi <vscale x 2 x i1> [ %active.lane.mask.entry1, %for.body.preheader ], [ %active.lane.mask.next6, %vector.body ]
+  %6 = getelementptr inbounds double, ptr %src, i64 %index
+  %wide.masked.load = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr %6, i32 8, <vscale x 2 x i1> %active.lane.mask, <vscale x 2 x double> poison)
+  %7 = getelementptr inbounds double, ptr %6, i64 %3
+  %wide.masked.load4 = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull %7, i32 8, <vscale x 2 x i1> %active.lane.mask3, <vscale x 2 x double> poison)
+  %8 = fmul <vscale x 2 x double> %wide.masked.load, shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+  %9 = fmul <vscale x 2 x double> %wide.masked.load4, shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+  %10 = getelementptr inbounds double, ptr %dst, i64 %index
+  tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> %8, ptr %10, i32 8, <vscale x 2 x i1> %active.lane.mask)
+  %11 = getelementptr inbounds double, ptr %10, i64 %5
+  tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> %9, ptr %11, i32 8, <vscale x 2 x i1> %active.lane.mask3)
+  %index.next = add i64 %index, %1
+  %index.next5 = add i64 %index, %1
+  %active.lane.mask.next = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %active.lane.mask.next6 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.next, i64 2)
+  %active.lane.mask.next7 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.next, i64 0)
+  %12 = extractelement <vscale x 2 x i1> %active.lane.mask.next7, i64 0
+  br i1 %12, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+; Byte elements, UF=4
+define void @f_b_4(ptr noalias %dst, ptr %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_b_4:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB6_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    rdvl x14, #2
+; CHECK-SVE-NEXT:    mov w8, w2
+; CHECK-SVE-NEXT:    mov x9, xzr
+; CHECK-SVE-NEXT:    add x11, x0, x14
+; CHECK-SVE-NEXT:    rdvl x13, #3
+; CHECK-SVE-NEXT:    rdvl x15, #1
+; CHECK-SVE-NEXT:    add x10, x0, x13
+; CHECK-SVE-NEXT:    add x12, x0, x15
+; CHECK-SVE-NEXT:    rdvl x16, #4
+; CHECK-SVE-NEXT:    rdvl x17, #5
+; CHECK-SVE-NEXT:    rdvl x18, #6
+; CHECK-SVE-NEXT:    rdvl x2, #7
+; CHECK-SVE-NEXT:    whilelo p0.b, x13, x8
+; CHECK-SVE-NEXT:    add x13, x1, x13
+; CHECK-SVE-NEXT:    whilelo p1.b, x14, x8
+; CHECK-SVE-NEXT:    add x14, x1, x14
+; CHECK-SVE-NEXT:    whilelo p2.b, x15, x8
+; CHECK-SVE-NEXT:    add x15, x1, x15
+; CHECK-SVE-NEXT:    whilelo p3.b, xzr, x8
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB6_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1b { z0.b }, p3/z, [x1, x9]
+; CHECK-SVE-NEXT:    ld1b { z1.b }, p2/z, [x15, x9]
+; CHECK-SVE-NEXT:    add x3, x2, x9
+; CHECK-SVE-NEXT:    mul z0.b, z0.b, #3
+; CHECK-SVE-NEXT:    ld1b { z2.b }, p1/z, [x14, x9]
+; CHECK-SVE-NEXT:    ld1b { z3.b }, p0/z, [x13, x9]
+; CHECK-SVE-NEXT:    mul z1.b, z1.b, #3
+; CHECK-SVE-NEXT:    mul z2.b, z2.b, #3
+; CHECK-SVE-NEXT:    mul z3.b, z3.b, #3
+; CHECK-SVE-NEXT:    st1b { z0.b }, p3, [x0, x9]
+; CHECK-SVE-NEXT:    st1b { z1.b }, p2, [x12, x9]
+; CHECK-SVE-NEXT:    st1b { z2.b }, p1, [x11, x9]
+; CHECK-SVE-NEXT:    st1b { z3.b }, p0, [x10, x9]
+; CHECK-SVE-NEXT:    whilelo p0.b, x3, x8
+; CHECK-SVE-NEXT:    add x3, x18, x9
+; CHECK-SVE-NEXT:    whilelo p1.b, x3, x8
+; CHECK-SVE-NEXT:    add x3, x17, x9
+; CHECK-SVE-NEXT:    add x9, x16, x9
+; CHECK-SVE-NEXT:    whilelo p2.b, x3, x8
+; CHECK-SVE-NEXT:    whilelo p3.b, x9, x8
+; CHECK-SVE-NEXT:    b.mi .LBB6_2
+; CHECK-SVE-NEXT:  .LBB6_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_b_4:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB6_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    rdvl x10, #1
+; CHECK-SVE2p1-NEXT:    mov w8, w2
+; CHECK-SVE2p1-NEXT:    mov x9, xzr
+; CHECK-SVE2p1-NEXT:    rdvl x11, #2
+; CHECK-SVE2p1-NEXT:    rdvl x12, #3
+; CHECK-SVE2p1-NEXT:    addvl x13, x1, #3
+; CHECK-SVE2p1-NEXT:    addvl x14, x1, #2
+; CHECK-SVE2p1-NEXT:    addvl x15, x1, #1
+; CHECK-SVE2p1-NEXT:    whilelo p3.b, xzr, x8
+; CHECK-SVE2p1-NEXT:    whilelo p0.b, x12, x8
+; CHECK-SVE2p1-NEXT:    addvl x12, x0, #1
+; CHECK-SVE2p1-NEXT:    whilelo p1.b, x11, x8
+; CHECK-SVE2p1-NEXT:    addvl x11, x0, #2
+; CHECK-SVE2p1-NEXT:    whilelo p2.b, x10, x8
+; CHECK-SVE2p1-NEXT:    addvl x10, x0, #3
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB6_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1b { z0.b }, p3/z, [x1, x9]
+; CHECK-SVE2p1-NEXT:    ld1b { z1.b }, p2/z, [x15, x9]
+; CHECK-SVE2p1-NEXT:    addvl x16, x9, #5
+; CHECK-SVE2p1-NEXT:    addvl x17, x9, #6
+; CHECK-SVE2p1-NEXT:    ld1b { z2.b }, p1/z, [x14, x9]
+; CHECK-SVE2p1-NEXT:    ld1b { z3.b }, p0/z, [x13, x9]
+; CHECK-SVE2p1-NEXT:    mul z0.b, z0.b, #3
+; CHECK-SVE2p1-NEXT:    mul z1.b, z1.b, #3
+; CHECK-SVE2p1-NEXT:    mul z2.b, z2.b, #3
+; CHECK-SVE2p1-NEXT:    mul z3.b, z3.b, #3
+; CHECK-SVE2p1-NEXT:    addvl x18, x9, #7
+; CHECK-SVE2p1-NEXT:    st1b { z0.b }, p3, [x0, x9]
+; CHECK-SVE2p1-NEXT:    st1b { z1.b }, p2, [x12, x9]
+; CHECK-SVE2p1-NEXT:    st1b { z2.b }, p1, [x11, x9]
+; CHECK-SVE2p1-NEXT:    st1b { z3.b }, p0, [x10, x9]
+; CHECK-SVE2p1-NEXT:    addvl x9, x9, #4
+; CHECK-SVE2p1-NEXT:    whilelo p0.b, x18, x8
+; CHECK-SVE2p1-NEXT:    whilelo p1.b, x17, x8
+; CHECK-SVE2p1-NEXT:    whilelo p2.b, x16, x8
+; CHECK-SVE2p1-NEXT:    whilelo p3.b, x9, x8
+; CHECK-SVE2p1-NEXT:    b.mi .LBB6_2
+; CHECK-SVE2p1-NEXT:  .LBB6_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp7 = icmp sgt i32 %n, 0
+  br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 6
+  %2 = tail call i64 @llvm.vscale.i64()
+  %3 = shl nuw nsw i64 %2, 4
+  %4 = tail call i64 @llvm.vscale.i64()
+  %5 = shl nuw nsw i64 %4, 5
+  %6 = tail call i64 @llvm.vscale.i64()
+  %7 = mul nuw nsw i64 %6, 48
+  %active.lane.mask.entry = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %7, i64 %wide.trip.count)
+  %active.lane.mask.entry12 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %5, i64 %wide.trip.count)
+  %active.lane.mask.entry13 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %3, i64 %wide.trip.count)
+  %active.lane.mask.entry14 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 %wide.trip.count)
+  %8 = tail call i64 @llvm.vscale.i64()
+  %9 = shl nuw nsw i64 %8, 4
+  %10 = tail call i64 @llvm.vscale.i64()
+  %11 = shl nuw nsw i64 %10, 5
+  %12 = tail call i64 @llvm.vscale.i64()
+  %13 = mul nuw nsw i64 %12, 48
+  %14 = tail call i64 @llvm.vscale.i64()
+  %15 = shl nuw nsw i64 %14, 4
+  %16 = tail call i64 @llvm.vscale.i64()
+  %17 = shl nuw nsw i64 %16, 5
+  %18 = tail call i64 @llvm.vscale.i64()
+  %19 = mul nuw nsw i64 %18, 48
+  %20 = tail call i64 @llvm.vscale.i64()
+  %21 = shl nuw nsw i64 %20, 4
+  %22 = tail call i64 @llvm.vscale.i64()
+  %23 = shl nuw nsw i64 %22, 5
+  %24 = tail call i64 @llvm.vscale.i64()
+  %25 = mul nuw nsw i64 %24, 48
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next23, %vector.body ]
+  %active.lane.mask = phi <vscale x 16 x i1> [ %active.lane.mask.entry14, %for.body.preheader ], [ %active.lane.mask.next26, %vector.body ]
+  %active.lane.mask15 = phi <vscale x 16 x i1> [ %active.lane.mask.entry13, %for.body.preheader ], [ %active.lane.mask.next25, %vector.body ]
+  %active.lane.mask16 = phi <vscale x 16 x i1> [ %active.lane.mask.entry12, %for.body.preheader ], [ %active.lane.mask.next24, %vector.body ]
+  %active.lane.mask17 = phi <vscale x 16 x i1> [ %active.lane.mask.entry, %for.body.preheader ], [ %active.lane.mask.next, %vector.body ]
+  %26 = getelementptr inbounds i8, ptr %src, i64 %index
+  %27 = getelementptr inbounds i8, ptr %26, i64 %9
+  %28 = getelementptr inbounds i8, ptr %26, i64 %11
+  %29 = getelementptr inbounds i8, ptr %26, i64 %13
+  %wide.masked.load = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr %26, i32 1, <vscale x 16 x i1> %active.lane.mask, <vscale x 16 x i8> poison)
+  %wide.masked.load18 = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull %27, i32 1, <vscale x 16 x i1> %active.lane.mask15, <vscale x 16 x i8> poison)
+  %wide.masked.load19 = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull %28, i32 1, <vscale x 16 x i1> %active.lane.mask16, <vscale x 16 x i8> poison)
+  %wide.masked.load20 = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull %29, i32 1, <vscale x 16 x i1> %active.lane.mask17, <vscale x 16 x i8> poison)
+  %30 = mul <vscale x 16 x i8> %wide.masked.load, shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+  %31 = mul <vscale x 16 x i8> %wide.masked.load18, shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+  %32 = mul <vscale x 16 x i8> %wide.masked.load19, shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+  %33 = mul <vscale x 16 x i8> %wide.masked.load20, shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+  %34 = getelementptr inbounds i8, ptr %dst, i64 %index
+  %35 = getelementptr inbounds i8, ptr %34, i64 %15
+  %36 = getelementptr inbounds i8, ptr %34, i64 %17
+  %37 = getelementptr inbounds i8, ptr %34, i64 %19
+  tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> %30, ptr %34, i32 1, <vscale x 16 x i1> %active.lane.mask)
+  tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> %31, ptr %35, i32 1, <vscale x 16 x i1> %active.lane.mask15)
+  tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> %32, ptr %36, i32 1, <vscale x 16 x i1> %active.lane.mask16)
+  tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> %33, ptr %37, i32 1, <vscale x 16 x i1> %active.lane.mask17)
+  %index.next = add i64 %index, %1
+  %index.next23 = add i64 %index, %1
+  %38 = add i64 %index.next, %21
+  %39 = add i64 %index.next, %23
+  %40 = add i64 %index.next, %25
+  %active.lane.mask.next = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %40, i64 %wide.trip.count)
+  %active.lane.mask.next24 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %39, i64 %wide.trip.count)
+  %active.lane.mask.next25 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %38, i64 %wide.trip.count)
+  %active.lane.mask.next26 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %41 = extractelement <vscale x 16 x i1> %active.lane.mask.next26, i64 0
+  br i1 %41, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+; Halfword elements, UF=4
+define void @f_h_4(ptr noalias %dst, ptr %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_h_4:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB7_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    rdvl x17, #1
+; CHECK-SVE-NEXT:    mov w8, w2
+; CHECK-SVE-NEXT:    mov x10, xzr
+; CHECK-SVE-NEXT:    add x14, x0, x17
+; CHECK-SVE-NEXT:    whilelo p1.b, x17, x8
+; CHECK-SVE-NEXT:    add x17, x1, x17
+; CHECK-SVE-NEXT:    rdvl x9, #2
+; CHECK-SVE-NEXT:    add x13, x0, x9
+; CHECK-SVE-NEXT:    add x16, x1, x9
+; CHECK-SVE-NEXT:    rdvl x11, #3
+; CHECK-SVE-NEXT:    add x12, x0, x11
+; CHECK-SVE-NEXT:    add x15, x1, x11
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    whilelo p3.b, xzr, x8
+; CHECK-SVE-NEXT:    punpkhi p2.h, p3.b
+; CHECK-SVE-NEXT:    punpklo p3.h, p3.b
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB7_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1h { z0.h }, p3/z, [x1, x10, lsl #1]
+; CHECK-SVE-NEXT:    ld1h { z1.h }, p2/z, [x17, x10, lsl #1]
+; CHECK-SVE-NEXT:    add x18, x11, x10
+; CHECK-SVE-NEXT:    mul z0.h, z0.h, #3
+; CHECK-SVE-NEXT:    ld1h { z2.h }, p1/z, [x16, x10, lsl #1]
+; CHECK-SVE-NEXT:    ld1h { z3.h }, p0/z, [x15, x10, lsl #1]
+; CHECK-SVE-NEXT:    mul z1.h, z1.h, #3
+; CHECK-SVE-NEXT:    mul z2.h, z2.h, #3
+; CHECK-SVE-NEXT:    mul z3.h, z3.h, #3
+; CHECK-SVE-NEXT:    st1h { z0.h }, p3, [x0, x10, lsl #1]
+; CHECK-SVE-NEXT:    st1h { z1.h }, p2, [x14, x10, lsl #1]
+; CHECK-SVE-NEXT:    st1h { z2.h }, p1, [x13, x10, lsl #1]
+; CHECK-SVE-NEXT:    st1h { z3.h }, p0, [x12, x10, lsl #1]
+; CHECK-SVE-NEXT:    add x10, x9, x10
+; CHECK-SVE-NEXT:    whilelo p1.b, x18, x8
+; CHECK-SVE-NEXT:    whilelo p3.b, x10, x8
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    punpkhi p2.h, p3.b
+; CHECK-SVE-NEXT:    punpklo p3.h, p3.b
+; CHECK-SVE-NEXT:    mov z0.h, p3/z, #1 // =0x1
+; CHECK-SVE-NEXT:    fmov w18, s0
+; CHECK-SVE-NEXT:    tbnz w18, #0, .LBB7_2
+; CHECK-SVE-NEXT:  .LBB7_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_h_4:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB7_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    rdvl x10, #1
+; CHECK-SVE2p1-NEXT:    mov w8, w2
+; CHECK-SVE2p1-NEXT:    mov x9, xzr
+; CHECK-SVE2p1-NEXT:    whilelo { p2.h, p3.h }, xzr, x8
+; CHECK-SVE2p1-NEXT:    addvl x11, x0, #2
+; CHECK-SVE2p1-NEXT:    whilelo { p0.h, p1.h }, x10, x8
+; CHECK-SVE2p1-NEXT:    addvl x10, x0, #3
+; CHECK-SVE2p1-NEXT:    addvl x12, x0, #1
+; CHECK-SVE2p1-NEXT:    addvl x13, x1, #3
+; CHECK-SVE2p1-NEXT:    addvl x14, x1, #2
+; CHECK-SVE2p1-NEXT:    addvl x15, x1, #1
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB7_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1h { z0.h }, p2/z, [x1, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    ld1h { z1.h }, p3/z, [x15, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    addvl x16, x9, #3
+; CHECK-SVE2p1-NEXT:    mul z0.h, z0.h, #3
+; CHECK-SVE2p1-NEXT:    ld1h { z2.h }, p0/z, [x14, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    ld1h { z3.h }, p1/z, [x13, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    mul z1.h, z1.h, #3
+; CHECK-SVE2p1-NEXT:    mul z2.h, z2.h, #3
+; CHECK-SVE2p1-NEXT:    mul z3.h, z3.h, #3
+; CHECK-SVE2p1-NEXT:    st1h { z0.h }, p2, [x0, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    st1h { z1.h }, p3, [x12, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    st1h { z2.h }, p0, [x11, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    st1h { z3.h }, p1, [x10, x9, lsl #1]
+; CHECK-SVE2p1-NEXT:    addvl x9, x9, #2
+; CHECK-SVE2p1-NEXT:    whilelo { p0.h, p1.h }, x16, x8
+; CHECK-SVE2p1-NEXT:    whilelo { p2.h, p3.h }, x9, x8
+; CHECK-SVE2p1-NEXT:    mov z0.h, p2/z, #1 // =0x1
+; CHECK-SVE2p1-NEXT:    fmov w16, s0
+; CHECK-SVE2p1-NEXT:    tbnz w16, #0, .LBB7_2
+; CHECK-SVE2p1-NEXT:  .LBB7_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp7 = icmp sgt i32 %n, 0
+  br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 5
+  %2 = tail call i64 @llvm.vscale.i64()
+  %3 = shl nuw nsw i64 %2, 4
+  %active.lane.mask.entry = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %3, i64 %wide.trip.count)
+  %active.lane.mask.entry12 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 %wide.trip.count)
+  %active.lane.mask.entry13 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.entry, i64 8)
+  %active.lane.mask.entry14 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.entry, i64 0)
+  %active.lane.mask.entry15 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.entry12, i64 8)
+  %active.lane.mask.entry16 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.entry12, i64 0)
+  %4 = tail call i64 @llvm.vscale.i64()
+  %5 = shl nuw nsw i64 %4, 3
+  %6 = tail call i64 @llvm.vscale.i64()
+  %7 = shl nuw nsw i64 %6, 4
+  %8 = tail call i64 @llvm.vscale.i64()
+  %9 = mul nuw nsw i64 %8, 24
+  %10 = tail call i64 @llvm.vscale.i64()
+  %11 = shl nuw nsw i64 %10, 3
+  %12 = tail call i64 @llvm.vscale.i64()
+  %13 = shl nuw nsw i64 %12, 4
+  %14 = tail call i64 @llvm.vscale.i64()
+  %15 = mul nuw nsw i64 %14, 24
+  %16 = tail call i64 @llvm.vscale.i64()
+  %17 = shl nuw nsw i64 %16, 4
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next25, %vector.body ]
+  %active.lane.mask = phi <vscale x 8 x i1> [ %active.lane.mask.entry16, %for.body.preheader ], [ %active.lane.mask.next30, %vector.body ]
+  %active.lane.mask17 = phi <vscale x 8 x i1> [ %active.lane.mask.entry15, %for.body.preheader ], [ %active.lane.mask.next29, %vector.body ]
+  %active.lane.mask18 = phi <vscale x 8 x i1> [ %active.lane.mask.entry14, %for.body.preheader ], [ %active.lane.mask.next28, %vector.body ]
+  %active.lane.mask19 = phi <vscale x 8 x i1> [ %active.lane.mask.entry13, %for.body.preheader ], [ %active.lane.mask.next27, %vector.body ]
+  %18 = getelementptr inbounds i16, ptr %src, i64 %index
+  %19 = getelementptr inbounds i16, ptr %18, i64 %5
+  %20 = getelementptr inbounds i16, ptr %18, i64 %7
+  %21 = getelementptr inbounds i16, ptr %18, i64 %9
+  %wide.masked.load = tail call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr %18, i32 2, <vscale x 8 x i1> %active.lane.mask, <vscale x 8 x i16> poison)
+  %wide.masked.load20 = tail call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr nonnull %19, i32 2, <vscale x 8 x i1> %active.lane.mask17, <vscale x 8 x i16> poison)
+  %wide.masked.load21 = tail call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr nonnull %20, i32 2, <vscale x 8 x i1> %active.lane.mask18, <vscale x 8 x i16> poison)
+  %wide.masked.load22 = tail call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr nonnull %21, i32 2, <vscale x 8 x i1> %active.lane.mask19, <vscale x 8 x i16> poison)
+  %22 = mul <vscale x 8 x i16> %wide.masked.load, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 3, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %23 = mul <vscale x 8 x i16> %wide.masked.load20, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 3, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %24 = mul <vscale x 8 x i16> %wide.masked.load21, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 3, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %25 = mul <vscale x 8 x i16> %wide.masked.load22, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 3, i64 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %26 = getelementptr inbounds i16, ptr %dst, i64 %index
+  %27 = getelementptr inbounds i16, ptr %26, i64 %11
+  %28 = getelementptr inbounds i16, ptr %26, i64 %13
+  %29 = getelementptr inbounds i16, ptr %26, i64 %15
+  tail call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> %22, ptr %26, i32 2, <vscale x 8 x i1> %active.lane.mask)
+  tail call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> %23, ptr %27, i32 2, <vscale x 8 x i1> %active.lane.mask17)
+  tail call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> %24, ptr %28, i32 2, <vscale x 8 x i1> %active.lane.mask18)
+  tail call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> %25, ptr %29, i32 2, <vscale x 8 x i1> %active.lane.mask19)
+  %index.next = add i64 %index, %1
+  %index.next25 = add i64 %index, %1
+  %30 = add i64 %index.next, %17
+  %active.lane.mask.next = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %30, i64 %wide.trip.count)
+  %active.lane.mask.next26 = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %active.lane.mask.next27 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.next, i64 8)
+  %active.lane.mask.next28 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.next, i64 0)
+  %active.lane.mask.next29 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.next26, i64 8)
+  %active.lane.mask.next30 = tail call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> %active.lane.mask.next26, i64 0)
+  %31 = extractelement <vscale x 8 x i1> %active.lane.mask.next30, i64 0
+  br i1 %31, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+
+; Double-precision float elements, UF=4
+define void @f_d_4(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-LABEL: f_d_4:
+; CHECK-SVE:       // %bb.0: // %entry
+; CHECK-SVE-NEXT:    cmp w2, #1
+; CHECK-SVE-NEXT:    b.lt .LBB8_3
+; CHECK-SVE-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE-NEXT:    cntw x10
+; CHECK-SVE-NEXT:    mov w8, w2
+; CHECK-SVE-NEXT:    fmov z0.d, #3.00000000
+; CHECK-SVE-NEXT:    mov x9, xzr
+; CHECK-SVE-NEXT:    rdvl x13, #3
+; CHECK-SVE-NEXT:    rdvl x14, #2
+; CHECK-SVE-NEXT:    add x11, x0, x14
+; CHECK-SVE-NEXT:    add x14, x1, x14
+; CHECK-SVE-NEXT:    rdvl x15, #1
+; CHECK-SVE-NEXT:    add x12, x0, x15
+; CHECK-SVE-NEXT:    add x15, x1, x15
+; CHECK-SVE-NEXT:    cnth x16
+; CHECK-SVE-NEXT:    cntw x17, all, mul #3
+; CHECK-SVE-NEXT:    whilelo p1.s, x10, x8
+; CHECK-SVE-NEXT:    add x10, x0, x13
+; CHECK-SVE-NEXT:    add x13, x1, x13
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    whilelo p3.s, xzr, x8
+; CHECK-SVE-NEXT:    punpkhi p2.h, p3.b
+; CHECK-SVE-NEXT:    punpklo p3.h, p3.b
+; CHECK-SVE-NEXT:    .p2align 5, , 16
+; CHECK-SVE-NEXT:  .LBB8_2: // %vector.body
+; CHECK-SVE-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE-NEXT:    ld1d { z1.d }, p3/z, [x1, x9, lsl #3]
+; CHECK-SVE-NEXT:    ld1d { z2.d }, p2/z, [x15, x9, lsl #3]
+; CHECK-SVE-NEXT:    add x18, x17, x9
+; CHECK-SVE-NEXT:    fmul z1.d, z1.d, z0.d
+; CHECK-SVE-NEXT:    ld1d { z3.d }, p1/z, [x14, x9, lsl #3]
+; CHECK-SVE-NEXT:    ld1d { z4.d }, p0/z, [x13, x9, lsl #3]
+; CHECK-SVE-NEXT:    fmul z2.d, z2.d, z0.d
+; CHECK-SVE-NEXT:    fmul z3.d, z3.d, z0.d
+; CHECK-SVE-NEXT:    fmul z4.d, z4.d, z0.d
+; CHECK-SVE-NEXT:    st1d { z1.d }, p3, [x0, x9, lsl #3]
+; CHECK-SVE-NEXT:    st1d { z2.d }, p2, [x12, x9, lsl #3]
+; CHECK-SVE-NEXT:    st1d { z3.d }, p1, [x11, x9, lsl #3]
+; CHECK-SVE-NEXT:    st1d { z4.d }, p0, [x10, x9, lsl #3]
+; CHECK-SVE-NEXT:    add x9, x16, x9
+; CHECK-SVE-NEXT:    whilelo p1.s, x18, x8
+; CHECK-SVE-NEXT:    whilelo p3.s, x9, x8
+; CHECK-SVE-NEXT:    punpkhi p0.h, p1.b
+; CHECK-SVE-NEXT:    punpklo p1.h, p1.b
+; CHECK-SVE-NEXT:    punpkhi p2.h, p3.b
+; CHECK-SVE-NEXT:    punpklo p3.h, p3.b
+; CHECK-SVE-NEXT:    mov z1.d, p3/z, #1 // =0x1
+; CHECK-SVE-NEXT:    fmov x18, d1
+; CHECK-SVE-NEXT:    tbnz w18, #0, .LBB8_2
+; CHECK-SVE-NEXT:  .LBB8_3: // %for.cond.cleanup
+; CHECK-SVE-NEXT:    ret
+;
+; CHECK-SVE2p1-LABEL: f_d_4:
+; CHECK-SVE2p1:       // %bb.0: // %entry
+; CHECK-SVE2p1-NEXT:    cmp w2, #1
+; CHECK-SVE2p1-NEXT:    b.lt .LBB8_3
+; CHECK-SVE2p1-NEXT:  // %bb.1: // %for.body.preheader
+; CHECK-SVE2p1-NEXT:    cntw x10
+; CHECK-SVE2p1-NEXT:    mov w8, w2
+; CHECK-SVE2p1-NEXT:    fmov z0.d, #3.00000000
+; CHECK-SVE2p1-NEXT:    mov x9, xzr
+; CHECK-SVE2p1-NEXT:    mov x15, xzr
+; CHECK-SVE2p1-NEXT:    whilelo { p2.d, p3.d }, xzr, x8
+; CHECK-SVE2p1-NEXT:    addvl x11, x0, #2
+; CHECK-SVE2p1-NEXT:    whilelo { p0.d, p1.d }, x10, x8
+; CHECK-SVE2p1-NEXT:    addvl x10, x0, #3
+; CHECK-SVE2p1-NEXT:    addvl x12, x0, #1
+; CHECK-SVE2p1-NEXT:    addvl x13, x1, #3
+; CHECK-SVE2p1-NEXT:    addvl x14, x1, #2
+; CHECK-SVE2p1-NEXT:    addvl x16, x1, #1
+; CHECK-SVE2p1-NEXT:    .p2align 5, , 16
+; CHECK-SVE2p1-NEXT:  .LBB8_2: // %vector.body
+; CHECK-SVE2p1-NEXT:    // =>This Inner Loop Header: Depth=1
+; CHECK-SVE2p1-NEXT:    ld1d { z1.d }, p2/z, [x1, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    ld1d { z2.d }, p3/z, [x16, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    inch x15
+; CHECK-SVE2p1-NEXT:    fmul z1.d, z1.d, z0.d
+; CHECK-SVE2p1-NEXT:    ld1d { z3.d }, p0/z, [x14, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    ld1d { z4.d }, p1/z, [x13, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    fmul z2.d, z2.d, z0.d
+; CHECK-SVE2p1-NEXT:    fmul z3.d, z3.d, z0.d
+; CHECK-SVE2p1-NEXT:    fmul z4.d, z4.d, z0.d
+; CHECK-SVE2p1-NEXT:    st1d { z1.d }, p2, [x0, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    st1d { z2.d }, p3, [x12, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    whilelo { p2.d, p3.d }, x15, x8
+; CHECK-SVE2p1-NEXT:    mov z1.d, p2/z, #1 // =0x1
+; CHECK-SVE2p1-NEXT:    st1d { z3.d }, p0, [x11, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    st1d { z4.d }, p1, [x10, x9, lsl #3]
+; CHECK-SVE2p1-NEXT:    incw x9, all, mul #3
+; CHECK-SVE2p1-NEXT:    whilelo { p0.d, p1.d }, x9, x8
+; CHECK-SVE2p1-NEXT:    mov x9, x15
+; CHECK-SVE2p1-NEXT:    fmov x17, d1
+; CHECK-SVE2p1-NEXT:    tbnz w17, #0, .LBB8_2
+; CHECK-SVE2p1-NEXT:  .LBB8_3: // %for.cond.cleanup
+; CHECK-SVE2p1-NEXT:    ret
+entry:
+  %cmp6 = icmp sgt i32 %n, 0
+  br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  %0 = tail call i64 @llvm.vscale.i64()
+  %1 = shl nuw nsw i64 %0, 3
+  %2 = tail call i64 @llvm.vscale.i64()
+  %3 = shl nuw nsw i64 %2, 2
+  %active.lane.mask.entry = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 %3, i64 %wide.trip.count)
+  %active.lane.mask.entry3 = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 %wide.trip.count)
+  %active.lane.mask.entry4 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.entry, i64 2)
+  %active.lane.mask.entry5 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.entry, i64 0)
+  %active.lane.mask.entry6 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.entry3, i64 2)
+  %active.lane.mask.entry7 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.entry3, i64 0)
+  %4 = tail call i64 @llvm.vscale.i64()
+  %5 = shl nuw nsw i64 %4, 1
+  %6 = tail call i64 @llvm.vscale.i64()
+  %7 = shl nuw nsw i64 %6, 2
+  %8 = tail call i64 @llvm.vscale.i64()
+  %9 = mul nuw nsw i64 %8, 6
+  %10 = tail call i64 @llvm.vscale.i64()
+  %11 = shl nuw nsw i64 %10, 1
+  %12 = tail call i64 @llvm.vscale.i64()
+  %13 = shl nuw nsw i64 %12, 2
+  %14 = tail call i64 @llvm.vscale.i64()
+  %15 = mul nuw nsw i64 %14, 6
+  %16 = tail call i64 @llvm.vscale.i64()
+  %17 = shl nuw nsw i64 %16, 2
+  br label %vector.body
+
+vector.body:
+  %index = phi i64 [ 0, %for.body.preheader ], [ %index.next16, %vector.body ]
+  %active.lane.mask = phi <vscale x 2 x i1> [ %active.lane.mask.entry7, %for.body.preheader ], [ %active.lane.mask.next21, %vector.body ]
+  %active.lane.mask8 = phi <vscale x 2 x i1> [ %active.lane.mask.entry6, %for.body.preheader ], [ %active.lane.mask.next20, %vector.body ]
+  %active.lane.mask9 = phi <vscale x 2 x i1> [ %active.lane.mask.entry5, %for.body.preheader ], [ %active.lane.mask.next19, %vector.body ]
+  %active.lane.mask10 = phi <vscale x 2 x i1> [ %active.lane.mask.entry4, %for.body.preheader ], [ %active.lane.mask.next18, %vector.body ]
+  %18 = getelementptr inbounds double, ptr %src, i64 %index
+  %wide.masked.load = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr %18, i32 8, <vscale x 2 x i1> %active.lane.mask, <vscale x 2 x double> poison)
+  %19 = getelementptr inbounds double, ptr %18, i64 %5
+  %wide.masked.load11 = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull %19, i32 8, <vscale x 2 x i1> %active.lane.mask8, <vscale x 2 x double> poison)
+  %20 = getelementptr inbounds double, ptr %18, i64 %7
+  %wide.masked.load12 = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull %20, i32 8, <vscale x 2 x i1> %active.lane.mask9, <vscale x 2 x double> poison)
+  %21 = getelementptr inbounds double, ptr %18, i64 %9
+  %wide.masked.load13 = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull %21, i32 8, <vscale x 2 x i1> %active.lane.mask10, <vscale x 2 x double> poison)
+  %22 = fmul <vscale x 2 x double> %wide.masked.load, shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+  %23 = fmul <vscale x 2 x double> %wide.masked.load11, shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+  %24 = fmul <vscale x 2 x double> %wide.masked.load12, shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+  %25 = fmul <vscale x 2 x double> %wide.masked.load13, shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+  %26 = getelementptr inbounds double, ptr %dst, i64 %index
+  tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> %22, ptr %26, i32 8, <vscale x 2 x i1> %active.lane.mask)
+  %27 = getelementptr inbounds double, ptr %26, i64 %11
+  tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> %23, ptr %27, i32 8, <vscale x 2 x i1> %active.lane.mask8)
+  %28 = getelementptr inbounds double, ptr %26, i64 %13
+  tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> %24, ptr %28, i32 8, <vscale x 2 x i1> %active.lane.mask9)
+  %29 = getelementptr inbounds double, ptr %26, i64 %15
+  tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> %25, ptr %29, i32 8, <vscale x 2 x i1> %active.lane.mask10)
+  %index.next = add i64 %index, %1
+  %index.next16 = add i64 %index, %1
+  %30 = add i64 %index.next, %17
+  %active.lane.mask.next = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 %30, i64 %wide.trip.count)
+  %active.lane.mask.next17 = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 %index.next, i64 %wide.trip.count)
+  %active.lane.mask.next18 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.next, i64 2)
+  %active.lane.mask.next19 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.next, i64 0)
+  %active.lane.mask.next20 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.next17, i64 2)
+  %active.lane.mask.next21 = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> %active.lane.mask.next17, i64 0)
+  %31 = extractelement <vscale x 2 x i1> %active.lane.mask.next21, i64 0
+  br i1 %31, label %vector.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+}
+
+declare <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64, i64)
+declare <vscale x 16 x i1> @llvm.vector.extract.nxv16i1.nxv32i1(<vscale x 32 x i1>, i64 immarg)
+declare <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nocapture, i32 immarg, <vscale x 16 x i1>, <vscale x 16 x i8>)
+declare <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nocapture, i32 immarg, <vscale x 2 x i1>, <vscale x 2 x double>)
+declare <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64, i64)
+declare <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1>, i64 immarg)
+declare <vscale x 32 x i1> @llvm.get.active.lane.mask.nxv32i1.i64(i64, i64)
+declare <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64, i64)
+declare i64 @llvm.vscale.i64()
+declare void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8>, ptr nocapture, i32 immarg, <vscale x 16 x i1>)
+declare void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double>, ptr nocapture, i32 immarg, <vscale x 2 x i1>)
+
+attributes #0 = { nounwind vscale_range(1,16) "target-cpu"="neoverse-v1" }
+
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll b/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
index 200c2adcf0e66..a36fce5f8f690 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
@@ -93,38 +93,35 @@ define void @test_widen(ptr noalias %a, ptr readnone %b) #4 {
 ; TFA_INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
 ; TFA_INTERLEAVE-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; TFA_INTERLEAVE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
-; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP8]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
 ; TFA_INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; TFA_INTERLEAVE:       vector.body:
 ; TFA_INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP9]], i64 [[TMP11]]
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP9]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP12]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD3]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP15]], i64 [[TMP17]]
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP13]], ptr [[TMP15]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP14]], ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], [[TMP6]]
-; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX_NEXT]], [[TMP20]]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP21]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP23:%.*]] = extractelement <vscale x 2 x i1> [[TMP22]], i32 0
-; TFA_INTERLEAVE-NEXT:    br i1 [[TMP23]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[TMP7]], i64 [[TMP9]]
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP7]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD4]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[TMP13]], i64 [[TMP15]]
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP11]], ptr [[TMP13]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP12]], ptr [[TMP16]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT6]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = extractelement <vscale x 2 x i1> [[TMP17]], i32 0
+; TFA_INTERLEAVE-NEXT:    br i1 [[TMP18]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; TFA_INTERLEAVE:       for.cond.cleanup:
 ; TFA_INTERLEAVE-NEXT:    ret void
 ;
@@ -224,13 +221,13 @@ define void @test_if_then(ptr noalias %a, ptr readnone %b) #4 {
 ; TFCOMMON-NEXT:    [[TMP11:%.*]] = xor <vscale x 2 x i1> [[TMP8]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
 ; TFCOMMON-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i1> zeroinitializer
 ; TFCOMMON-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP12]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP10]]
-; TFCOMMON-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFCOMMON-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP14]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFCOMMON-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
+; TFCOMMON-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP13]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
 ; TFCOMMON-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
 ; TFCOMMON-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
-; TFCOMMON-NEXT:    [[TMP15:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFCOMMON-NEXT:    [[TMP16:%.*]] = extractelement <vscale x 2 x i1> [[TMP15]], i32 0
-; TFCOMMON-NEXT:    br i1 [[TMP16]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; TFCOMMON-NEXT:    [[TMP14:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFCOMMON-NEXT:    [[TMP15:%.*]] = extractelement <vscale x 2 x i1> [[TMP14]], i32 0
+; TFCOMMON-NEXT:    br i1 [[TMP15]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; TFCOMMON:       for.cond.cleanup:
 ; TFCOMMON-NEXT:    ret void
 ;
@@ -246,48 +243,45 @@ define void @test_if_then(ptr noalias %a, ptr readnone %b) #4 {
 ; TFA_INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
 ; TFA_INTERLEAVE-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; TFA_INTERLEAVE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
-; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP8]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
 ; TFA_INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; TFA_INTERLEAVE:       vector.body:
 ; TFA_INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP9]], i64 [[TMP11]]
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP9]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP12]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD3]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP13]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i1> [[TMP14]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP15]])
-; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD3]], <vscale x 2 x i1> [[TMP16]])
-; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = xor <vscale x 2 x i1> [[TMP13]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = xor <vscale x 2 x i1> [[TMP14]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP19]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i1> [[TMP20]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP21]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP17]]
-; TFA_INTERLEAVE-NEXT:    [[PREDPHI4:%.*]] = select <vscale x 2 x i1> [[TMP22]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP18]]
-; TFA_INTERLEAVE-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP26:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP27:%.*]] = mul i64 [[TMP26]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i64, ptr [[TMP25]], i64 [[TMP27]]
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP25]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI4]], ptr [[TMP28]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], [[TMP6]]
-; TFA_INTERLEAVE-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX_NEXT]], [[TMP30]]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP31]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[TMP32:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP33:%.*]] = extractelement <vscale x 2 x i1> [[TMP32]], i32 0
-; TFA_INTERLEAVE-NEXT:    br i1 [[TMP33]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i64 [[TMP9]]
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP7]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD4]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i1> [[TMP12]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP13]])
+; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD4]], <vscale x 2 x i1> [[TMP14]])
+; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = xor <vscale x 2 x i1> [[TMP11]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = xor <vscale x 2 x i1> [[TMP12]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP17]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i1> [[TMP18]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP19]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP15]]
+; TFA_INTERLEAVE-NEXT:    [[PREDPHI5:%.*]] = select <vscale x 2 x i1> [[TMP20]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP16]]
+; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP23:%.*]] = mul i64 [[TMP22]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i64, ptr [[TMP21]], i64 [[TMP23]]
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP21]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI5]], ptr [[TMP24]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT7]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; TFA_INTERLEAVE-NEXT:    [[TMP25:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT7]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP26:%.*]] = extractelement <vscale x 2 x i1> [[TMP25]], i32 0
+; TFA_INTERLEAVE-NEXT:    br i1 [[TMP26]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; TFA_INTERLEAVE:       for.cond.cleanup:
 ; TFA_INTERLEAVE-NEXT:    ret void
 ;
@@ -403,13 +397,13 @@ define void @test_widen_if_then_else(ptr noalias %a, ptr readnone %b) #4 {
 ; TFCOMMON-NEXT:    [[TMP12:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP8]], <vscale x 2 x i1> zeroinitializer
 ; TFCOMMON-NEXT:    [[TMP13:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP12]])
 ; TFCOMMON-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP10]], <vscale x 2 x i64> [[TMP11]], <vscale x 2 x i64> [[TMP13]]
-; TFCOMMON-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFCOMMON-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP15]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFCOMMON-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
+; TFCOMMON-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP14]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
 ; TFCOMMON-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
 ; TFCOMMON-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
-; TFCOMMON-NEXT:    [[TMP16:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFCOMMON-NEXT:    [[TMP17:%.*]] = extractelement <vscale x 2 x i1> [[TMP16]], i32 0
-; TFCOMMON-NEXT:    br i1 [[TMP17]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; TFCOMMON-NEXT:    [[TMP15:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFCOMMON-NEXT:    [[TMP16:%.*]] = extractelement <vscale x 2 x i1> [[TMP15]], i32 0
+; TFCOMMON-NEXT:    br i1 [[TMP16]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; TFCOMMON:       for.cond.cleanup:
 ; TFCOMMON-NEXT:    ret void
 ;
@@ -425,50 +419,47 @@ define void @test_widen_if_then_else(ptr noalias %a, ptr readnone %b) #4 {
 ; TFA_INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
 ; TFA_INTERLEAVE-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; TFA_INTERLEAVE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
-; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP8]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
 ; TFA_INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; TFA_INTERLEAVE:       vector.body:
 ; TFA_INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP9]], i64 [[TMP11]]
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP9]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP12]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD3]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = xor <vscale x 2 x i1> [[TMP13]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = xor <vscale x 2 x i1> [[TMP14]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP15]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i1> [[TMP16]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> [[TMP17]])
-; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> [[TMP18]])
-; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP13]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i1> [[TMP14]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT:    [[TMP23:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP21]])
-; TFA_INTERLEAVE-NEXT:    [[TMP24:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD3]], <vscale x 2 x i1> [[TMP22]])
-; TFA_INTERLEAVE-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP17]], <vscale x 2 x i64> [[TMP19]], <vscale x 2 x i64> [[TMP23]]
-; TFA_INTERLEAVE-NEXT:    [[PREDPHI4:%.*]] = select <vscale x 2 x i1> [[TMP18]], <vscale x 2 x i64> [[TMP20]], <vscale x 2 x i64> [[TMP24]]
-; TFA_INTERLEAVE-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP29:%.*]] = mul i64 [[TMP28]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i64, ptr [[TMP27]], i64 [[TMP29]]
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP27]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI4]], ptr [[TMP30]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], [[TMP6]]
-; TFA_INTERLEAVE-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX_NEXT]], [[TMP32]]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP33]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[TMP34:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP35:%.*]] = extractelement <vscale x 2 x i1> [[TMP34]], i32 0
-; TFA_INTERLEAVE-NEXT:    br i1 [[TMP35]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i64 [[TMP9]]
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP7]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = icmp ugt <vscale x 2 x i64> [[WIDE_MASKED_LOAD4]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 50, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = xor <vscale x 2 x i1> [[TMP11]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = xor <vscale x 2 x i1> [[TMP12]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP13]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i1> [[TMP14]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> [[TMP15]])
+; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> [[TMP16]])
+; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i1> [[TMP12]], <vscale x 2 x i1> zeroinitializer
+; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP19]])
+; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD4]], <vscale x 2 x i1> [[TMP20]])
+; TFA_INTERLEAVE-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP15]], <vscale x 2 x i64> [[TMP17]], <vscale x 2 x i64> [[TMP21]]
+; TFA_INTERLEAVE-NEXT:    [[PREDPHI5:%.*]] = select <vscale x 2 x i1> [[TMP16]], <vscale x 2 x i64> [[TMP18]], <vscale x 2 x i64> [[TMP22]]
+; TFA_INTERLEAVE-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP25:%.*]] = mul i64 [[TMP24]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i64, ptr [[TMP23]], i64 [[TMP25]]
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP23]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI5]], ptr [[TMP26]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT7]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; TFA_INTERLEAVE-NEXT:    [[TMP27:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT7]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP28:%.*]] = extractelement <vscale x 2 x i1> [[TMP27]], i32 0
+; TFA_INTERLEAVE-NEXT:    br i1 [[TMP28]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; TFA_INTERLEAVE:       for.cond.cleanup:
 ; TFA_INTERLEAVE-NEXT:    ret void
 ;
@@ -756,38 +747,35 @@ define void @test_widen_optmask(ptr noalias %a, ptr readnone %b) #4 {
 ; TFA_INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
 ; TFA_INTERLEAVE-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; TFA_INTERLEAVE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
-; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP8]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
 ; TFA_INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; TFA_INTERLEAVE:       vector.body:
 ; TFA_INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP9]], i64 [[TMP11]]
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP9]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP12]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i64> poison)
-; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD3]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP15]], i64 [[TMP17]]
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP13]], ptr [[TMP15]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP14]], ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], [[TMP6]]
-; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX_NEXT]], [[TMP20]]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP21]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP23:%.*]] = extractelement <vscale x 2 x i1> [[TMP22]], i32 0
-; TFA_INTERLEAVE-NEXT:    br i1 [[TMP23]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[TMP7]], i64 [[TMP9]]
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP7]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x i64> poison)
+; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD4]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[TMP13]], i64 [[TMP15]]
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP11]], ptr [[TMP13]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP12]], ptr [[TMP16]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT6]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = extractelement <vscale x 2 x i1> [[TMP17]], i32 0
+; TFA_INTERLEAVE-NEXT:    br i1 [[TMP18]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; TFA_INTERLEAVE:       for.cond.cleanup:
 ; TFA_INTERLEAVE-NEXT:    ret void
 ;
@@ -952,51 +940,48 @@ define double @test_widen_fmuladd_and_call(ptr noalias %a, ptr readnone %b, doub
 ; TFA_INTERLEAVE-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
 ; TFA_INTERLEAVE-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; TFA_INTERLEAVE-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
-; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP8]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
 ; TFA_INTERLEAVE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x double> poison, double [[M:%.*]], i64 0
 ; TFA_INTERLEAVE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x double> [[BROADCAST_SPLATINSERT]], <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer
 ; TFA_INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; TFA_INTERLEAVE:       vector.body:
 ; TFA_INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[VEC_PHI:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TMP26:%.*]], [[VECTOR_BODY]] ]
-; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = getelementptr double, ptr [[TMP9]], i64 [[TMP11]]
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP9]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
-; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP12]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x double> poison)
-; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
-; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD3]], [[BROADCAST_SPLAT]]
-; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = fptoui <vscale x 2 x double> [[WIDE_MASKED_LOAD]] to <vscale x 2 x i64>
-; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = fptoui <vscale x 2 x double> [[WIDE_MASKED_LOAD3]] to <vscale x 2 x i64>
-; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[TMP15]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[TMP16]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = mul i64 [[TMP20]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i64, ptr [[TMP19]], i64 [[TMP21]]
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP17]], ptr [[TMP19]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP18]], ptr [[TMP22]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; TFA_INTERLEAVE-NEXT:    [[TMP23:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> [[TMP13]], <vscale x 2 x double> shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double -0.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP24:%.*]] = call double @llvm.vector.reduce.fadd.nxv2f64(double [[VEC_PHI]], <vscale x 2 x double> [[TMP23]])
-; TFA_INTERLEAVE-NEXT:    [[TMP25:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x double> [[TMP14]], <vscale x 2 x double> shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double -0.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP26]] = call double @llvm.vector.reduce.fadd.nxv2f64(double [[TMP24]], <vscale x 2 x double> [[TMP25]])
-; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], [[TMP6]]
-; TFA_INTERLEAVE-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
-; TFA_INTERLEAVE-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 2
-; TFA_INTERLEAVE-NEXT:    [[TMP29:%.*]] = add i64 [[INDEX_NEXT]], [[TMP28]]
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP29]], i64 1025)
-; TFA_INTERLEAVE-NEXT:    [[TMP30:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TFA_INTERLEAVE-NEXT:    [[TMP31:%.*]] = extractelement <vscale x 2 x i1> [[TMP30]], i32 0
-; TFA_INTERLEAVE-NEXT:    br i1 [[TMP31]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[VEC_PHI:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ], [ [[TMP24:%.*]], [[VECTOR_BODY]] ]
+; TFA_INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr double, ptr [[TMP7]], i64 [[TMP9]]
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP7]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; TFA_INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x double> poison)
+; TFA_INTERLEAVE-NEXT:    [[TMP11:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
+; TFA_INTERLEAVE-NEXT:    [[TMP12:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD4]], [[BROADCAST_SPLAT]]
+; TFA_INTERLEAVE-NEXT:    [[TMP13:%.*]] = fptoui <vscale x 2 x double> [[WIDE_MASKED_LOAD]] to <vscale x 2 x i64>
+; TFA_INTERLEAVE-NEXT:    [[TMP14:%.*]] = fptoui <vscale x 2 x double> [[WIDE_MASKED_LOAD4]] to <vscale x 2 x i64>
+; TFA_INTERLEAVE-NEXT:    [[TMP15:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[TMP13]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    [[TMP16:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[TMP14]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; TFA_INTERLEAVE-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; TFA_INTERLEAVE-NEXT:    [[TMP19:%.*]] = mul i64 [[TMP18]], 2
+; TFA_INTERLEAVE-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[TMP17]], i64 [[TMP19]]
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP15]], ptr [[TMP17]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP16]], ptr [[TMP20]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; TFA_INTERLEAVE-NEXT:    [[TMP21:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> [[TMP11]], <vscale x 2 x double> shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double -0.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP22:%.*]] = call double @llvm.vector.reduce.fadd.nxv2f64(double [[VEC_PHI]], <vscale x 2 x double> [[TMP21]])
+; TFA_INTERLEAVE-NEXT:    [[TMP23:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x double> [[TMP12]], <vscale x 2 x double> shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double -0.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP24]] = call double @llvm.vector.reduce.fadd.nxv2f64(double [[TMP22]], <vscale x 2 x double> [[TMP23]])
+; TFA_INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT]], i64 1025)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; TFA_INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; TFA_INTERLEAVE-NEXT:    [[TMP25:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT6]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
+; TFA_INTERLEAVE-NEXT:    [[TMP26:%.*]] = extractelement <vscale x 2 x i1> [[TMP25]], i32 0
+; TFA_INTERLEAVE-NEXT:    br i1 [[TMP26]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; TFA_INTERLEAVE:       for.cond.cleanup:
-; TFA_INTERLEAVE-NEXT:    ret double [[TMP26]]
+; TFA_INTERLEAVE-NEXT:    ret double [[TMP24]]
 ;
 entry:
   br label %for.body
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/pr73894.ll b/llvm/test/Transforms/LoopVectorize/AArch64/pr73894.ll
index 809d2e8f7ea1a..34283ea296712 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/pr73894.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/pr73894.ll
@@ -16,8 +16,8 @@ define i32 @pr70988() {
 ; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[UMAX]], 1
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 2
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = icmp ult i64 0, [[UMAX]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = icmp ult i64 1, [[UMAX]]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = icmp ult i64 0, [[UMAX]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE5:%.*]] ]
@@ -51,8 +51,8 @@ define i32 @pr70988() {
 ; CHECK-NEXT:    [[TMP18]] = select i1 [[ACTIVE_LANE_MASK2]], i32 [[TMP16]], i32 [[VEC_PHI3]]
 ; CHECK-NEXT:    [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX_NEXT]], 1
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = icmp ult i64 [[INDEX_NEXT]], [[UMAX]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT7]] = icmp ult i64 [[TMP19]], [[UMAX]]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = icmp ult i64 [[INDEX_NEXT]], [[UMAX]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = xor i1 [[ACTIVE_LANE_MASK_NEXT]], true
 ; CHECK-NEXT:    [[TMP21:%.*]] = xor i1 [[ACTIVE_LANE_MASK_NEXT7]], true
 ; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
index bcf8096f1b738..0ae81efdf7d70 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
@@ -42,22 +42,22 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 8
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 8
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP7]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP8]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP9]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP7]])
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP9]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
@@ -89,18 +89,18 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 8
 ; CHECK-ORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-ORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 8
+; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 8
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP7]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
-; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP8]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP9]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
+; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
@@ -108,7 +108,7 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -120,7 +120,7 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[ADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fadd_strict
@@ -136,27 +136,27 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 8
 ; CHECK-ORDERED-TF-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-ORDERED-TF-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = sub i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = icmp ugt i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i64 [[TMP9]], i64 0
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[N]])
 ; CHECK-ORDERED-TF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP10]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP12]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP14]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP13]])
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP16]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP12]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP14]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP16]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP15]])
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP11]])
 ; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
 ; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = extractelement <vscale x 8 x i1> [[TMP17]], i32 0
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -164,7 +164,7 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED-TF:       scalar.ph:
 ; CHECK-ORDERED-TF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP16]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       for.body:
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -176,7 +176,7 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP16]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[ADD_LCSSA]]
 ;
 
@@ -230,60 +230,60 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 32
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-UNORDERED-NEXT:    [[TMP38:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP39:%.*]] = mul i64 [[TMP38]], 32
+; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 32
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP34:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP36:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP37:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], [[TMP13]]
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = add i64 [[TMP16]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], [[TMP18]]
-; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP14]]
-; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP19]]
-; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP26]]
-; CHECK-UNORDERED-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP29:%.*]] = mul i64 [[TMP28]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP29]]
-; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP32]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP30]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP33]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP34]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
-; CHECK-UNORDERED-NEXT:    [[TMP35]] = fadd <vscale x 8 x float> [[WIDE_LOAD4]], [[VEC_PHI1]]
-; CHECK-UNORDERED-NEXT:    [[TMP36]] = fadd <vscale x 8 x float> [[WIDE_LOAD5]], [[VEC_PHI2]]
-; CHECK-UNORDERED-NEXT:    [[TMP37]] = fadd <vscale x 8 x float> [[WIDE_LOAD6]], [[VEC_PHI3]]
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP39]]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP36:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP37:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP38:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP39:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[TMP8]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]]
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[TMP13]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], [[TMP15]]
+; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[TMP18]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], [[TMP20]]
+; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP16]]
+; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP21]]
+; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i32 0
+; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP28]]
+; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = mul i64 [[TMP30]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP31]]
+; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP34:%.*]] = mul i64 [[TMP33]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP34]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP26]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP29]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP32]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP35]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP36]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-UNORDERED-NEXT:    [[TMP37]] = fadd <vscale x 8 x float> [[WIDE_LOAD4]], [[VEC_PHI1]]
+; CHECK-UNORDERED-NEXT:    [[TMP38]] = fadd <vscale x 8 x float> [[WIDE_LOAD5]], [[VEC_PHI2]]
+; CHECK-UNORDERED-NEXT:    [[TMP39]] = fadd <vscale x 8 x float> [[WIDE_LOAD6]], [[VEC_PHI3]]
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-UNORDERED-NEXT:    [[TMP40:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP35]], [[TMP34]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX7:%.*]] = fadd <vscale x 8 x float> [[TMP36]], [[BIN_RDX]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX8:%.*]] = fadd <vscale x 8 x float> [[TMP37]], [[BIN_RDX7]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP37]], [[TMP36]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX7:%.*]] = fadd <vscale x 8 x float> [[TMP38]], [[BIN_RDX]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX8:%.*]] = fadd <vscale x 8 x float> [[TMP39]], [[BIN_RDX7]]
 ; CHECK-UNORDERED-NEXT:    [[TMP41:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX8]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
@@ -316,51 +316,51 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 32
 ; CHECK-ORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-ORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP39:%.*]] = mul i64 [[TMP38]], 32
+; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 32
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP37:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 0
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 1
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 0
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 1
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], [[TMP13]]
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 24
-; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = add i64 [[TMP16]], 0
-; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 1
-; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], [[TMP18]]
-; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP14]]
-; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP19]]
-; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP26]]
-; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = mul i64 [[TMP28]], 16
-; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP29]]
-; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 24
-; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP32]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP30]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP33]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
-; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP34]], <vscale x 8 x float> [[WIDE_LOAD1]])
-; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP35]], <vscale x 8 x float> [[WIDE_LOAD2]])
-; CHECK-ORDERED-NEXT:    [[TMP37]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP36]], <vscale x 8 x float> [[WIDE_LOAD3]])
-; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP39]]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP39:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 8
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[TMP8]], 0
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 1
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]]
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 16
+; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[TMP13]], 0
+; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 1
+; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], [[TMP15]]
+; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 24
+; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[TMP18]], 0
+; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 1
+; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], [[TMP20]]
+; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP16]]
+; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP21]]
+; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i32 0
+; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
+; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP28]]
+; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = mul i64 [[TMP30]], 16
+; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP31]]
+; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = mul i64 [[TMP33]], 24
+; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP34]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP26]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP29]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP32]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP35]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
+; CHECK-ORDERED-NEXT:    [[TMP37:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP36]], <vscale x 8 x float> [[WIDE_LOAD1]])
+; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP37]], <vscale x 8 x float> [[WIDE_LOAD2]])
+; CHECK-ORDERED-NEXT:    [[TMP39]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP38]], <vscale x 8 x float> [[WIDE_LOAD3]])
+; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-ORDERED-NEXT:    [[TMP40:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
@@ -368,7 +368,7 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP37]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -380,7 +380,7 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP37]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[ADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fadd_strict_unroll
@@ -396,113 +396,117 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP69:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP70:%.*]] = mul i64 [[TMP69]], 32
 ; CHECK-ORDERED-TF-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-ORDERED-TF-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 32
-; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = sub i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP26]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 16
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP28]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 24
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP30]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT]], i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT2]], i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 32
+; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = icmp ugt i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i64 [[TMP9]], i64 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 8
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP13]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 16
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 24
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP17]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY6:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY7:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 0)
 ; CHECK-ORDERED-TF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT12:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT13:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT14:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP68:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = mul i64 [[TMP32]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = add i64 [[TMP33]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = mul i64 [[TMP34]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = add i64 [[INDEX]], [[TMP35]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = mul i64 [[TMP37]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = add i64 [[TMP38]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = add i64 [[INDEX]], [[TMP40]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY7]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY6]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK10:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP55:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = add i64 [[TMP20]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = add i64 [[INDEX]], [[TMP22]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = mul i64 [[TMP24]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = add i64 [[TMP25]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = mul i64 [[TMP26]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = add i64 [[INDEX]], [[TMP27]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = add i64 [[TMP30]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX]], [[TMP32]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP23]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP28]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP33]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP40]]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = add i64 [[TMP43]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = add i64 [[INDEX]], [[TMP45]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP31]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP41]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP46]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = mul i64 [[TMP52]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP53]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = mul i64 [[TMP55]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP56]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = mul i64 [[TMP58]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP59]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP51]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP54]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP57]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP60]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP61]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP63:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP64:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP62]], <vscale x 8 x float> [[TMP63]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP65:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP66:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP64]], <vscale x 8 x float> [[TMP65]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP67:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP68]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP66]], <vscale x 8 x float> [[TMP67]])
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP70]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP71:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP72:%.*]] = mul i64 [[TMP71]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP73:%.*]] = add i64 [[INDEX]], [[TMP72]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP74:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP75:%.*]] = mul i64 [[TMP74]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP76:%.*]] = add i64 [[INDEX]], [[TMP75]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP77:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP78:%.*]] = mul i64 [[TMP77]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP79:%.*]] = add i64 [[INDEX]], [[TMP78]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT12]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP73]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT13]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP76]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT14]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP79]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP80:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP81:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT12]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP82:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT13]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP83:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT14]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP84:%.*]] = extractelement <vscale x 8 x i1> [[TMP80]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP84]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP43]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = mul i64 [[TMP45]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP46]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP38]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP41]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP44]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP47]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP48]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP49]], <vscale x 8 x float> [[TMP50]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> [[WIDE_MASKED_LOAD12]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP51]], <vscale x 8 x float> [[TMP52]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> [[WIDE_MASKED_LOAD13]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP55]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP53]], <vscale x 8 x float> [[TMP54]])
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = mul i64 [[TMP56]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = add i64 [[INDEX]], [[TMP57]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = mul i64 [[TMP59]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = add i64 [[INDEX]], [[TMP60]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP63:%.*]] = mul i64 [[TMP62]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP64:%.*]] = add i64 [[INDEX]], [[TMP63]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP61]], i64 [[TMP11]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT14:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP11]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT15]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT14]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT14]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[TMP65:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT18]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP66:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT17]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP67:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT16]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP68:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT15]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP69:%.*]] = extractelement <vscale x 8 x i1> [[TMP65]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP69]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED-TF:       scalar.ph:
 ; CHECK-ORDERED-TF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP68]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP55]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       for.body:
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP85:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[ADD]] = fadd float [[TMP85]], [[SUM_07]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP70:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[ADD]] = fadd float [[TMP70]], [[SUM_07]]
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP68]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP55]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[ADD_LCSSA]]
 ;
 
@@ -574,37 +578,37 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], [[TMP6]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
 ; CHECK-UNORDERED-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 4
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = insertelement <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float [[A2]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = insertelement <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float [[A1]], i32 0
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 4
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = insertelement <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float [[A2]], i32 0
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = insertelement <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float [[A1]], i32 0
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP7]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x float> [ [[TMP8]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP9]], [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x float> [ [[TMP10]], [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[OFFSET_IDX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_VEC:%.*]] = load <vscale x 8 x float>, ptr [[TMP11]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
+; CHECK-UNORDERED-NEXT:    [[WIDE_VEC:%.*]] = load <vscale x 8 x float>, ptr [[TMP13]], align 4
 ; CHECK-UNORDERED-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.vector.deinterleave2.nxv8f32(<vscale x 8 x float> [[WIDE_VEC]])
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP14]] = fadd <vscale x 4 x float> [[TMP12]], [[VEC_PHI1]]
-; CHECK-UNORDERED-NEXT:    [[TMP15]] = fadd <vscale x 4 x float> [[TMP13]], [[VEC_PHI]]
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP17]]
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP16]] = fadd <vscale x 4 x float> [[TMP14]], [[VEC_PHI1]]
+; CHECK-UNORDERED-NEXT:    [[TMP17]] = fadd <vscale x 4 x float> [[TMP15]], [[VEC_PHI]]
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
 ; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP15]])
-; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP14]])
+; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP17]])
+; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP16]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[A2]], [[ENTRY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi float [ [[A1]], [[ENTRY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[A2]], [[ENTRY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi float [ [[A1]], [[ENTRY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[ADD_PHI1:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD2:%.*]], [[FOR_BODY]] ]
@@ -621,8 +625,8 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK-UNORDERED:       for.end:
-; CHECK-UNORDERED-NEXT:    [[ADD1_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
-; CHECK-UNORDERED-NEXT:    [[ADD2_LCSSA:%.*]] = phi float [ [[ADD2]], [[FOR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[ADD1_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[ADD2_LCSSA:%.*]] = phi float [ [[ADD2]], [[FOR_BODY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    store float [[ADD1_LCSSA]], ptr [[A]], align 4
 ; CHECK-UNORDERED-NEXT:    store float [[ADD2_LCSSA]], ptr [[ARRAYIDXA]], align 4
 ; CHECK-UNORDERED-NEXT:    ret void
@@ -646,24 +650,24 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-ORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], [[TMP6]]
 ; CHECK-ORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
 ; CHECK-ORDERED-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 4
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 4
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[A2]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi float [ [[A1]], [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[A2]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi float [ [[A1]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 0
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP7]]
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_VEC:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP9]]
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
+; CHECK-ORDERED-NEXT:    [[WIDE_VEC:%.*]] = load <vscale x 8 x float>, ptr [[TMP11]], align 4
 ; CHECK-ORDERED-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.vector.deinterleave2.nxv8f32(<vscale x 8 x float> [[WIDE_VEC]])
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 0
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 1
-; CHECK-ORDERED-NEXT:    [[TMP12]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP11]])
-; CHECK-ORDERED-NEXT:    [[TMP13]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI1]], <vscale x 4 x float> [[TMP10]])
-; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP15]]
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 0
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 1
+; CHECK-ORDERED-NEXT:    [[TMP14]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP13]])
+; CHECK-ORDERED-NEXT:    [[TMP15]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI1]], <vscale x 4 x float> [[TMP12]])
+; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
 ; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
@@ -671,8 +675,8 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[A2]], [[ENTRY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi float [ [[A1]], [[ENTRY]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[A2]], [[ENTRY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi float [ [[A1]], [[ENTRY]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[ADD_PHI1:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD2:%.*]], [[FOR_BODY]] ]
@@ -689,8 +693,8 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[ADD1_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
-; CHECK-ORDERED-NEXT:    [[ADD2_LCSSA:%.*]] = phi float [ [[ADD2]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[ADD1_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[ADD2_LCSSA:%.*]] = phi float [ [[ADD2]], [[FOR_BODY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    store float [[ADD1_LCSSA]], ptr [[A]], align 4
 ; CHECK-ORDERED-NEXT:    store float [[ADD2_LCSSA]], ptr [[ARRAYIDXA]], align 4
 ; CHECK-ORDERED-NEXT:    ret void
@@ -715,35 +719,35 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-ORDERED-TF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
 ; CHECK-ORDERED-TF-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
-; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = mul i64 [[TMP22]], 4
 ; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 4
-; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = sub i64 [[TMP2]], [[TMP9]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = icmp ugt i64 [[TMP2]], [[TMP9]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i64 [[TMP10]], i64 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP2]], [[TMP11]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = icmp ugt i64 [[TMP2]], [[TMP11]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i64 [[TMP12]], i64 0
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[TMP2]])
 ; CHECK-ORDERED-TF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[A2]], [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI1:%.*]] = phi float [ [[A1]], [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[A2]], [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI1:%.*]] = phi float [ [[A1]], [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
-; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = add i64 [[OFFSET_IDX]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP13]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
 ; CHECK-ORDERED-TF-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 8 x i1> @llvm.vector.interleave2.nxv8i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP15]], i32 4, <vscale x 8 x i1> [[INTERLEAVED_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP17]], i32 4, <vscale x 8 x i1> [[INTERLEAVED_MASK]], <vscale x 8 x float> poison)
 ; CHECK-ORDERED-TF-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.vector.deinterleave2.nxv8f32(<vscale x 8 x float> [[WIDE_MASKED_VEC]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP17]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP19]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP18]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP16]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP21]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI1]], <vscale x 4 x float> [[TMP20]])
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP23]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP12]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } [[STRIDED_VEC]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP19]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP21]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP20]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP18]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP23]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI1]], <vscale x 4 x float> [[TMP22]])
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP9]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP14]])
 ; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = extractelement <vscale x 4 x i1> [[TMP24]], i32 0
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -751,8 +755,8 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-ORDERED-TF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED-TF:       scalar.ph:
 ; CHECK-ORDERED-TF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[A2]], [[ENTRY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi float [ [[A1]], [[ENTRY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[A2]], [[ENTRY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi float [ [[A1]], [[ENTRY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       for.body:
 ; CHECK-ORDERED-TF-NEXT:    [[ADD_PHI1:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD2:%.*]], [[FOR_BODY]] ]
@@ -769,8 +773,8 @@ define void @fadd_strict_interleave(ptr noalias nocapture readonly %a, ptr noali
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[ADD1_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ADD2_LCSSA:%.*]] = phi float [ [[ADD2]], [[FOR_BODY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ADD1_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ADD2_LCSSA:%.*]] = phi float [ [[ADD2]], [[FOR_BODY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    store float [[ADD1_LCSSA]], ptr [[A]], align 4
 ; CHECK-ORDERED-TF-NEXT:    store float [[ADD2_LCSSA]], ptr [[ARRAYIDXA]], align 4
 ; CHECK-ORDERED-TF-NEXT:    ret void
@@ -852,26 +856,26 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = mul i64 [[TMP3]], 4
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP4]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]]
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP5]]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP7]]
 ; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; CHECK-UNORDERED-NEXT:    [[TMP11]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[TMP10]]
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP13]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP7]]
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; CHECK-UNORDERED-NEXT:    [[TMP13]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[TMP12]]
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP11]])
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP13]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
@@ -914,22 +918,22 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = mul i64 [[TMP3]], 4
 ; CHECK-ORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP4]]
 ; CHECK-ORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]]
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP5]]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP7]]
 ; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; CHECK-ORDERED-NEXT:    [[TMP11]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP10]])
-; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP13]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP7]]
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; CHECK-ORDERED-NEXT:    [[TMP13]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP12]])
+; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
@@ -937,7 +941,7 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -952,7 +956,7 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK-ORDERED:       for.end.loopexit:
-; CHECK-ORDERED-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_END]]
 ; CHECK-ORDERED:       for.end:
 ; CHECK-ORDERED-NEXT:    [[RES:%.*]] = phi float [ 0.000000e+00, [[ENTRY:%.*]] ], [ [[RDX_LCSSA]], [[FOR_END_LOOPEXIT]] ]
@@ -976,31 +980,31 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-TF-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP5]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP2]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 4
 ; CHECK-ORDERED-TF-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 4
-; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = sub i64 [[N]], [[TMP7]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = icmp ugt i64 [[N]], [[TMP7]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = select i1 [[TMP9]], i64 [[TMP8]], i64 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = sub i64 [[N]], [[TMP9]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = icmp ugt i64 [[N]], [[TMP9]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = select i1 [[TMP11]], i64 [[TMP10]], i64 0
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[N]])
 ; CHECK-ORDERED-TF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP11]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP13]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP11]]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP13]]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP15]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = fadd <vscale x 4 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD1]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP16]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP18]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP17]])
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP20]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP10]])
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP15]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP13]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP17]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = fadd <vscale x 4 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD1]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP18]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP20]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP19]])
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP7]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP12]])
 ; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = extractelement <vscale x 4 x i1> [[TMP21]], i32 0
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -1008,7 +1012,7 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-TF-NEXT:    br i1 true, label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED-TF:       scalar.ph:
 ; CHECK-ORDERED-TF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       for.body:
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1023,7 +1027,7 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end.loopexit:
-; CHECK-ORDERED-TF-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END]]
 ; CHECK-ORDERED-TF:       for.end:
 ; CHECK-ORDERED-TF-NEXT:    [[RES:%.*]] = phi float [ 0.000000e+00, [[ENTRY:%.*]] ], [ [[RDX_LCSSA]], [[FOR_END_LOOPEXIT]] ]
@@ -1095,28 +1099,28 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-UNORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float 1.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr float, ptr [[TMP8]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP9]], i32 4, <vscale x 4 x i1> [[TMP7]], <vscale x 4 x float> poison)
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = xor <vscale x 4 x i1> [[TMP7]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-UNORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP10]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x float> [[WIDE_MASKED_LOAD]]
-; CHECK-UNORDERED-NEXT:    [[TMP11]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[PREDPHI]]
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP13]]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), float 1.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[A]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[TMP10]], i32 0
+; CHECK-UNORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> poison)
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = xor <vscale x 4 x i1> [[TMP9]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-UNORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP12]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x float> [[WIDE_MASKED_LOAD]]
+; CHECK-UNORDERED-NEXT:    [[TMP13]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[PREDPHI]]
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP11]])
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP13]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
@@ -1156,24 +1160,24 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-ORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 4
 ; CHECK-ORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-ORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr float, ptr [[TMP8]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP9]], i32 4, <vscale x 4 x i1> [[TMP7]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = xor <vscale x 4 x i1> [[TMP7]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-ORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP10]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x float> [[WIDE_MASKED_LOAD]]
-; CHECK-ORDERED-NEXT:    [[TMP11]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[PREDPHI]])
-; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP13]]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[A]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[TMP10]], i32 0
+; CHECK-ORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = xor <vscale x 4 x i1> [[TMP9]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP12]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x float> [[WIDE_MASKED_LOAD]]
+; CHECK-ORDERED-NEXT:    [[TMP13]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[PREDPHI]])
+; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
@@ -1181,7 +1185,7 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 1.000000e+00, [[ENTRY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 1.000000e+00, [[ENTRY]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -1201,7 +1205,7 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[RDX]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fadd_conditional
@@ -1217,63 +1221,63 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-ORDERED-TF-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = mul i64 [[TMP22]], 4
 ; CHECK-ORDERED-TF-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-ORDERED-TF-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
-; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = sub i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = icmp ugt i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i64 [[TMP9]], i64 0
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[N]])
 ; CHECK-ORDERED-TF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP10]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_MASKED_LOAD]], zeroinitializer
-; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> zeroinitializer
-; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr float, ptr [[A]], i64 [[TMP10]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = getelementptr float, ptr [[TMP14]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP16]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = xor <vscale x 4 x i1> [[TMP13]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP17]], <vscale x 4 x i1> zeroinitializer
-; CHECK-ORDERED-TF-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP18]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x float> [[WIDE_MASKED_LOAD1]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[PREDPHI]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP21]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP20]])
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP23]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = extractelement <vscale x 4 x i1> [[TMP24]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP12]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_MASKED_LOAD]], zeroinitializer
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i1> zeroinitializer
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = getelementptr float, ptr [[A]], i64 [[TMP12]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = getelementptr float, ptr [[TMP17]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP18]], i32 4, <vscale x 4 x i1> [[TMP16]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = xor <vscale x 4 x i1> [[TMP15]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP19]], <vscale x 4 x i1> zeroinitializer
+; CHECK-ORDERED-TF-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP20]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x float> [[WIDE_MASKED_LOAD1]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[PREDPHI]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP22]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP21]])
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP11]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = extractelement <vscale x 4 x i1> [[TMP23]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED-TF:       scalar.ph:
 ; CHECK-ORDERED-TF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 1.000000e+00, [[ENTRY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 1.000000e+00, [[ENTRY]] ], [ [[TMP22]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       for.body:
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[RES:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[FADD:%.*]], [[FOR_INC]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP26]], 0.000000e+00
+; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP25]], 0.000000e+00
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[FOR_INC]]
 ; CHECK-ORDERED-TF:       if.then:
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_INC]]
 ; CHECK-ORDERED-TF:       for.inc:
-; CHECK-ORDERED-TF-NEXT:    [[PHI:%.*]] = phi float [ [[TMP27]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[PHI:%.*]] = phi float [ [[TMP26]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[FADD]] = fadd float [[RES]], [[PHI]]
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP22]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[RDX]]
 ;
 
@@ -1342,26 +1346,26 @@ define float @fadd_multiple(ptr noalias nocapture %a, ptr noalias nocapture %b,
 ; CHECK-UNORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 8
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = mul i64 [[TMP11]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 8
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float -0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = fadd <vscale x 8 x float> [[VEC_PHI]], [[WIDE_LOAD]]
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP10]] = fadd <vscale x 8 x float> [[TMP7]], [[WIDE_LOAD1]]
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float -0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP8]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = fadd <vscale x 8 x float> [[VEC_PHI]], [[WIDE_LOAD]]
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP11]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP12]] = fadd <vscale x 8 x float> [[TMP9]], [[WIDE_LOAD1]]
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP10]])
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP12]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
@@ -1482,78 +1486,78 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 32
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-UNORDERED-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP53:%.*]] = mul i64 [[TMP52]], 32
+; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 32
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP50:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP51:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], [[TMP13]]
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = add i64 [[TMP16]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], [[TMP18]]
-; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP14]]
-; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP19]]
-; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP26]]
-; CHECK-UNORDERED-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP29:%.*]] = mul i64 [[TMP28]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP29]]
-; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP32]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP30]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP33]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP14]]
-; CHECK-UNORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP19]]
-; CHECK-UNORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP40]]
-; CHECK-UNORDERED-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP43]]
-; CHECK-UNORDERED-NEXT:    [[TMP45:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP46:%.*]] = mul i64 [[TMP45]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP46]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP38]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP41]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP44]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP47]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP48]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
-; CHECK-UNORDERED-NEXT:    [[TMP49]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
-; CHECK-UNORDERED-NEXT:    [[TMP50]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
-; CHECK-UNORDERED-NEXT:    [[TMP51]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP53]]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP50:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP51:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP52:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP53:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[TMP8]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]]
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[TMP13]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], [[TMP15]]
+; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[TMP18]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], [[TMP20]]
+; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP16]]
+; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP21]]
+; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i32 0
+; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP28]]
+; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = mul i64 [[TMP30]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP31]]
+; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP34:%.*]] = mul i64 [[TMP33]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP34]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP26]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP29]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP32]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP35]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP16]]
+; CHECK-UNORDERED-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP21]]
+; CHECK-UNORDERED-NEXT:    [[TMP40:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i32 0
+; CHECK-UNORDERED-NEXT:    [[TMP41:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP41]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP42]]
+; CHECK-UNORDERED-NEXT:    [[TMP44:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP46:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP45]]
+; CHECK-UNORDERED-NEXT:    [[TMP47:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP48:%.*]] = mul i64 [[TMP47]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP48]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP40]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP43]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP46]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP49]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP50]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
+; CHECK-UNORDERED-NEXT:    [[TMP51]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
+; CHECK-UNORDERED-NEXT:    [[TMP52]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
+; CHECK-UNORDERED-NEXT:    [[TMP53]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-UNORDERED-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP54]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP49]], [[TMP48]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd <vscale x 8 x float> [[TMP50]], [[BIN_RDX]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd <vscale x 8 x float> [[TMP51]], [[BIN_RDX11]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP51]], [[TMP50]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd <vscale x 8 x float> [[TMP52]], [[BIN_RDX]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd <vscale x 8 x float> [[TMP53]], [[BIN_RDX11]]
 ; CHECK-UNORDERED-NEXT:    [[TMP55:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX12]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
@@ -1588,73 +1592,73 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 32
 ; CHECK-ORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-ORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-ORDERED-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP57:%.*]] = mul i64 [[TMP56]], 32
+; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 32
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP55:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 0
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 1
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 0
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 1
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], [[TMP13]]
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 24
-; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = add i64 [[TMP16]], 0
-; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 1
-; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], [[TMP18]]
-; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP14]]
-; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP19]]
-; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP26]]
-; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = mul i64 [[TMP28]], 16
-; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP29]]
-; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 24
-; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP32]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP30]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP33]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP14]]
-; CHECK-ORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP19]]
-; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
-; CHECK-ORDERED-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP40]]
-; CHECK-ORDERED-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 16
-; CHECK-ORDERED-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP43]]
-; CHECK-ORDERED-NEXT:    [[TMP45:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP46:%.*]] = mul i64 [[TMP45]], 24
-; CHECK-ORDERED-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP46]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP38]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP41]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP44]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP47]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP48:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
-; CHECK-ORDERED-NEXT:    [[TMP49:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
-; CHECK-ORDERED-NEXT:    [[TMP50:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
-; CHECK-ORDERED-NEXT:    [[TMP51:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; CHECK-ORDERED-NEXT:    [[TMP52:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP48]])
-; CHECK-ORDERED-NEXT:    [[TMP53:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP52]], <vscale x 8 x float> [[TMP49]])
-; CHECK-ORDERED-NEXT:    [[TMP54:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP53]], <vscale x 8 x float> [[TMP50]])
-; CHECK-ORDERED-NEXT:    [[TMP55]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP54]], <vscale x 8 x float> [[TMP51]])
-; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP57]]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP57:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 8
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[TMP8]], 0
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 1
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]]
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 16
+; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[TMP13]], 0
+; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 1
+; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], [[TMP15]]
+; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 24
+; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[TMP18]], 0
+; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 1
+; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], [[TMP20]]
+; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP16]]
+; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP21]]
+; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i32 0
+; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
+; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP28]]
+; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = mul i64 [[TMP30]], 16
+; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP31]]
+; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = mul i64 [[TMP33]], 24
+; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP34]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP26]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP29]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP32]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP35]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP16]]
+; CHECK-ORDERED-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP21]]
+; CHECK-ORDERED-NEXT:    [[TMP40:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i32 0
+; CHECK-ORDERED-NEXT:    [[TMP41:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP41]], 8
+; CHECK-ORDERED-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP42]]
+; CHECK-ORDERED-NEXT:    [[TMP44:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 16
+; CHECK-ORDERED-NEXT:    [[TMP46:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP45]]
+; CHECK-ORDERED-NEXT:    [[TMP47:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP48:%.*]] = mul i64 [[TMP47]], 24
+; CHECK-ORDERED-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP48]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP40]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP43]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP46]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP49]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP50:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
+; CHECK-ORDERED-NEXT:    [[TMP51:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; CHECK-ORDERED-NEXT:    [[TMP52:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; CHECK-ORDERED-NEXT:    [[TMP53:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
+; CHECK-ORDERED-NEXT:    [[TMP54:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP50]])
+; CHECK-ORDERED-NEXT:    [[TMP55:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP54]], <vscale x 8 x float> [[TMP51]])
+; CHECK-ORDERED-NEXT:    [[TMP56:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP55]], <vscale x 8 x float> [[TMP52]])
+; CHECK-ORDERED-NEXT:    [[TMP57]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP56]], <vscale x 8 x float> [[TMP53]])
+; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-ORDERED-NEXT:    [[TMP58:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[TMP58]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
@@ -1662,7 +1666,7 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP55]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP57]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -1676,7 +1680,7 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP55]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP57]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fmuladd_strict
@@ -1692,137 +1696,141 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP87:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP88:%.*]] = mul i64 [[TMP87]], 32
 ; CHECK-ORDERED-TF-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-ORDERED-TF-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 32
-; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = sub i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP26]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 16
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP28]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 24
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP30]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT]], i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT2]], i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 32
+; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = icmp ugt i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i64 [[TMP9]], i64 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 8
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP13]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 16
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 24
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP17]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY6:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY7:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 0)
 ; CHECK-ORDERED-TF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP86:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = mul i64 [[TMP32]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = add i64 [[TMP33]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = mul i64 [[TMP34]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = add i64 [[INDEX]], [[TMP35]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = mul i64 [[TMP37]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = add i64 [[TMP38]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = add i64 [[INDEX]], [[TMP40]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY7]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT22:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY6]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT21:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK10:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT19:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP73:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = add i64 [[TMP20]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = add i64 [[INDEX]], [[TMP22]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = mul i64 [[TMP24]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = add i64 [[TMP25]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = mul i64 [[TMP26]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = add i64 [[INDEX]], [[TMP27]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = add i64 [[TMP30]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX]], [[TMP32]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP23]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP28]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP33]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP40]]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = add i64 [[TMP43]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = add i64 [[INDEX]], [[TMP45]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP31]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP41]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP46]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = mul i64 [[TMP52]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP53]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = mul i64 [[TMP55]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP56]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = mul i64 [[TMP58]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP59]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP51]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP54]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP57]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP60]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP31]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP63:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP41]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP64:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP46]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP65:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP66:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP67:%.*]] = mul i64 [[TMP66]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP68:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i64 [[TMP67]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP69:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP70:%.*]] = mul i64 [[TMP69]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP71:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i64 [[TMP70]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP72:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP73:%.*]] = mul i64 [[TMP72]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP74:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i64 [[TMP73]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP65]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP68]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP71]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP74]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP75:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD12]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP76:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], [[WIDE_MASKED_LOAD13]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP77:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], [[WIDE_MASKED_LOAD14]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP78:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP79:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP75]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP80:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP79]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP81:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[TMP76]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP82:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP80]], <vscale x 8 x float> [[TMP81]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP83:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[TMP77]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP84:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP82]], <vscale x 8 x float> [[TMP83]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP85:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP78]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP86]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP84]], <vscale x 8 x float> [[TMP85]])
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP88]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP89:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP90:%.*]] = mul i64 [[TMP89]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP91:%.*]] = add i64 [[INDEX]], [[TMP90]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP92:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP93:%.*]] = mul i64 [[TMP92]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP94:%.*]] = add i64 [[INDEX]], [[TMP93]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP95:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP96:%.*]] = mul i64 [[TMP95]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP97:%.*]] = add i64 [[INDEX]], [[TMP96]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP91]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP94]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP97]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP98:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP99:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT16]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP100:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT17]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP101:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT18]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP102:%.*]] = extractelement <vscale x 8 x i1> [[TMP98]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP102]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP43]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = mul i64 [[TMP45]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP46]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP38]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP41]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP44]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP47]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP23]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP28]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP33]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = mul i64 [[TMP53]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 [[TMP54]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = mul i64 [[TMP56]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 [[TMP57]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = mul i64 [[TMP59]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 [[TMP60]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP52]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP55]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD16:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP58]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD17:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP61]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD14]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP63:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP64:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD12]], [[WIDE_MASKED_LOAD16]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP65:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD13]], [[WIDE_MASKED_LOAD17]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP66:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP62]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP67:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP66]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP68:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP63]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP69:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP67]], <vscale x 8 x float> [[TMP68]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP70:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> [[TMP64]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP71:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP69]], <vscale x 8 x float> [[TMP70]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP72:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> [[TMP65]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP73]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP71]], <vscale x 8 x float> [[TMP72]])
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP74:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP75:%.*]] = mul i64 [[TMP74]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP76:%.*]] = add i64 [[INDEX]], [[TMP75]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP77:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP78:%.*]] = mul i64 [[TMP77]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP79:%.*]] = add i64 [[INDEX]], [[TMP78]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP80:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP81:%.*]] = mul i64 [[TMP80]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP82:%.*]] = add i64 [[INDEX]], [[TMP81]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP79]], i64 [[TMP11]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP11]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT19]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT20]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT21]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT18]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT22]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT18]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[TMP83:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT22]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP84:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT21]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP85:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT20]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP86:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT19]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP87:%.*]] = extractelement <vscale x 8 x i1> [[TMP83]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP87]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED-TF:       scalar.ph:
 ; CHECK-ORDERED-TF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP86]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP73]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       for.body:
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP103:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP88:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP104:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP103]], float [[TMP104]], float [[SUM_07]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP89:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP88]], float [[TMP89]], float [[SUM_07]])
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP86]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP73]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 
@@ -1881,78 +1889,78 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 32
 ; CHECK-UNORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-UNORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-UNORDERED-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP53:%.*]] = mul i64 [[TMP52]], 32
+; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 32
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP50:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP51:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], [[TMP13]]
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = add i64 [[TMP16]], 0
-; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 1
-; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], [[TMP18]]
-; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP14]]
-; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP19]]
-; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP26]]
-; CHECK-UNORDERED-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP29:%.*]] = mul i64 [[TMP28]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP29]]
-; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP32]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP30]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP33]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
-; CHECK-UNORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP14]]
-; CHECK-UNORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP19]]
-; CHECK-UNORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP40]]
-; CHECK-UNORDERED-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP43]]
-; CHECK-UNORDERED-NEXT:    [[TMP45:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP46:%.*]] = mul i64 [[TMP45]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP46]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP38]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP41]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP44]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP47]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP48]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
-; CHECK-UNORDERED-NEXT:    [[TMP49]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
-; CHECK-UNORDERED-NEXT:    [[TMP50]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
-; CHECK-UNORDERED-NEXT:    [[TMP51]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
-; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP53]]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP50:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP51:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP52:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer), [[VECTOR_PH]] ], [ [[TMP53:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[TMP8]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]]
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[TMP13]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], [[TMP15]]
+; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[TMP18]], 0
+; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 1
+; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], [[TMP20]]
+; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP16]]
+; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP21]]
+; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i32 0
+; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP28]]
+; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = mul i64 [[TMP30]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP31]]
+; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP34:%.*]] = mul i64 [[TMP33]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP34]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP26]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP29]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP32]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP35]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]
+; CHECK-UNORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP16]]
+; CHECK-UNORDERED-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP21]]
+; CHECK-UNORDERED-NEXT:    [[TMP40:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i32 0
+; CHECK-UNORDERED-NEXT:    [[TMP41:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP41]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP42]]
+; CHECK-UNORDERED-NEXT:    [[TMP44:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP46:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP45]]
+; CHECK-UNORDERED-NEXT:    [[TMP47:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP48:%.*]] = mul i64 [[TMP47]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP48]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP40]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP43]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP46]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP49]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP50]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
+; CHECK-UNORDERED-NEXT:    [[TMP51]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
+; CHECK-UNORDERED-NEXT:    [[TMP52]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
+; CHECK-UNORDERED-NEXT:    [[TMP53]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
+; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-UNORDERED-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[TMP54]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd nnan <vscale x 8 x float> [[TMP49]], [[TMP48]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd nnan <vscale x 8 x float> [[TMP50]], [[BIN_RDX]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd nnan <vscale x 8 x float> [[TMP51]], [[BIN_RDX11]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd nnan <vscale x 8 x float> [[TMP51]], [[TMP50]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd nnan <vscale x 8 x float> [[TMP52]], [[BIN_RDX]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd nnan <vscale x 8 x float> [[TMP53]], [[BIN_RDX11]]
 ; CHECK-UNORDERED-NEXT:    [[TMP55:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX12]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
@@ -1987,73 +1995,73 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 32
 ; CHECK-ORDERED-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
 ; CHECK-ORDERED-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
-; CHECK-ORDERED-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP57:%.*]] = mul i64 [[TMP56]], 32
+; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 32
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP55:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 8
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = add i64 [[TMP6]], 0
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 1
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], 0
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 1
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], [[TMP13]]
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP15]], 24
-; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = add i64 [[TMP16]], 0
-; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 1
-; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX]], [[TMP18]]
-; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP14]]
-; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP19]]
-; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP26]]
-; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = mul i64 [[TMP28]], 16
-; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP29]]
-; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 24
-; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 [[TMP32]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP30]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP33]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
-; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP14]]
-; CHECK-ORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP19]]
-; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
-; CHECK-ORDERED-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP40]]
-; CHECK-ORDERED-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 16
-; CHECK-ORDERED-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP43]]
-; CHECK-ORDERED-NEXT:    [[TMP45:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP46:%.*]] = mul i64 [[TMP45]], 24
-; CHECK-ORDERED-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP46]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP38]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP41]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP44]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP47]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP48:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
-; CHECK-ORDERED-NEXT:    [[TMP49:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
-; CHECK-ORDERED-NEXT:    [[TMP50:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
-; CHECK-ORDERED-NEXT:    [[TMP51:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; CHECK-ORDERED-NEXT:    [[TMP52:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP48]])
-; CHECK-ORDERED-NEXT:    [[TMP53:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP52]], <vscale x 8 x float> [[TMP49]])
-; CHECK-ORDERED-NEXT:    [[TMP54:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP53]], <vscale x 8 x float> [[TMP50]])
-; CHECK-ORDERED-NEXT:    [[TMP55]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP54]], <vscale x 8 x float> [[TMP51]])
-; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP57]]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP57:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 8
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = add i64 [[TMP8]], 0
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 1
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]]
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 16
+; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = add i64 [[TMP13]], 0
+; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 1
+; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], [[TMP15]]
+; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul i64 [[TMP17]], 24
+; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = add i64 [[TMP18]], 0
+; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 1
+; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], [[TMP20]]
+; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP16]]
+; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP21]]
+; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i32 0
+; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
+; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP28]]
+; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = mul i64 [[TMP30]], 16
+; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP31]]
+; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = mul i64 [[TMP33]], 24
+; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i64 [[TMP34]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP26]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP29]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP32]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP35]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]
+; CHECK-ORDERED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP16]]
+; CHECK-ORDERED-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP21]]
+; CHECK-ORDERED-NEXT:    [[TMP40:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i32 0
+; CHECK-ORDERED-NEXT:    [[TMP41:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP41]], 8
+; CHECK-ORDERED-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP42]]
+; CHECK-ORDERED-NEXT:    [[TMP44:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 16
+; CHECK-ORDERED-NEXT:    [[TMP46:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP45]]
+; CHECK-ORDERED-NEXT:    [[TMP47:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP48:%.*]] = mul i64 [[TMP47]], 24
+; CHECK-ORDERED-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[TMP36]], i64 [[TMP48]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP40]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP43]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP46]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP49]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP50:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
+; CHECK-ORDERED-NEXT:    [[TMP51:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; CHECK-ORDERED-NEXT:    [[TMP52:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; CHECK-ORDERED-NEXT:    [[TMP53:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
+; CHECK-ORDERED-NEXT:    [[TMP54:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP50]])
+; CHECK-ORDERED-NEXT:    [[TMP55:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP54]], <vscale x 8 x float> [[TMP51]])
+; CHECK-ORDERED-NEXT:    [[TMP56:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP55]], <vscale x 8 x float> [[TMP52]])
+; CHECK-ORDERED-NEXT:    [[TMP57]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP56]], <vscale x 8 x float> [[TMP53]])
+; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-ORDERED-NEXT:    [[TMP58:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[TMP58]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
@@ -2061,7 +2069,7 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP55]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP57]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -2075,7 +2083,7 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP55]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP57]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fmuladd_strict_fmf
@@ -2091,137 +2099,141 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
 ; CHECK-ORDERED-TF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP87:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP88:%.*]] = mul i64 [[TMP87]], 32
 ; CHECK-ORDERED-TF-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-ORDERED-TF-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 32
-; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = sub i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[N]], [[TMP6]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 8
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP26]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 16
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP28]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 24
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP30]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT]], i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[N]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT2]], i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 32
+; CHECK-ORDERED-TF-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = icmp ugt i64 [[N]], [[TMP8]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i64 [[TMP9]], i64 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 8
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP13]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 16
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 24
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP17]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[N]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY6:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_ENTRY7:%.*]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 0)
 ; CHECK-ORDERED-TF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP86:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = mul i64 [[TMP32]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = add i64 [[TMP33]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = mul i64 [[TMP34]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = add i64 [[INDEX]], [[TMP35]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = mul i64 [[TMP37]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = add i64 [[TMP38]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = add i64 [[INDEX]], [[TMP40]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY7]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT22:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY6]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT21:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK10:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT19:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP73:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = add i64 [[INDEX]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = mul i64 [[TMP19]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = add i64 [[TMP20]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = mul i64 [[TMP21]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = add i64 [[INDEX]], [[TMP22]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = mul i64 [[TMP24]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = add i64 [[TMP25]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = mul i64 [[TMP26]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = add i64 [[INDEX]], [[TMP27]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = add i64 [[TMP30]], 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 1
+; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX]], [[TMP32]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP23]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP28]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP33]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP40]]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = add i64 [[TMP43]], 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 1
-; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = add i64 [[INDEX]], [[TMP45]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP31]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP41]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP46]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = mul i64 [[TMP52]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP53]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = mul i64 [[TMP55]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP56]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = mul i64 [[TMP58]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = getelementptr inbounds float, ptr [[TMP47]], i64 [[TMP59]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP51]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP54]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP57]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP60]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP31]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP63:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP41]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP64:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP46]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP65:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP66:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP67:%.*]] = mul i64 [[TMP66]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP68:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i64 [[TMP67]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP69:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP70:%.*]] = mul i64 [[TMP69]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP71:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i64 [[TMP70]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP72:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP73:%.*]] = mul i64 [[TMP72]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP74:%.*]] = getelementptr inbounds float, ptr [[TMP61]], i64 [[TMP73]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP65]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP68]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP71]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP74]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP75:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD12]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP76:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], [[WIDE_MASKED_LOAD13]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP77:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], [[WIDE_MASKED_LOAD14]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP78:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP79:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP75]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP80:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP79]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP81:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[TMP76]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP82:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP80]], <vscale x 8 x float> [[TMP81]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP83:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[TMP77]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP84:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP82]], <vscale x 8 x float> [[TMP83]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP85:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP78]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP86]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP84]], <vscale x 8 x float> [[TMP85]])
-; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP88]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP89:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP90:%.*]] = mul i64 [[TMP89]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP91:%.*]] = add i64 [[INDEX]], [[TMP90]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP92:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP93:%.*]] = mul i64 [[TMP92]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP94:%.*]] = add i64 [[INDEX]], [[TMP93]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP95:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP96:%.*]] = mul i64 [[TMP95]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP97:%.*]] = add i64 [[INDEX]], [[TMP96]]
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP91]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP94]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP97]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP98:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP99:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT16]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP100:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT17]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP101:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT18]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
-; CHECK-ORDERED-TF-NEXT:    [[TMP102:%.*]] = extractelement <vscale x 8 x i1> [[TMP98]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP102]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP43]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = mul i64 [[TMP45]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP34]], i64 [[TMP46]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP38]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP41]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP44]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP47]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP23]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP28]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP33]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 0
+; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = mul i64 [[TMP53]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 [[TMP54]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = mul i64 [[TMP56]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 [[TMP57]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = mul i64 [[TMP59]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i64 [[TMP60]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP52]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP55]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD16:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP58]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD17:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP61]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD14]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP63:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP64:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD12]], [[WIDE_MASKED_LOAD16]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP65:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD13]], [[WIDE_MASKED_LOAD17]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP66:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP62]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP67:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP66]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP68:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP63]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP69:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP67]], <vscale x 8 x float> [[TMP68]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP70:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 8 x float> [[TMP64]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP71:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP69]], <vscale x 8 x float> [[TMP70]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP72:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 8 x float> [[TMP65]], <vscale x 8 x float> shufflevector (<vscale x 8 x float> insertelement (<vscale x 8 x float> poison, float -0.000000e+00, i64 0), <vscale x 8 x float> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP73]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP71]], <vscale x 8 x float> [[TMP72]])
+; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP74:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP75:%.*]] = mul i64 [[TMP74]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP76:%.*]] = add i64 [[INDEX]], [[TMP75]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP77:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP78:%.*]] = mul i64 [[TMP77]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP79:%.*]] = add i64 [[INDEX]], [[TMP78]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP80:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP81:%.*]] = mul i64 [[TMP80]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP82:%.*]] = add i64 [[INDEX]], [[TMP81]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP79]], i64 [[TMP11]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP11]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT19]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT20]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT21]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT18]], i64 8)
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT22]] = call <vscale x 8 x i1> @llvm.vector.extract.nxv8i1.nxv16i1(<vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT18]], i64 0)
+; CHECK-ORDERED-TF-NEXT:    [[TMP83:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT22]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP84:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT21]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP85:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT20]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP86:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT19]], shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT:    [[TMP87:%.*]] = extractelement <vscale x 8 x i1> [[TMP83]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP87]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED-TF:       scalar.ph:
 ; CHECK-ORDERED-TF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP86]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP73]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED-TF:       for.body:
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP103:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP88:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP104:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP103]], float [[TMP104]], float [[SUM_07]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP89:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP88]], float [[TMP89]], float [[SUM_07]])
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP86]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP73]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll
index 2acc1ddbffea4..0c1b587c95259 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll
@@ -18,87 +18,91 @@ define void @simple_memset(i32 %val, ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[UMAX]], [[TMP4]]
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-NEXT:    [[TMP61:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP62:%.*]] = mul i64 [[TMP61]], 16
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 16
-; CHECK-NEXT:    [[TMP7:%.*]] = sub i64 [[UMAX]], [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[UMAX]], [[TMP6]]
+; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
+; CHECK-NEXT:    [[TMP7:%.*]] = sub i64 [[UMAX]], [[TMP11]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[UMAX]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
-; CHECK-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 4
-; CHECK-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP26]]
 ; CHECK-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
-; CHECK-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP28]]
+; CHECK-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 4
+; CHECK-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP28]]
 ; CHECK-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 12
-; CHECK-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP30]]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[UMAX]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_PART_NEXT]], i64 [[UMAX]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[UMAX]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_PART_NEXT2]], i64 [[UMAX]])
+; CHECK-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 8
+; CHECK-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP30]]
+; CHECK-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 12
+; CHECK-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP32]]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[UMAX]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[UMAX]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY6:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY7:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 0)
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT10:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT11:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT12:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT13:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX6]], 0
-; CHECK-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP33:%.*]] = mul i64 [[TMP32]], 4
-; CHECK-NEXT:    [[TMP34:%.*]] = add i64 [[TMP33]], 0
-; CHECK-NEXT:    [[TMP35:%.*]] = mul i64 [[TMP34]], 1
-; CHECK-NEXT:    [[TMP36:%.*]] = add i64 [[INDEX6]], [[TMP35]]
-; CHECK-NEXT:    [[TMP37:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP38:%.*]] = mul i64 [[TMP37]], 8
-; CHECK-NEXT:    [[TMP39:%.*]] = add i64 [[TMP38]], 0
-; CHECK-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 1
-; CHECK-NEXT:    [[TMP41:%.*]] = add i64 [[INDEX6]], [[TMP40]]
-; CHECK-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 12
-; CHECK-NEXT:    [[TMP44:%.*]] = add i64 [[TMP43]], 0
-; CHECK-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 1
-; CHECK-NEXT:    [[TMP46:%.*]] = add i64 [[INDEX6]], [[TMP45]]
-; CHECK-NEXT:    [[TMP47:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP31]]
-; CHECK-NEXT:    [[TMP48:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP36]]
-; CHECK-NEXT:    [[TMP49:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP41]]
-; CHECK-NEXT:    [[TMP50:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP46]]
-; CHECK-NEXT:    [[TMP51:%.*]] = getelementptr i32, ptr [[TMP47]], i32 0
-; CHECK-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP53:%.*]] = mul i64 [[TMP52]], 4
-; CHECK-NEXT:    [[TMP54:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP53]]
-; CHECK-NEXT:    [[TMP55:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP56:%.*]] = mul i64 [[TMP55]], 8
-; CHECK-NEXT:    [[TMP57:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP56]]
-; CHECK-NEXT:    [[TMP58:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP59:%.*]] = mul i64 [[TMP58]], 12
-; CHECK-NEXT:    [[TMP60:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP59]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP51]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP54]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK7]])
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP57]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK8]])
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP60]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]])
-; CHECK-NEXT:    [[INDEX_NEXT10]] = add i64 [[INDEX6]], [[TMP62]]
+; CHECK-NEXT:    [[INDEX8:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT12:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY7]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY6]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK10:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK11:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT14:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX8]], 0
+; CHECK-NEXT:    [[TMP34:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP35:%.*]] = mul i64 [[TMP34]], 4
+; CHECK-NEXT:    [[TMP36:%.*]] = add i64 [[TMP35]], 0
+; CHECK-NEXT:    [[TMP37:%.*]] = mul i64 [[TMP36]], 1
+; CHECK-NEXT:    [[TMP38:%.*]] = add i64 [[INDEX8]], [[TMP37]]
+; CHECK-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
+; CHECK-NEXT:    [[TMP41:%.*]] = add i64 [[TMP40]], 0
+; CHECK-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP41]], 1
+; CHECK-NEXT:    [[TMP43:%.*]] = add i64 [[INDEX8]], [[TMP42]]
+; CHECK-NEXT:    [[TMP44:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 12
+; CHECK-NEXT:    [[TMP46:%.*]] = add i64 [[TMP45]], 0
+; CHECK-NEXT:    [[TMP47:%.*]] = mul i64 [[TMP46]], 1
+; CHECK-NEXT:    [[TMP48:%.*]] = add i64 [[INDEX8]], [[TMP47]]
+; CHECK-NEXT:    [[TMP49:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP33]]
+; CHECK-NEXT:    [[TMP50:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP38]]
+; CHECK-NEXT:    [[TMP51:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP43]]
+; CHECK-NEXT:    [[TMP52:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP48]]
+; CHECK-NEXT:    [[TMP53:%.*]] = getelementptr i32, ptr [[TMP49]], i32 0
+; CHECK-NEXT:    [[TMP54:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP55:%.*]] = mul i64 [[TMP54]], 4
+; CHECK-NEXT:    [[TMP56:%.*]] = getelementptr i32, ptr [[TMP49]], i64 [[TMP55]]
+; CHECK-NEXT:    [[TMP57:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP58:%.*]] = mul i64 [[TMP57]], 8
+; CHECK-NEXT:    [[TMP59:%.*]] = getelementptr i32, ptr [[TMP49]], i64 [[TMP58]]
+; CHECK-NEXT:    [[TMP60:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP61:%.*]] = mul i64 [[TMP60]], 12
+; CHECK-NEXT:    [[TMP62:%.*]] = getelementptr i32, ptr [[TMP49]], i64 [[TMP61]]
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP53]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP56]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP59]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK10]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP62]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK11]])
+; CHECK-NEXT:    [[INDEX_NEXT12]] = add i64 [[INDEX8]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP63:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP64:%.*]] = mul i64 [[TMP63]], 4
-; CHECK-NEXT:    [[TMP65:%.*]] = add i64 [[INDEX6]], [[TMP64]]
+; CHECK-NEXT:    [[TMP65:%.*]] = add i64 [[INDEX8]], [[TMP64]]
 ; CHECK-NEXT:    [[TMP66:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP67:%.*]] = mul i64 [[TMP66]], 8
-; CHECK-NEXT:    [[TMP68:%.*]] = add i64 [[INDEX6]], [[TMP67]]
+; CHECK-NEXT:    [[TMP68:%.*]] = add i64 [[INDEX8]], [[TMP67]]
 ; CHECK-NEXT:    [[TMP69:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP70:%.*]] = mul i64 [[TMP69]], 12
-; CHECK-NEXT:    [[TMP71:%.*]] = add i64 [[INDEX6]], [[TMP70]]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX6]], i64 [[TMP9]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT11]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP65]], i64 [[TMP9]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT12]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP68]], i64 [[TMP9]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT13]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP71]], i64 [[TMP9]])
-; CHECK-NEXT:    [[TMP72:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-NEXT:    [[TMP73:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT11]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-NEXT:    [[TMP74:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT12]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-NEXT:    [[TMP75:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT13]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP71:%.*]] = add i64 [[INDEX8]], [[TMP70]]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP68]], i64 [[TMP9]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT13:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX8]], i64 [[TMP9]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT14]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT15]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT13]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT13]], i64 0)
+; CHECK-NEXT:    [[TMP72:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT17]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP73:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT16]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP74:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT15]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP75:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT14]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-NEXT:    [[TMP76:%.*]] = extractelement <vscale x 4 x i1> [[TMP72]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP76]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
@@ -145,113 +149,117 @@ define void @cond_memset(i32 %val, ptr noalias readonly %cond_ptr, ptr noalias %
 ; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[UMAX]], [[TMP4]]
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP1]]
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-NEXT:    [[TMP83:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP84:%.*]] = mul i64 [[TMP83]], 16
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 16
-; CHECK-NEXT:    [[TMP7:%.*]] = sub i64 [[UMAX]], [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[UMAX]], [[TMP6]]
+; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 16
+; CHECK-NEXT:    [[TMP7:%.*]] = sub i64 [[UMAX]], [[TMP11]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp ugt i64 [[UMAX]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = select i1 [[TMP8]], i64 [[TMP7]], i64 0
-; CHECK-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP26:%.*]] = mul i64 [[TMP25]], 4
-; CHECK-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP26]]
 ; CHECK-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 8
-; CHECK-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP28]]
+; CHECK-NEXT:    [[TMP28:%.*]] = mul i64 [[TMP27]], 4
+; CHECK-NEXT:    [[INDEX_PART_NEXT:%.*]] = add i64 0, [[TMP28]]
 ; CHECK-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 12
-; CHECK-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP30]]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[UMAX]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_PART_NEXT]], i64 [[UMAX]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[UMAX]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_PART_NEXT2]], i64 [[UMAX]])
+; CHECK-NEXT:    [[TMP30:%.*]] = mul i64 [[TMP29]], 8
+; CHECK-NEXT:    [[INDEX_PART_NEXT1:%.*]] = add i64 0, [[TMP30]]
+; CHECK-NEXT:    [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP32:%.*]] = mul i64 [[TMP31]], 12
+; CHECK-NEXT:    [[INDEX_PART_NEXT2:%.*]] = add i64 0, [[TMP32]]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_PART_NEXT1]], i64 [[UMAX]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[UMAX]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY6:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_ENTRY7:%.*]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 0)
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[VAL:%.*]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT13:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT14:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT15:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX6]], 0
-; CHECK-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP33:%.*]] = mul i64 [[TMP32]], 4
-; CHECK-NEXT:    [[TMP34:%.*]] = add i64 [[TMP33]], 0
-; CHECK-NEXT:    [[TMP35:%.*]] = mul i64 [[TMP34]], 1
-; CHECK-NEXT:    [[TMP36:%.*]] = add i64 [[INDEX6]], [[TMP35]]
-; CHECK-NEXT:    [[TMP37:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP38:%.*]] = mul i64 [[TMP37]], 8
-; CHECK-NEXT:    [[TMP39:%.*]] = add i64 [[TMP38]], 0
-; CHECK-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 1
-; CHECK-NEXT:    [[TMP41:%.*]] = add i64 [[INDEX6]], [[TMP40]]
-; CHECK-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP43:%.*]] = mul i64 [[TMP42]], 12
-; CHECK-NEXT:    [[TMP44:%.*]] = add i64 [[TMP43]], 0
-; CHECK-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 1
-; CHECK-NEXT:    [[TMP46:%.*]] = add i64 [[INDEX6]], [[TMP45]]
-; CHECK-NEXT:    [[TMP47:%.*]] = getelementptr i32, ptr [[COND_PTR:%.*]], i64 [[TMP31]]
-; CHECK-NEXT:    [[TMP48:%.*]] = getelementptr i32, ptr [[COND_PTR]], i64 [[TMP36]]
-; CHECK-NEXT:    [[TMP49:%.*]] = getelementptr i32, ptr [[COND_PTR]], i64 [[TMP41]]
-; CHECK-NEXT:    [[TMP50:%.*]] = getelementptr i32, ptr [[COND_PTR]], i64 [[TMP46]]
-; CHECK-NEXT:    [[TMP51:%.*]] = getelementptr i32, ptr [[TMP47]], i32 0
-; CHECK-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP53:%.*]] = mul i64 [[TMP52]], 4
-; CHECK-NEXT:    [[TMP54:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP53]]
-; CHECK-NEXT:    [[TMP55:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP56:%.*]] = mul i64 [[TMP55]], 8
-; CHECK-NEXT:    [[TMP57:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP56]]
-; CHECK-NEXT:    [[TMP58:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP59:%.*]] = mul i64 [[TMP58]], 12
-; CHECK-NEXT:    [[TMP60:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP59]]
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP51]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP54]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 4 x i32> poison)
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP57]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 4 x i32> poison)
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP60]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 4 x i32> poison)
-; CHECK-NEXT:    [[TMP61:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP62:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD10]], zeroinitializer
-; CHECK-NEXT:    [[TMP63:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD11]], zeroinitializer
+; CHECK-NEXT:    [[INDEX8:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY7]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY6]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT19:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK10:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK11:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX8]], 0
+; CHECK-NEXT:    [[TMP34:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP35:%.*]] = mul i64 [[TMP34]], 4
+; CHECK-NEXT:    [[TMP36:%.*]] = add i64 [[TMP35]], 0
+; CHECK-NEXT:    [[TMP37:%.*]] = mul i64 [[TMP36]], 1
+; CHECK-NEXT:    [[TMP38:%.*]] = add i64 [[INDEX8]], [[TMP37]]
+; CHECK-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP40:%.*]] = mul i64 [[TMP39]], 8
+; CHECK-NEXT:    [[TMP41:%.*]] = add i64 [[TMP40]], 0
+; CHECK-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP41]], 1
+; CHECK-NEXT:    [[TMP43:%.*]] = add i64 [[INDEX8]], [[TMP42]]
+; CHECK-NEXT:    [[TMP44:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP44]], 12
+; CHECK-NEXT:    [[TMP46:%.*]] = add i64 [[TMP45]], 0
+; CHECK-NEXT:    [[TMP47:%.*]] = mul i64 [[TMP46]], 1
+; CHECK-NEXT:    [[TMP48:%.*]] = add i64 [[INDEX8]], [[TMP47]]
+; CHECK-NEXT:    [[TMP49:%.*]] = getelementptr i32, ptr [[COND_PTR:%.*]], i64 [[TMP33]]
+; CHECK-NEXT:    [[TMP50:%.*]] = getelementptr i32, ptr [[COND_PTR]], i64 [[TMP38]]
+; CHECK-NEXT:    [[TMP51:%.*]] = getelementptr i32, ptr [[COND_PTR]], i64 [[TMP43]]
+; CHECK-NEXT:    [[TMP52:%.*]] = getelementptr i32, ptr [[COND_PTR]], i64 [[TMP48]]
+; CHECK-NEXT:    [[TMP53:%.*]] = getelementptr i32, ptr [[TMP49]], i32 0
+; CHECK-NEXT:    [[TMP54:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP55:%.*]] = mul i64 [[TMP54]], 4
+; CHECK-NEXT:    [[TMP56:%.*]] = getelementptr i32, ptr [[TMP49]], i64 [[TMP55]]
+; CHECK-NEXT:    [[TMP57:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP58:%.*]] = mul i64 [[TMP57]], 8
+; CHECK-NEXT:    [[TMP59:%.*]] = getelementptr i32, ptr [[TMP49]], i64 [[TMP58]]
+; CHECK-NEXT:    [[TMP60:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP61:%.*]] = mul i64 [[TMP60]], 12
+; CHECK-NEXT:    [[TMP62:%.*]] = getelementptr i32, ptr [[TMP49]], i64 [[TMP61]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP53]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP56]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP59]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP62]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK11]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[TMP63:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP64:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD12]], zeroinitializer
-; CHECK-NEXT:    [[TMP69:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP61]], <vscale x 4 x i1> zeroinitializer
-; CHECK-NEXT:    [[TMP70:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 4 x i1> [[TMP62]], <vscale x 4 x i1> zeroinitializer
-; CHECK-NEXT:    [[TMP71:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 4 x i1> [[TMP63]], <vscale x 4 x i1> zeroinitializer
-; CHECK-NEXT:    [[TMP72:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 4 x i1> [[TMP64]], <vscale x 4 x i1> zeroinitializer
-; CHECK-NEXT:    [[TMP65:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP31]]
-; CHECK-NEXT:    [[TMP66:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP36]]
-; CHECK-NEXT:    [[TMP67:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP41]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP46]]
-; CHECK-NEXT:    [[TMP73:%.*]] = getelementptr i32, ptr [[TMP65]], i32 0
-; CHECK-NEXT:    [[TMP74:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP75:%.*]] = mul i64 [[TMP74]], 4
-; CHECK-NEXT:    [[TMP76:%.*]] = getelementptr i32, ptr [[TMP65]], i64 [[TMP75]]
-; CHECK-NEXT:    [[TMP77:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP78:%.*]] = mul i64 [[TMP77]], 8
-; CHECK-NEXT:    [[TMP79:%.*]] = getelementptr i32, ptr [[TMP65]], i64 [[TMP78]]
-; CHECK-NEXT:    [[TMP80:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP81:%.*]] = mul i64 [[TMP80]], 12
-; CHECK-NEXT:    [[TMP82:%.*]] = getelementptr i32, ptr [[TMP65]], i64 [[TMP81]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP73]], i32 4, <vscale x 4 x i1> [[TMP69]])
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP76]], i32 4, <vscale x 4 x i1> [[TMP70]])
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP79]], i32 4, <vscale x 4 x i1> [[TMP71]])
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP82]], i32 4, <vscale x 4 x i1> [[TMP72]])
-; CHECK-NEXT:    [[INDEX_NEXT13]] = add i64 [[INDEX6]], [[TMP84]]
+; CHECK-NEXT:    [[TMP65:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD13]], zeroinitializer
+; CHECK-NEXT:    [[TMP66:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD14]], zeroinitializer
+; CHECK-NEXT:    [[TMP67:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP63]], <vscale x 4 x i1> zeroinitializer
+; CHECK-NEXT:    [[TMP68:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 4 x i1> [[TMP64]], <vscale x 4 x i1> zeroinitializer
+; CHECK-NEXT:    [[TMP69:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 4 x i1> [[TMP65]], <vscale x 4 x i1> zeroinitializer
+; CHECK-NEXT:    [[TMP70:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK11]], <vscale x 4 x i1> [[TMP66]], <vscale x 4 x i1> zeroinitializer
+; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[TMP33]]
+; CHECK-NEXT:    [[TMP72:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP38]]
+; CHECK-NEXT:    [[TMP73:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP43]]
+; CHECK-NEXT:    [[TMP74:%.*]] = getelementptr i32, ptr [[PTR]], i64 [[TMP48]]
+; CHECK-NEXT:    [[TMP75:%.*]] = getelementptr i32, ptr [[TMP71]], i32 0
+; CHECK-NEXT:    [[TMP76:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP77:%.*]] = mul i64 [[TMP76]], 4
+; CHECK-NEXT:    [[TMP78:%.*]] = getelementptr i32, ptr [[TMP71]], i64 [[TMP77]]
+; CHECK-NEXT:    [[TMP79:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP80:%.*]] = mul i64 [[TMP79]], 8
+; CHECK-NEXT:    [[TMP81:%.*]] = getelementptr i32, ptr [[TMP71]], i64 [[TMP80]]
+; CHECK-NEXT:    [[TMP82:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP83:%.*]] = mul i64 [[TMP82]], 12
+; CHECK-NEXT:    [[TMP84:%.*]] = getelementptr i32, ptr [[TMP71]], i64 [[TMP83]]
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP75]], i32 4, <vscale x 4 x i1> [[TMP67]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP78]], i32 4, <vscale x 4 x i1> [[TMP68]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP81]], i32 4, <vscale x 4 x i1> [[TMP69]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP84]], i32 4, <vscale x 4 x i1> [[TMP70]])
+; CHECK-NEXT:    [[INDEX_NEXT15]] = add i64 [[INDEX8]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP85:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP86:%.*]] = mul i64 [[TMP85]], 4
-; CHECK-NEXT:    [[TMP87:%.*]] = add i64 [[INDEX6]], [[TMP86]]
+; CHECK-NEXT:    [[TMP87:%.*]] = add i64 [[INDEX8]], [[TMP86]]
 ; CHECK-NEXT:    [[TMP88:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP89:%.*]] = mul i64 [[TMP88]], 8
-; CHECK-NEXT:    [[TMP90:%.*]] = add i64 [[INDEX6]], [[TMP89]]
+; CHECK-NEXT:    [[TMP90:%.*]] = add i64 [[INDEX8]], [[TMP89]]
 ; CHECK-NEXT:    [[TMP91:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP92:%.*]] = mul i64 [[TMP91]], 12
-; CHECK-NEXT:    [[TMP93:%.*]] = add i64 [[INDEX6]], [[TMP92]]
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX6]], i64 [[TMP9]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT14]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP87]], i64 [[TMP9]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT15]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP90]], i64 [[TMP9]])
-; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP93]], i64 [[TMP9]])
-; CHECK-NEXT:    [[TMP94:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-NEXT:    [[TMP95:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT14]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-NEXT:    [[TMP96:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT15]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
-; CHECK-NEXT:    [[TMP97:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT16]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP93:%.*]] = add i64 [[INDEX8]], [[TMP92]]
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP90]], i64 [[TMP9]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT16:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX8]], i64 [[TMP9]])
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT19]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT16]], i64 4)
+; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT20]] = call <vscale x 4 x i1> @llvm.vector.extract.nxv4i1.nxv8i1(<vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT16]], i64 0)
+; CHECK-NEXT:    [[TMP94:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT20]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP95:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT19]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP96:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT18]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP97:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT17]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
 ; CHECK-NEXT:    [[TMP98:%.*]] = extractelement <vscale x 4 x i1> [[TMP94]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP98]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-wide-lane-mask.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-wide-lane-mask.ll
new file mode 100644
index 0000000000000..f766cfeefb0d8
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-wide-lane-mask.ll
@@ -0,0 +1,656 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S --passes="default<O3>" -sve-tail-folding-insn-threshold=0 -force-vector-interleave=0                < %s | FileCheck %s -check-prefix CHECK-SVE-UF0
+; RUN: opt -S --passes="default<O3>" -sve-tail-folding-insn-threshold=0 -force-vector-interleave=2                < %s | FileCheck %s -check-prefix CHECK-SVE-UF2
+; RUN: opt -S --passes="default<O3>" -sve-tail-folding-insn-threshold=0 -force-vector-interleave=4                < %s | FileCheck %s -check-prefix CHECK-SVE-UF4
+; RUN: opt -S --passes="default<O3>" -sve-tail-folding-insn-threshold=0 -force-vector-interleave=0 -mattr=+sve2p1 < %s | FileCheck %s -check-prefix CHECK-SVE2p1-UF0
+; RUN: opt -S --passes="default<O3>" -sve-tail-folding-insn-threshold=0 -force-vector-interleave=2 -mattr=+sve2p1 < %s | FileCheck %s -check-prefix CHECK-SVE2p1-UF2
+; RUN: opt -S --passes="default<O3>" -sve-tail-folding-insn-threshold=0 -force-vector-interleave=4 -mattr=+sve2p1 < %s | FileCheck %s -check-prefix CHECK-SVE2p1-UF4
+
+target triple = "aarch64-unknown-linux"
+
+define void @f0(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-UF0-LABEL: define void @f0(
+; CHECK-SVE-UF0-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-SVE-UF0-NEXT:  entry:
+; CHECK-SVE-UF0-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE-UF0-NEXT:    br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE-UF0:       for.body.preheader:
+; CHECK-SVE-UF0-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE-UF0-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF0-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
+; CHECK-SVE-UF0-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF0-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE-UF0:       vector.body:
+; CHECK-SVE-UF0-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF0-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF0-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE-UF0-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP2]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-SVE-UF0-NEXT:    [[TMP3:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE-UF0-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE-UF0-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP3]], ptr [[TMP4]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE-UF0-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE-UF0-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF0-NEXT:    [[TMP5:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
+; CHECK-SVE-UF0-NEXT:    br i1 [[TMP5]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-SVE-UF0:       for.cond.cleanup:
+; CHECK-SVE-UF0-NEXT:    ret void
+;
+; CHECK-SVE-UF2-LABEL: define void @f0(
+; CHECK-SVE-UF2-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-SVE-UF2-NEXT:  entry:
+; CHECK-SVE-UF2-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE-UF2-NEXT:    br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE-UF2:       for.body.preheader:
+; CHECK-SVE-UF2-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE-UF2-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 5
+; CHECK-SVE-UF2-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 4
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP3]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF2-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 4
+; CHECK-SVE-UF2-NEXT:    [[TMP6:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP7:%.*]] = shl nuw nsw i64 [[TMP6]], 4
+; CHECK-SVE-UF2-NEXT:    [[TMP8:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP9:%.*]] = shl nuw nsw i64 [[TMP8]], 4
+; CHECK-SVE-UF2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE-UF2:       vector.body:
+; CHECK-SVE-UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE-UF2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i64 [[TMP5]]
+; CHECK-SVE-UF2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP10]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-SVE-UF2-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP11]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 16 x i8> poison)
+; CHECK-SVE-UF2-NEXT:    [[TMP12:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE-UF2-NEXT:    [[TMP13:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD3]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE-UF2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE-UF2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i64 [[TMP7]]
+; CHECK-SVE-UF2-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP12]], ptr [[TMP14]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE-UF2-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP13]], ptr [[TMP15]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK2]])
+; CHECK-SVE-UF2-NEXT:    [[INDEX_NEXT4]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE-UF2-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX_NEXT4]], [[TMP9]]
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP16]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_NEXT4]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF2-NEXT:    [[TMP17:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT5]], i64 0
+; CHECK-SVE-UF2-NEXT:    br i1 [[TMP17]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-SVE-UF2:       for.cond.cleanup:
+; CHECK-SVE-UF2-NEXT:    ret void
+;
+; CHECK-SVE-UF4-LABEL: define void @f0(
+; CHECK-SVE-UF4-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-SVE-UF4-NEXT:  entry:
+; CHECK-SVE-UF4-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE-UF4-NEXT:    br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE-UF4:       for.body.preheader:
+; CHECK-SVE-UF4-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE-UF4-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 6
+; CHECK-SVE-UF4-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 4
+; CHECK-SVE-UF4-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 5
+; CHECK-SVE-UF4-NEXT:    [[TMP6:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP7:%.*]] = mul nuw nsw i64 [[TMP6]], 48
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP7]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP5]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP3]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[TMP8:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP9:%.*]] = shl nuw nsw i64 [[TMP8]], 4
+; CHECK-SVE-UF4-NEXT:    [[TMP10:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP11:%.*]] = shl nuw nsw i64 [[TMP10]], 5
+; CHECK-SVE-UF4-NEXT:    [[TMP12:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP13:%.*]] = mul nuw nsw i64 [[TMP12]], 48
+; CHECK-SVE-UF4-NEXT:    [[TMP14:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP15:%.*]] = shl nuw nsw i64 [[TMP14]], 4
+; CHECK-SVE-UF4-NEXT:    [[TMP16:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP17:%.*]] = shl nuw nsw i64 [[TMP16]], 5
+; CHECK-SVE-UF4-NEXT:    [[TMP18:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP19:%.*]] = mul nuw nsw i64 [[TMP18]], 48
+; CHECK-SVE-UF4-NEXT:    [[TMP20:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP21:%.*]] = shl nuw nsw i64 [[TMP20]], 4
+; CHECK-SVE-UF4-NEXT:    [[TMP22:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP23:%.*]] = shl nuw nsw i64 [[TMP22]], 5
+; CHECK-SVE-UF4-NEXT:    [[TMP24:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP25:%.*]] = mul nuw nsw i64 [[TMP24]], 48
+; CHECK-SVE-UF4-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE-UF4:       vector.body:
+; CHECK-SVE-UF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT14:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE-UF4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i64 [[TMP9]]
+; CHECK-SVE-UF4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i64 [[TMP11]]
+; CHECK-SVE-UF4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i64 [[TMP13]]
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP26]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP27]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 16 x i8> poison)
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP28]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 16 x i8> poison)
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP29]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 16 x i8> poison)
+; CHECK-SVE-UF4-NEXT:    [[TMP30:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP31:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD9]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP32:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD10]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP33:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD11]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE-UF4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i8, ptr [[TMP34]], i64 [[TMP15]]
+; CHECK-SVE-UF4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i8, ptr [[TMP34]], i64 [[TMP17]]
+; CHECK-SVE-UF4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds i8, ptr [[TMP34]], i64 [[TMP19]]
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP30]], ptr [[TMP34]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP31]], ptr [[TMP35]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK6]])
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP32]], ptr [[TMP36]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK7]])
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP33]], ptr [[TMP37]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK8]])
+; CHECK-SVE-UF4-NEXT:    [[INDEX_NEXT14]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE-UF4-NEXT:    [[TMP38:%.*]] = add i64 [[INDEX_NEXT14]], [[TMP21]]
+; CHECK-SVE-UF4-NEXT:    [[TMP39:%.*]] = add i64 [[INDEX_NEXT14]], [[TMP23]]
+; CHECK-SVE-UF4-NEXT:    [[TMP40:%.*]] = add i64 [[INDEX_NEXT14]], [[TMP25]]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP40]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT15]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP39]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP38]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_NEXT14]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[TMP41:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT17]], i64 0
+; CHECK-SVE-UF4-NEXT:    br i1 [[TMP41]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-SVE-UF4:       for.cond.cleanup:
+; CHECK-SVE-UF4-NEXT:    ret void
+;
+; CHECK-SVE2p1-UF0-LABEL: define void @f0(
+; CHECK-SVE2p1-UF0-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-SVE2p1-UF0-NEXT:  entry:
+; CHECK-SVE2p1-UF0-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE2p1-UF0-NEXT:    br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE2p1-UF0:       for.body.preheader:
+; CHECK-SVE2p1-UF0-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 4
+; CHECK-SVE2p1-UF0-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF0-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE2p1-UF0:       vector.body:
+; CHECK-SVE2p1-UF0-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF0-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF0-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP2]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP3:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF0-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP3]], ptr [[TMP4]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE2p1-UF0-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE2p1-UF0-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP5:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
+; CHECK-SVE2p1-UF0-NEXT:    br i1 [[TMP5]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-SVE2p1-UF0:       for.cond.cleanup:
+; CHECK-SVE2p1-UF0-NEXT:    ret void
+;
+; CHECK-SVE2p1-UF2-LABEL: define void @f0(
+; CHECK-SVE2p1-UF2-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-SVE2p1-UF2-NEXT:  entry:
+; CHECK-SVE2p1-UF2-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE2p1-UF2-NEXT:    br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE2p1-UF2:       for.body.preheader:
+; CHECK-SVE2p1-UF2-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 5
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 4
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP3]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 4
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP6:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP7:%.*]] = shl nuw nsw i64 [[TMP6]], 4
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP8:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP9:%.*]] = shl nuw nsw i64 [[TMP8]], 4
+; CHECK-SVE2p1-UF2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE2p1-UF2:       vector.body:
+; CHECK-SVE2p1-UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT4:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i64 [[TMP5]]
+; CHECK-SVE2p1-UF2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP10]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-SVE2p1-UF2-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP11]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 16 x i8> poison)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP12:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP13:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD3]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i64 [[TMP7]]
+; CHECK-SVE2p1-UF2-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP12]], ptr [[TMP14]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE2p1-UF2-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP13]], ptr [[TMP15]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK2]])
+; CHECK-SVE2p1-UF2-NEXT:    [[INDEX_NEXT4]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX_NEXT4]], [[TMP9]]
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP16]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_NEXT4]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP17:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT5]], i64 0
+; CHECK-SVE2p1-UF2-NEXT:    br i1 [[TMP17]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-SVE2p1-UF2:       for.cond.cleanup:
+; CHECK-SVE2p1-UF2-NEXT:    ret void
+;
+; CHECK-SVE2p1-UF4-LABEL: define void @f0(
+; CHECK-SVE2p1-UF4-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+; CHECK-SVE2p1-UF4-NEXT:  entry:
+; CHECK-SVE2p1-UF4-NEXT:    [[CMP7:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE2p1-UF4-NEXT:    br i1 [[CMP7]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE2p1-UF4:       for.body.preheader:
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 6
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 4
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 5
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP6:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP7:%.*]] = mul nuw nsw i64 [[TMP6]], 48
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP7]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP5]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP3]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP8:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP9:%.*]] = shl nuw nsw i64 [[TMP8]], 4
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP10:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP11:%.*]] = shl nuw nsw i64 [[TMP10]], 5
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP12:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP13:%.*]] = mul nuw nsw i64 [[TMP12]], 48
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP14:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP15:%.*]] = shl nuw nsw i64 [[TMP14]], 4
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP16:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP17:%.*]] = shl nuw nsw i64 [[TMP16]], 5
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP18:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP19:%.*]] = mul nuw nsw i64 [[TMP18]], 48
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP20:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP21:%.*]] = shl nuw nsw i64 [[TMP20]], 4
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP22:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP23:%.*]] = shl nuw nsw i64 [[TMP22]], 5
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP24:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP25:%.*]] = mul nuw nsw i64 [[TMP24]], 48
+; CHECK-SVE2p1-UF4-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE2p1-UF4:       vector.body:
+; CHECK-SVE2p1-UF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT14:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT15:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i64 [[TMP9]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i64 [[TMP11]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i64 [[TMP13]]
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP26]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP27]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 16 x i8> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP28]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 16 x i8> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP29]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 16 x i8> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP30:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP31:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD9]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP32:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD10]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP33:%.*]] = mul <vscale x 16 x i8> [[WIDE_MASKED_LOAD11]], shufflevector (<vscale x 16 x i8> insertelement (<vscale x 16 x i8> poison, i8 3, i64 0), <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i8, ptr [[TMP34]], i64 [[TMP15]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i8, ptr [[TMP34]], i64 [[TMP17]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds i8, ptr [[TMP34]], i64 [[TMP19]]
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP30]], ptr [[TMP34]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP31]], ptr [[TMP35]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK6]])
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP32]], ptr [[TMP36]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK7]])
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP33]], ptr [[TMP37]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK8]])
+; CHECK-SVE2p1-UF4-NEXT:    [[INDEX_NEXT14]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP38:%.*]] = add i64 [[INDEX_NEXT14]], [[TMP21]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP39:%.*]] = add i64 [[INDEX_NEXT14]], [[TMP23]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP40:%.*]] = add i64 [[INDEX_NEXT14]], [[TMP25]]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP40]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT15]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP39]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[TMP38]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = tail call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_NEXT14]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP41:%.*]] = extractelement <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT17]], i64 0
+; CHECK-SVE2p1-UF4-NEXT:    br i1 [[TMP41]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-SVE2p1-UF4:       for.cond.cleanup:
+; CHECK-SVE2p1-UF4-NEXT:    ret void
+;
+entry:
+  %cmp7 = icmp sgt i32 %n, 0
+  br i1 %cmp7, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  br label %for.body
+
+for.cond.cleanup:
+  ret void
+
+for.body:
+  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds i8, ptr %src, i64 %indvars.iv
+  %0 = load i8, ptr %arrayidx, align 1
+  %mul = mul i8 %0, 3
+  %arrayidx3 = getelementptr inbounds i8, ptr %dst, i64 %indvars.iv
+  store i8 %mul, ptr %arrayidx3, align 1
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+}
+
+define void @f1(ptr noalias %dst, ptr readonly %src, i32 %n) #0 {
+; CHECK-SVE-UF0-LABEL: define void @f1(
+; CHECK-SVE-UF0-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-SVE-UF0-NEXT:  entry:
+; CHECK-SVE-UF0-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE-UF0-NEXT:    br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE-UF0:       for.body.preheader:
+; CHECK-SVE-UF0-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE-UF0-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF0-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 1
+; CHECK-SVE-UF0-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF0-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE-UF0:       vector.body:
+; CHECK-SVE-UF0-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF0-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF0-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE-UF0-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP2]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; CHECK-SVE-UF0-NEXT:    [[TMP3:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE-UF0-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE-UF0-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP3]], ptr [[TMP4]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE-UF0-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE-UF0-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF0-NEXT:    [[TMP5:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
+; CHECK-SVE-UF0-NEXT:    br i1 [[TMP5]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-SVE-UF0:       for.cond.cleanup:
+; CHECK-SVE-UF0-NEXT:    ret void
+;
+; CHECK-SVE-UF2-LABEL: define void @f1(
+; CHECK-SVE-UF2-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-SVE-UF2-NEXT:  entry:
+; CHECK-SVE-UF2-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE-UF2-NEXT:    br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE-UF2:       for.body.preheader:
+; CHECK-SVE-UF2-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE-UF2-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 2
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-SVE-UF2-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 1
+; CHECK-SVE-UF2-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF2-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 1
+; CHECK-SVE-UF2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE-UF2:       vector.body:
+; CHECK-SVE-UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE-UF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP6]], i64 [[TMP3]]
+; CHECK-SVE-UF2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP6]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; CHECK-SVE-UF2-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP7]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x double> poison)
+; CHECK-SVE-UF2-NEXT:    [[TMP8:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE-UF2-NEXT:    [[TMP9:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD4]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE-UF2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE-UF2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, ptr [[TMP10]], i64 [[TMP5]]
+; CHECK-SVE-UF2-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP8]], ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE-UF2-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP9]], ptr [[TMP11]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; CHECK-SVE-UF2-NEXT:    [[INDEX_NEXT5]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT5]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; CHECK-SVE-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT7]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-SVE-UF2-NEXT:    [[TMP12:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT7]], i64 0
+; CHECK-SVE-UF2-NEXT:    br i1 [[TMP12]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-SVE-UF2:       for.cond.cleanup:
+; CHECK-SVE-UF2-NEXT:    ret void
+;
+; CHECK-SVE-UF4-LABEL: define void @f1(
+; CHECK-SVE-UF4-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-SVE-UF4-NEXT:  entry:
+; CHECK-SVE-UF4-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE-UF4-NEXT:    br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE-UF4:       for.body.preheader:
+; CHECK-SVE-UF4-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE-UF4-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 3
+; CHECK-SVE-UF4-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 2
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP3]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY6:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 2)
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY7:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 0)
+; CHECK-SVE-UF4-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 1
+; CHECK-SVE-UF4-NEXT:    [[TMP6:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP7:%.*]] = shl nuw nsw i64 [[TMP6]], 2
+; CHECK-SVE-UF4-NEXT:    [[TMP8:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP9:%.*]] = mul nuw nsw i64 [[TMP8]], 6
+; CHECK-SVE-UF4-NEXT:    [[TMP10:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP11:%.*]] = shl nuw nsw i64 [[TMP10]], 1
+; CHECK-SVE-UF4-NEXT:    [[TMP12:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP13:%.*]] = shl nuw nsw i64 [[TMP12]], 2
+; CHECK-SVE-UF4-NEXT:    [[TMP14:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP15:%.*]] = mul nuw nsw i64 [[TMP14]], 6
+; CHECK-SVE-UF4-NEXT:    [[TMP16:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE-UF4-NEXT:    [[TMP17:%.*]] = shl nuw nsw i64 [[TMP16]], 2
+; CHECK-SVE-UF4-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE-UF4:       vector.body:
+; CHECK-SVE-UF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY7]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT21:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY6]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT19:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK10:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE-UF4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE-UF4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, ptr [[TMP18]], i64 [[TMP5]]
+; CHECK-SVE-UF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds double, ptr [[TMP18]], i64 [[TMP7]]
+; CHECK-SVE-UF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds double, ptr [[TMP18]], i64 [[TMP9]]
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP19]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 2 x double> poison)
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP20]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 2 x double> poison)
+; CHECK-SVE-UF4-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP21]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 2 x double> poison)
+; CHECK-SVE-UF4-NEXT:    [[TMP22:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP23:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD11]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP24:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD12]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP25:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD13]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE-UF4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE-UF4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds double, ptr [[TMP26]], i64 [[TMP11]]
+; CHECK-SVE-UF4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds double, ptr [[TMP26]], i64 [[TMP13]]
+; CHECK-SVE-UF4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds double, ptr [[TMP26]], i64 [[TMP15]]
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP22]], ptr [[TMP26]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP23]], ptr [[TMP27]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK8]])
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP24]], ptr [[TMP28]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK9]])
+; CHECK-SVE-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP25]], ptr [[TMP29]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK10]])
+; CHECK-SVE-UF4-NEXT:    [[INDEX_NEXT16]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE-UF4-NEXT:    [[TMP30:%.*]] = add i64 [[INDEX_NEXT16]], [[TMP17]]
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP30]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT17:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT16]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT19]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT20]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT17]], i64 2)
+; CHECK-SVE-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT21]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT17]], i64 0)
+; CHECK-SVE-UF4-NEXT:    [[TMP31:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT21]], i64 0
+; CHECK-SVE-UF4-NEXT:    br i1 [[TMP31]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-SVE-UF4:       for.cond.cleanup:
+; CHECK-SVE-UF4-NEXT:    ret void
+;
+; CHECK-SVE2p1-UF0-LABEL: define void @f1(
+; CHECK-SVE2p1-UF0-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-SVE2p1-UF0-NEXT:  entry:
+; CHECK-SVE2p1-UF0-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE2p1-UF0-NEXT:    br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE2p1-UF0:       for.body.preheader:
+; CHECK-SVE2p1-UF0-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 1
+; CHECK-SVE2p1-UF0-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF0-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE2p1-UF0:       vector.body:
+; CHECK-SVE2p1-UF0-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF0-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF0-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP2]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP3:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF0-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP3]], ptr [[TMP4]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE2p1-UF0-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE2p1-UF0-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = tail call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF0-NEXT:    [[TMP5:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
+; CHECK-SVE2p1-UF0-NEXT:    br i1 [[TMP5]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-SVE2p1-UF0:       for.cond.cleanup:
+; CHECK-SVE2p1-UF0-NEXT:    ret void
+;
+; CHECK-SVE2p1-UF2-LABEL: define void @f1(
+; CHECK-SVE2p1-UF2-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-SVE2p1-UF2-NEXT:  entry:
+; CHECK-SVE2p1-UF2-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE2p1-UF2-NEXT:    br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE2p1-UF2:       for.body.preheader:
+; CHECK-SVE2p1-UF2-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 2
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 1
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 1
+; CHECK-SVE2p1-UF2-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE2p1-UF2:       vector.body:
+; CHECK-SVE2p1-UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT7:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP6]], i64 [[TMP3]]
+; CHECK-SVE2p1-UF2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP6]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; CHECK-SVE2p1-UF2-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP7]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x double> poison)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP8:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP9:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD4]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, ptr [[TMP10]], i64 [[TMP5]]
+; CHECK-SVE2p1-UF2-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP8]], ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE2p1-UF2-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP9]], ptr [[TMP11]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; CHECK-SVE2p1-UF2-NEXT:    [[INDEX_NEXT5]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT5]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; CHECK-SVE2p1-UF2-NEXT:    [[ACTIVE_LANE_MASK_NEXT7]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-SVE2p1-UF2-NEXT:    [[TMP12:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT7]], i64 0
+; CHECK-SVE2p1-UF2-NEXT:    br i1 [[TMP12]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-SVE2p1-UF2:       for.cond.cleanup:
+; CHECK-SVE2p1-UF2-NEXT:    ret void
+;
+; CHECK-SVE2p1-UF4-LABEL: define void @f1(
+; CHECK-SVE2p1-UF4-SAME: ptr noalias nocapture writeonly [[DST:%.*]], ptr nocapture readonly [[SRC:%.*]], i32 [[N:%.*]]) local_unnamed_addr #[[ATTR0]] {
+; CHECK-SVE2p1-UF4-NEXT:  entry:
+; CHECK-SVE2p1-UF4-NEXT:    [[CMP6:%.*]] = icmp sgt i32 [[N]], 0
+; CHECK-SVE2p1-UF4-NEXT:    br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
+; CHECK-SVE2p1-UF4:       for.body.preheader:
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext nneg i32 [[N]] to i64
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP0:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP1:%.*]] = shl nuw nsw i64 [[TMP0]], 3
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP2:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 2
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP3]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY3:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY4:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY5:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY6:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 2)
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_ENTRY7:%.*]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY3]], i64 0)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP4:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 1
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP6:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP7:%.*]] = shl nuw nsw i64 [[TMP6]], 2
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP8:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP9:%.*]] = mul nuw nsw i64 [[TMP8]], 6
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP10:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP11:%.*]] = shl nuw nsw i64 [[TMP10]], 1
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP12:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP13:%.*]] = shl nuw nsw i64 [[TMP12]], 2
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP14:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP15:%.*]] = mul nuw nsw i64 [[TMP14]], 6
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP16:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP17:%.*]] = shl nuw nsw i64 [[TMP16]], 2
+; CHECK-SVE2p1-UF4-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-SVE2p1-UF4:       vector.body:
+; CHECK-SVE2p1-UF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDEX_NEXT16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY7]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT21:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY6]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT19:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK10:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[FOR_BODY_PREHEADER]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, ptr [[SRC]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, ptr [[TMP18]], i64 [[TMP5]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds double, ptr [[TMP18]], i64 [[TMP7]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds double, ptr [[TMP18]], i64 [[TMP9]]
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP19]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 2 x double> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP20]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 2 x double> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = tail call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr nonnull [[TMP21]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK10]], <vscale x 2 x double> poison)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP22:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP23:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD11]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP24:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD12]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP25:%.*]] = fmul <vscale x 2 x double> [[WIDE_MASKED_LOAD13]], shufflevector (<vscale x 2 x double> insertelement (<vscale x 2 x double> poison, double 3.000000e+00, i64 0), <vscale x 2 x double> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds double, ptr [[TMP26]], i64 [[TMP11]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds double, ptr [[TMP26]], i64 [[TMP13]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds double, ptr [[TMP26]], i64 [[TMP15]]
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP22]], ptr [[TMP26]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP23]], ptr [[TMP27]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK8]])
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP24]], ptr [[TMP28]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK9]])
+; CHECK-SVE2p1-UF4-NEXT:    tail call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP25]], ptr [[TMP29]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK10]])
+; CHECK-SVE2p1-UF4-NEXT:    [[INDEX_NEXT16]] = add i64 [[INDEX]], [[TMP1]]
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP30:%.*]] = add i64 [[INDEX_NEXT16]], [[TMP17]]
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[TMP30]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT17:%.*]] = tail call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT16]], i64 [[WIDE_TRIP_COUNT]])
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT19]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT20]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT17]], i64 2)
+; CHECK-SVE2p1-UF4-NEXT:    [[ACTIVE_LANE_MASK_NEXT21]] = tail call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT17]], i64 0)
+; CHECK-SVE2p1-UF4-NEXT:    [[TMP31:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT21]], i64 0
+; CHECK-SVE2p1-UF4-NEXT:    br i1 [[TMP31]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-SVE2p1-UF4:       for.cond.cleanup:
+; CHECK-SVE2p1-UF4-NEXT:    ret void
+;
+entry:
+  %cmp6 = icmp sgt i32 %n, 0
+  br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup
+
+for.body.preheader:
+  %wide.trip.count = zext nneg i32 %n to i64
+  br label %for.body
+
+for.cond.cleanup:
+  ret void
+
+for.body:
+  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds double, ptr %src, i64 %indvars.iv
+  %0 = load double, ptr %arrayidx, align 8
+  %mul = fmul double %0, 3.000000e+00
+  %arrayidx2 = getelementptr inbounds double, ptr %dst, i64 %indvars.iv
+  store double %mul, ptr %arrayidx2, align 8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+}
+
+attributes #0 = {nounwind vscale_range(1,16) "target-cpu"="neoverse-v1" }
+;.
+; CHECK-SVE-UF0: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-SVE-UF0: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-SVE-UF0: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-SVE-UF0: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]}
+;.
+; CHECK-SVE-UF2: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-SVE-UF2: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-SVE-UF2: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-SVE-UF2: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]}
+;.
+; CHECK-SVE-UF4: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-SVE-UF4: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-SVE-UF4: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-SVE-UF4: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]}
+;.
+; CHECK-SVE2p1-UF0: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-SVE2p1-UF0: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-SVE2p1-UF0: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-SVE2p1-UF0: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]}
+;.
+; CHECK-SVE2p1-UF2: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-SVE2p1-UF2: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-SVE2p1-UF2: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-SVE2p1-UF2: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]}
+;.
+; CHECK-SVE2p1-UF4: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK-SVE2p1-UF4: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK-SVE2p1-UF4: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK-SVE2p1-UF4: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/uniform-args-call-variants.ll b/llvm/test/Transforms/LoopVectorize/AArch64/uniform-args-call-variants.ll
index d8f14f30295b6..8c1747c26bdf9 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/uniform-args-call-variants.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/uniform-args-call-variants.ll
@@ -40,36 +40,33 @@ define void @test_uniform(ptr noalias %dst, ptr readonly %src, i64 %uniform , i6
 ; INTERLEAVE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = shl i64 [[TMP2]], 2
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[N]], i64 [[TMP3]])
-; INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP9:%.*]] = shl i64 [[TMP8]], 1
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 [[N]])
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP9]], i64 [[N]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[N]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
 ; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; INTERLEAVE:       vector.body:
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT4:%.*]], [[VECTOR_BODY]] ]
-; INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP12:%.*]] = shl i64 [[TMP11]], 1
-; INTERLEAVE-NEXT:    [[TMP13:%.*]] = getelementptr double, ptr [[TMP10]], i64 [[TMP12]]
-; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
-; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP13]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x double> poison)
-; INTERLEAVE-NEXT:    [[TMP14:%.*]] = call <vscale x 2 x double> @foo_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD]], i64 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; INTERLEAVE-NEXT:    [[TMP15:%.*]] = call <vscale x 2 x double> @foo_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD3]], i64 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; INTERLEAVE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP18:%.*]] = shl i64 [[TMP17]], 1
-; INTERLEAVE-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, ptr [[TMP16]], i64 [[TMP18]]
-; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP14]], ptr [[TMP16]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP15]], ptr [[TMP19]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = shl i64 [[TMP6]], 1
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = getelementptr double, ptr [[TMP5]], i64 [[TMP7]]
+; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP5]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP8]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x double> poison)
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = call <vscale x 2 x double> @foo_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD]], i64 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = call <vscale x 2 x double> @foo_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD4]], i64 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = shl i64 [[TMP12]], 1
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds double, ptr [[TMP11]], i64 [[TMP13]]
+; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP9]], ptr [[TMP11]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP10]], ptr [[TMP14]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]
-; INTERLEAVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP21:%.*]] = shl i64 [[TMP20]], 1
-; INTERLEAVE-NEXT:    [[TMP22:%.*]] = add i64 [[INDEX]], [[TMP21]]
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX]], i64 [[TMP4]])
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT4]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP22]], i64 [[TMP4]])
-; INTERLEAVE-NEXT:    [[TMP23:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP4]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT6]], i64 0
 ; INTERLEAVE-NEXT:    br i1 [[TMP23]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP:%.*]], !llvm.loop [[LOOP0:![0-9]+]]
 ; INTERLEAVE:       for.cond.cleanup:
 ; INTERLEAVE-NEXT:    ret void
@@ -126,36 +123,33 @@ define void @test_uniform_smaller_scalar(ptr noalias %dst, ptr readonly %src, i3
 ; INTERLEAVE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = shl i64 [[TMP2]], 2
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[N]], i64 [[TMP3]])
-; INTERLEAVE-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP9:%.*]] = shl i64 [[TMP8]], 1
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 0, i64 [[N]])
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP9]], i64 [[N]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 0, i64 [[N]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 2)
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY2:%.*]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_ENTRY]], i64 0)
 ; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; INTERLEAVE:       vector.body:
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT4:%.*]], [[VECTOR_BODY]] ]
-; INTERLEAVE-NEXT:    [[TMP10:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP12:%.*]] = shl i64 [[TMP11]], 1
-; INTERLEAVE-NEXT:    [[TMP13:%.*]] = getelementptr double, ptr [[TMP10]], i64 [[TMP12]]
-; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP10]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
-; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP13]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x double> poison)
-; INTERLEAVE-NEXT:    [[TMP14:%.*]] = call <vscale x 2 x double> @bar_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD]], i32 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; INTERLEAVE-NEXT:    [[TMP15:%.*]] = call <vscale x 2 x double> @bar_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD3]], i32 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
-; INTERLEAVE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP18:%.*]] = shl i64 [[TMP17]], 1
-; INTERLEAVE-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, ptr [[TMP16]], i64 [[TMP18]]
-; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP14]], ptr [[TMP16]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
-; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP15]], ptr [[TMP19]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY2]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT6:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK3:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY1]], [[ENTRY]] ], [ [[ACTIVE_LANE_MASK_NEXT5:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-NEXT:    [[TMP7:%.*]] = shl i64 [[TMP6]], 1
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = getelementptr double, ptr [[TMP5]], i64 [[TMP7]]
+; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP5]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x double> poison)
+; INTERLEAVE-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP8]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]], <vscale x 2 x double> poison)
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = call <vscale x 2 x double> @bar_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD]], i32 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; INTERLEAVE-NEXT:    [[TMP10:%.*]] = call <vscale x 2 x double> @bar_uniform(<vscale x 2 x double> [[WIDE_MASKED_LOAD4]], i32 [[UNIFORM]], <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
+; INTERLEAVE-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-NEXT:    [[TMP13:%.*]] = shl i64 [[TMP12]], 1
+; INTERLEAVE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds double, ptr [[TMP11]], i64 [[TMP13]]
+; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP9]], ptr [[TMP11]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; INTERLEAVE-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP10]], ptr [[TMP14]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK3]])
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]
-; INTERLEAVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
-; INTERLEAVE-NEXT:    [[TMP21:%.*]] = shl i64 [[TMP20]], 1
-; INTERLEAVE-NEXT:    [[TMP22:%.*]] = add i64 [[INDEX]], [[TMP21]]
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX]], i64 [[TMP4]])
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT4]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[TMP22]], i64 [[TMP4]])
-; INTERLEAVE-NEXT:    [[TMP23:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP4]])
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 2)
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT6]] = call <vscale x 2 x i1> @llvm.vector.extract.nxv2i1.nxv4i1(<vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], i64 0)
+; INTERLEAVE-NEXT:    [[TMP23:%.*]] = extractelement <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT6]], i64 0
 ; INTERLEAVE-NEXT:    br i1 [[TMP23]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP:%.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; INTERLEAVE:       for.cond.cleanup:
 ; INTERLEAVE-NEXT:    ret void
@@ -201,8 +195,8 @@ define void @test_uniform_not_invariant(ptr noalias %dst, ptr readonly %src, i64
 ; INTERLEAVE-SAME: (ptr noalias [[DST:%.*]], ptr readonly [[SRC:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; INTERLEAVE-NEXT:  entry:
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.usub.sat.i64(i64 [[N]], i64 2)
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = icmp ne i64 [[N]], 0
 ; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY1:%.*]] = icmp ugt i64 [[N]], 1
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_ENTRY:%.*]] = icmp ne i64 [[N]], 0
 ; INTERLEAVE-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; INTERLEAVE:       vector.body:
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE4:%.*]] ]
@@ -229,8 +223,8 @@ define void @test_uniform_not_invariant(ptr noalias %dst, ptr readonly %src, i64
 ; INTERLEAVE:       pred.store.continue4:
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 2
 ; INTERLEAVE-NEXT:    [[TMP11:%.*]] = or disjoint i64 [[INDEX]], 1
-; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = icmp ult i64 [[INDEX]], [[TMP0]]
 ; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT5]] = icmp ult i64 [[TMP11]], [[TMP0]]
+; INTERLEAVE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = icmp ult i64 [[INDEX]], [[TMP0]]
 ; INTERLEAVE-NEXT:    br i1 [[ACTIVE_LANE_MASK_NEXT]], label [[VECTOR_BODY]], label [[FOR_COND_CLEANUP:%.*]], !llvm.loop [[LOOP4:![0-9]+]]
 ; INTERLEAVE:       for.cond.cleanup:
 ; INTERLEAVE-NEXT:    ret void
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-prefer-flag.ll b/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-prefer-flag.ll
index 8f3cfc5b20ee9..c9645646ad0ac 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-prefer-flag.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-prefer-flag.ll
@@ -56,10 +56,10 @@ define dso_local void @interleave4(ptr noalias nocapture %A, ptr noalias nocaptu
 ; PREDFLAG:  %[[ADD2:.*]] = add i32 %index, 4
 ; PREDFLAG:  %[[ADD3:.*]] = add i32 %index, 8
 ; PREDFLAG:  %[[ADD4:.*]] = add i32 %index, 12
-; PREDFLAG:  %[[ALM1:active.lane.mask.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %[[ADD1]], i32 %N)
-; PREDFLAG:  %[[ALM2:active.lane.mask.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %[[ADD2]], i32 %N)
-; PREDFLAG:  %[[ALM3:active.lane.mask.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %[[ADD3]], i32 %N)
 ; PREDFLAG:  %[[ALM4:active.lane.mask.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %[[ADD4]], i32 %N)
+; PREDFLAG:  %[[ALM3:active.lane.mask.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %[[ADD3]], i32 %N)
+; PREDFLAG:  %[[ALM2:active.lane.mask.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %[[ADD2]], i32 %N)
+; PREDFLAG:  %[[ALM1:active.lane.mask.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %[[ADD1]], i32 %N)
 ;
 ; PREDFLAG:  call <4 x i32> @llvm.masked.load.v4i32.p0({{.*}}, <4 x i1> %[[ALM1]],{{.*}}
 ; PREDFLAG:  call <4 x i32> @llvm.masked.load.v4i32.p0({{.*}}, <4 x i1> %[[ALM2]],{{.*}}
diff --git a/llvm/test/Transforms/LoopVectorize/strict-fadd-interleave-only.ll b/llvm/test/Transforms/LoopVectorize/strict-fadd-interleave-only.ll
index 97c84f251fef1..8770a19ce9cfa 100644
--- a/llvm/test/Transforms/LoopVectorize/strict-fadd-interleave-only.ll
+++ b/llvm/test/Transforms/LoopVectorize/strict-fadd-interleave-only.ll
@@ -49,8 +49,8 @@ define float @pr70988() {
 ; CHECK-ALM-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ALM-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX1]], 0
 ; CHECK-ALM-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX1]], 1
-; CHECK-ALM-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = icmp ult i32 [[TMP0]], 1021
 ; CHECK-ALM-NEXT:    [[ACTIVE_LANE_MASK2:%.*]] = icmp ult i32 [[TMP1]], 1021
+; CHECK-ALM-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = icmp ult i32 [[TMP0]], 1021
 ; CHECK-ALM-NEXT:    [[TMP2:%.*]] = select contract i1 [[ACTIVE_LANE_MASK]], float 1.000000e+00, float -0.000000e+00
 ; CHECK-ALM-NEXT:    [[TMP3:%.*]] = fadd contract float [[VEC_PHI]], [[TMP2]]
 ; CHECK-ALM-NEXT:    [[TMP4:%.*]] = select contract i1 [[ACTIVE_LANE_MASK2]], float 1.000000e+00, float -0.000000e+00
@@ -158,8 +158,8 @@ define float @pr72720reduction_using_active_lane_mask(ptr %src) {
 ; CHECK-ALM-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[PRED_LOAD_CONTINUE3]] ]
 ; CHECK-ALM-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
 ; CHECK-ALM-NEXT:    [[TMP1:%.*]] = add i32 [[INDEX]], 1
-; CHECK-ALM-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = icmp ult i32 [[TMP0]], 15
 ; CHECK-ALM-NEXT:    [[ACTIVE_LANE_MASK1:%.*]] = icmp ult i32 [[TMP1]], 15
+; CHECK-ALM-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = icmp ult i32 [[TMP0]], 15
 ; CHECK-ALM-NEXT:    br i1 [[ACTIVE_LANE_MASK]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
 ; CHECK-ALM:       pred.load.if:
 ; CHECK-ALM-NEXT:    [[TMP2:%.*]] = getelementptr float, ptr [[SRC]], i32 [[TMP0]]

>From bd6adbbb91e7cf42163e9e92942cae805cfe62b8 Mon Sep 17 00:00:00 2001
From: Momchil Velikov <momchil.velikov at arm.com>
Date: Fri, 5 Apr 2024 10:45:15 +0100
Subject: [PATCH 2/2] [AArch64] Refactor redundant PTEST optimisations (NFC)

---
 llvm/lib/Target/AArch64/AArch64InstrInfo.cpp | 170 ++++++++++---------
 llvm/lib/Target/AArch64/AArch64InstrInfo.h   |   3 +
 2 files changed, 95 insertions(+), 78 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index aa0b7c93f8661..54f13e5ce12dd 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -1354,48 +1354,52 @@ static bool areCFlagsAccessedBetweenInstrs(
   return false;
 }
 
-/// optimizePTestInstr - Attempt to remove a ptest of a predicate-generating
-/// operation which could set the flags in an identical manner
-bool AArch64InstrInfo::optimizePTestInstr(
-    MachineInstr *PTest, unsigned MaskReg, unsigned PredReg,
-    const MachineRegisterInfo *MRI) const {
-  auto *Mask = MRI->getUniqueVRegDef(MaskReg);
-  auto *Pred = MRI->getUniqueVRegDef(PredReg);
-  auto NewOp = Pred->getOpcode();
-  bool OpChanged = false;
-
+std::pair<bool, unsigned>
+AArch64InstrInfo::canRemovePTestInstr(MachineInstr *PTest, MachineInstr *Mask,
+                                      MachineInstr *Pred,
+                                      const MachineRegisterInfo *MRI) const {
   unsigned MaskOpcode = Mask->getOpcode();
   unsigned PredOpcode = Pred->getOpcode();
   bool PredIsPTestLike = isPTestLikeOpcode(PredOpcode);
   bool PredIsWhileLike = isWhileOpcode(PredOpcode);
 
-  if (isPTrueOpcode(MaskOpcode) && (PredIsPTestLike || PredIsWhileLike) &&
-      getElementSizeForOpcode(MaskOpcode) ==
-          getElementSizeForOpcode(PredOpcode) &&
-      Mask->getOperand(1).getImm() == 31) {
+  if (PredIsWhileLike) {
+    // For PTEST(PG, PG), PTEST is redundant when PG is the result of a WHILEcc
+    // instruction and the condition is "any" since WHILcc does an implicit
+    // PTEST(ALL, PG) check and PG is always a subset of ALL.
+    if ((Mask == Pred) && PTest->getOpcode() == AArch64::PTEST_PP_ANY)
+      return {true, 0};
+
     // For PTEST(PTRUE_ALL, WHILE), if the element size matches, the PTEST is
     // redundant since WHILE performs an implicit PTEST with an all active
-    // mask. Must be an all active predicate of matching element size.
+    // mask.
+    if (isPTrueOpcode(MaskOpcode) && Mask->getOperand(1).getImm() == 31 &&
+        getElementSizeForOpcode(MaskOpcode) ==
+            getElementSizeForOpcode(PredOpcode))
+      return {true, 0};
+
+    return {false, 0};
+  }
+
+  if (PredIsPTestLike) {
+    // For PTEST(PG, PG), PTEST is redundant when PG is the result of an
+    // instruction that sets the flags as PTEST would and the condition is
+    // "any" since PG is always a subset of the governing predicate of the
+    // ptest-like instruction.
+    if ((Mask == Pred) && PTest->getOpcode() == AArch64::PTEST_PP_ANY)
+      return {true, 0};
 
     // For PTEST(PTRUE_ALL, PTEST_LIKE), the PTEST is redundant if the
-    // PTEST_LIKE instruction uses the same all active mask and the element
-    // size matches. If the PTEST has a condition of any then it is always
-    // redundant.
-    if (PredIsPTestLike) {
+    // the element size matches and either the PTEST_LIKE instruction uses
+    // the same all active mask or the condition is "any".
+    if (isPTrueOpcode(MaskOpcode) && Mask->getOperand(1).getImm() == 31 &&
+        getElementSizeForOpcode(MaskOpcode) ==
+            getElementSizeForOpcode(PredOpcode)) {
       auto PTestLikeMask = MRI->getUniqueVRegDef(Pred->getOperand(1).getReg());
-      if (Mask != PTestLikeMask && PTest->getOpcode() != AArch64::PTEST_PP_ANY)
-        return false;
+      if (Mask == PTestLikeMask || PTest->getOpcode() == AArch64::PTEST_PP_ANY)
+        return {true, 0};
     }
 
-    // Fallthough to simply remove the PTEST.
-  } else if ((Mask == Pred) && (PredIsPTestLike || PredIsWhileLike) &&
-             PTest->getOpcode() == AArch64::PTEST_PP_ANY) {
-    // For PTEST(PG, PG), PTEST is redundant when PG is the result of an
-    // instruction that sets the flags as PTEST would. This is only valid when
-    // the condition is any.
-
-    // Fallthough to simply remove the PTEST.
-  } else if (PredIsPTestLike) {
     // For PTEST(PG, PTEST_LIKE(PG, ...)), the PTEST is redundant since the
     // flags are set based on the same mask 'PG', but PTEST_LIKE must operate
     // on 8-bit predicates like the PTEST.  Otherwise, for instructions like
@@ -1420,55 +1424,65 @@ bool AArch64InstrInfo::optimizePTestInstr(
     // identical regardless of element size.
     auto PTestLikeMask = MRI->getUniqueVRegDef(Pred->getOperand(1).getReg());
     uint64_t PredElementSize = getElementSizeForOpcode(PredOpcode);
-    if ((Mask != PTestLikeMask) ||
-        (PredElementSize != AArch64::ElementSizeB &&
-         PTest->getOpcode() != AArch64::PTEST_PP_ANY))
-      return false;
+    if (Mask == PTestLikeMask && (PredElementSize == AArch64::ElementSizeB ||
+                                  PTest->getOpcode() == AArch64::PTEST_PP_ANY))
+      return {true, 0};
 
-    // Fallthough to simply remove the PTEST.
-  } else {
-    // If OP in PTEST(PG, OP(PG, ...)) has a flag-setting variant change the
-    // opcode so the PTEST becomes redundant.
-    switch (PredOpcode) {
-    case AArch64::AND_PPzPP:
-    case AArch64::BIC_PPzPP:
-    case AArch64::EOR_PPzPP:
-    case AArch64::NAND_PPzPP:
-    case AArch64::NOR_PPzPP:
-    case AArch64::ORN_PPzPP:
-    case AArch64::ORR_PPzPP:
-    case AArch64::BRKA_PPzP:
-    case AArch64::BRKPA_PPzPP:
-    case AArch64::BRKB_PPzP:
-    case AArch64::BRKPB_PPzPP:
-    case AArch64::RDFFR_PPz: {
-      // Check to see if our mask is the same. If not the resulting flag bits
-      // may be different and we can't remove the ptest.
-      auto *PredMask = MRI->getUniqueVRegDef(Pred->getOperand(1).getReg());
-      if (Mask != PredMask)
-        return false;
-      break;
-    }
-    case AArch64::BRKN_PPzP: {
-      // BRKN uses an all active implicit mask to set flags unlike the other
-      // flag-setting instructions.
-      // PTEST(PTRUE_B(31), BRKN(PG, A, B)) -> BRKNS(PG, A, B).
-      if ((MaskOpcode != AArch64::PTRUE_B) ||
-          (Mask->getOperand(1).getImm() != 31))
-        return false;
-      break;
-    }
-    case AArch64::PTRUE_B:
-      // PTEST(OP=PTRUE_B(A), OP) -> PTRUES_B(A)
-      break;
-    default:
-      // Bail out if we don't recognize the input
-      return false;
-    }
+    return {false, 0};
+  }
 
-    NewOp = convertToFlagSettingOpc(PredOpcode);
-    OpChanged = true;
+  // If OP in PTEST(PG, OP(PG, ...)) has a flag-setting variant change the
+  // opcode so the PTEST becomes redundant.
+  switch (PredOpcode) {
+  case AArch64::AND_PPzPP:
+  case AArch64::BIC_PPzPP:
+  case AArch64::EOR_PPzPP:
+  case AArch64::NAND_PPzPP:
+  case AArch64::NOR_PPzPP:
+  case AArch64::ORN_PPzPP:
+  case AArch64::ORR_PPzPP:
+  case AArch64::BRKA_PPzP:
+  case AArch64::BRKPA_PPzPP:
+  case AArch64::BRKB_PPzP:
+  case AArch64::BRKPB_PPzPP:
+  case AArch64::RDFFR_PPz: {
+    // Check to see if our mask is the same. If not the resulting flag bits
+    // may be different and we can't remove the ptest.
+    auto *PredMask = MRI->getUniqueVRegDef(Pred->getOperand(1).getReg());
+    if (Mask != PredMask)
+      return {false, 0};
+    break;
   }
+  case AArch64::BRKN_PPzP: {
+    // BRKN uses an all active implicit mask to set flags unlike the other
+    // flag-setting instructions.
+    // PTEST(PTRUE_B(31), BRKN(PG, A, B)) -> BRKNS(PG, A, B).
+    if ((MaskOpcode != AArch64::PTRUE_B) ||
+        (Mask->getOperand(1).getImm() != 31))
+      return {false, 0};
+    break;
+  }
+  case AArch64::PTRUE_B:
+    // PTEST(OP=PTRUE_B(A), OP) -> PTRUES_B(A)
+    break;
+  default:
+    // Bail out if we don't recognize the input
+    return {false, 0};
+  }
+
+  return {true, convertToFlagSettingOpc(PredOpcode)};
+}
+
+/// optimizePTestInstr - Attempt to remove a ptest of a predicate-generating
+/// operation which could set the flags in an identical manner
+bool AArch64InstrInfo::optimizePTestInstr(
+    MachineInstr *PTest, unsigned MaskReg, unsigned PredReg,
+    const MachineRegisterInfo *MRI) const {
+  auto *Mask = MRI->getUniqueVRegDef(MaskReg);
+  auto *Pred = MRI->getUniqueVRegDef(PredReg);
+  auto [canRemove, NewOp] = canRemovePTestInstr(PTest, Mask, Pred, MRI);
+  if (!canRemove)
+    return false;
 
   const TargetRegisterInfo *TRI = &getRegisterInfo();
 
@@ -1481,9 +1495,9 @@ bool AArch64InstrInfo::optimizePTestInstr(
   // as they are prior to PTEST. Sometimes this requires the tested PTEST
   // operand to be replaced with an equivalent instruction that also sets the
   // flags.
-  Pred->setDesc(get(NewOp));
   PTest->eraseFromParent();
-  if (OpChanged) {
+  if (NewOp) {
+    Pred->setDesc(get(NewOp));
     bool succeeded = UpdateOperandRegClass(*Pred);
     (void)succeeded;
     assert(succeeded && "Operands have incompatible register classes!");
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.h b/llvm/lib/Target/AArch64/AArch64InstrInfo.h
index f434799c3982b..9af8a39aae7bd 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.h
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.h
@@ -572,6 +572,9 @@ class AArch64InstrInfo final : public AArch64GenInstrInfo {
   bool optimizePTestInstr(MachineInstr *PTest, unsigned MaskReg,
                           unsigned PredReg,
                           const MachineRegisterInfo *MRI) const;
+  std::pair<bool, unsigned>
+  canRemovePTestInstr(MachineInstr *PTest, MachineInstr *Mask,
+                      MachineInstr *Pred, const MachineRegisterInfo *MRI) const;
 };
 
 struct UsedNZCV {



More information about the llvm-commits mailing list