[llvm] [LV][EVL] Support cast instruction with EVL-vectorization (PR #108351)

via llvm-commits llvm-commits at lists.llvm.org
Mon Dec 2 19:54:09 PST 2024


https://github.com/LiqinWeng updated https://github.com/llvm/llvm-project/pull/108351

>From 28e4d5c8b4505719a4cfd773c57ba0292934d1bc Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Thu, 12 Sep 2024 17:24:35 +0800
Subject: [PATCH 1/9] [LV][EVL] Support cast instruction with EVL-vectorization

---
 .../Transforms/Vectorize/LoopVectorize.cpp    |   1 +
 llvm/lib/Transforms/Vectorize/VPlan.h         |  77 +++++-
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp |  47 ++++
 .../Transforms/Vectorize/VPlanTransforms.cpp  |   9 +
 llvm/lib/Transforms/Vectorize/VPlanValue.h    |   1 +
 .../LoopVectorize/RISCV/inloop-reduction.ll   |  18 +-
 .../RISCV/vplan-vp-cast-intrinsics.ll         | 227 ++++++++++++++++++
 7 files changed, 366 insertions(+), 14 deletions(-)
 create mode 100644 llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index d68a26251ac9d4..6b6d338cdf8ebf 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -4516,6 +4516,7 @@ static bool willGenerateVectors(VPlan &Plan, ElementCount VF,
       case VPDef::VPWidenCallSC:
       case VPDef::VPWidenCanonicalIVSC:
       case VPDef::VPWidenCastSC:
+      case VPDef::VPWidenCastEVLSC:
       case VPDef::VPWidenGEPSC:
       case VPDef::VPWidenIntrinsicSC:
       case VPDef::VPWidenSC:
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 1b1630ebc6c23a..e5035a319f3441 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -866,6 +866,7 @@ class VPSingleDefRecipe : public VPRecipeBase, public VPValue {
     case VPRecipeBase::VPWidenCallSC:
     case VPRecipeBase::VPWidenCanonicalIVSC:
     case VPRecipeBase::VPWidenCastSC:
+    case VPRecipeBase::VPWidenCastEVLSC:
     case VPRecipeBase::VPWidenGEPSC:
     case VPRecipeBase::VPWidenIntrinsicSC:
     case VPRecipeBase::VPWidenSC:
@@ -1063,6 +1064,7 @@ class VPRecipeWithIRFlags : public VPSingleDefRecipe {
            R->getVPDefID() == VPRecipeBase::VPWidenEVLSC ||
            R->getVPDefID() == VPRecipeBase::VPWidenGEPSC ||
            R->getVPDefID() == VPRecipeBase::VPWidenCastSC ||
+           R->getVPDefID() == VPRecipeBase::VPWidenCastEVLSC ||
            R->getVPDefID() == VPRecipeBase::VPReplicateSC ||
            R->getVPDefID() == VPRecipeBase::VPReverseVectorPointerSC ||
            R->getVPDefID() == VPRecipeBase::VPVectorPointerSC;
@@ -1542,19 +1544,28 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
   /// Result type for the cast.
   Type *ResultTy;
 
-public:
-  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy,
-                    CastInst &UI)
+protected:
+  VPWidenCastRecipe(unsigned VPDefOpcode, Instruction::CastOps Opcode,
+                    VPValue *Op, Type *ResultTy, CastInst &UI)
       : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op, UI), Opcode(Opcode),
         ResultTy(ResultTy) {
     assert(UI.getOpcode() == Opcode &&
            "opcode of underlying cast doesn't match");
   }
 
-  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy)
+  VPWidenCastRecipe(unsigned VPDefOpcode, Instruction::CastOps Opcode,
+                    VPValue *Op, Type *ResultTy)
       : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op), Opcode(Opcode),
         ResultTy(ResultTy) {}
 
+public:
+  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy,
+                    CastInst &UI)
+      : VPWidenCastRecipe(VPDef::VPWidenCastSC, Opcode, Op, ResultTy, UI) {}
+
+  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy)
+      : VPWidenCastRecipe(VPDef::VPWidenCastSC, Opcode, Op, ResultTy) {}
+
   ~VPWidenCastRecipe() override = default;
 
   VPWidenCastRecipe *clone() override {
@@ -1565,7 +1576,15 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
     return new VPWidenCastRecipe(Opcode, getOperand(0), ResultTy);
   }
 
-  VP_CLASSOF_IMPL(VPDef::VPWidenCastSC)
+  static inline bool classof(const VPRecipeBase *R) {
+    return R->getVPDefID() == VPRecipeBase::VPWidenCastSC ||
+           R->getVPDefID() == VPRecipeBase::VPWidenCastEVLSC;
+  }
+
+  static inline bool classof(const VPUser *U) {
+    auto *R = dyn_cast<VPRecipeBase>(U);
+    return R && classof(R);
+  }
 
   /// Produce widened copies of the cast.
   void execute(VPTransformState &State) override;
@@ -1586,6 +1605,54 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
   Type *getResultType() const { return ResultTy; }
 };
 
+// A recipe for widening cast operation with vector-predication intrinsics with
+/// explicit vector length (EVL).
+class VPWidenCastEVLRecipe : public VPWidenCastRecipe {
+  using VPRecipeWithIRFlags::transferFlags;
+
+public:
+  VPWidenCastEVLRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy,
+                       VPValue &EVL)
+      : VPWidenCastRecipe(VPDef::VPWidenCastEVLSC, Opcode, Op, ResultTy) {
+    addOperand(&EVL);
+  }
+  VPWidenCastEVLRecipe(VPWidenCastRecipe &W, VPValue &EVL)
+      : VPWidenCastEVLRecipe(W.getOpcode(), W.getOperand(0), W.getResultType(),
+                             EVL) {
+    transferFlags(W);
+  }
+
+  ~VPWidenCastEVLRecipe() override = default;
+
+  VPWidenCastEVLRecipe *clone() final {
+    llvm_unreachable("VPWidenEVLRecipe cannot be cloned");
+    return nullptr;
+  }
+
+  VP_CLASSOF_IMPL(VPDef::VPWidenCastEVLSC)
+
+  VPValue *getEVL() { return getOperand(getNumOperands() - 1); }
+  const VPValue *getEVL() const { return getOperand(getNumOperands() - 1); }
+
+  /// Produce a vp-intrinsic copies of the cast.
+  void execute(VPTransformState &State) final;
+
+  /// Returns true if the recipe only uses the first lane of operand \p Op.
+  bool onlyFirstLaneUsed(const VPValue *Op) const override {
+    assert(is_contained(operands(), Op) &&
+           "Op must be an operand of the recipe");
+    // EVL in that recipe is always the last operand, thus any use before means
+    // the VPValue should be vectorized.
+    return getEVL() == Op;
+  }
+
+#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
+  /// Print the recipe.
+  void print(raw_ostream &O, const Twine &Indent,
+             VPSlotTracker &SlotTracker) const final;
+#endif
+};
+
 /// VPScalarCastRecipe is a recipe to create scalar cast instructions.
 class VPScalarCastRecipe : public VPSingleDefRecipe {
   Instruction::CastOps Opcode;
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 3b11ec77813f3e..f8303a64ca14ef 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -92,6 +92,7 @@ bool VPRecipeBase::mayWriteToMemory() const {
   case VPVectorPointerSC:
   case VPWidenCanonicalIVSC:
   case VPWidenCastSC:
+  case VPWidenCastEVLSC:
   case VPWidenGEPSC:
   case VPWidenIntOrFpInductionSC:
   case VPWidenLoadEVLSC:
@@ -138,6 +139,7 @@ bool VPRecipeBase::mayReadFromMemory() const {
   case VPVectorPointerSC:
   case VPWidenCanonicalIVSC:
   case VPWidenCastSC:
+  case VPWidenCastEVLSC:
   case VPWidenGEPSC:
   case VPWidenIntOrFpInductionSC:
   case VPWidenPHISC:
@@ -178,6 +180,7 @@ bool VPRecipeBase::mayHaveSideEffects() const {
   case VPVectorPointerSC:
   case VPWidenCanonicalIVSC:
   case VPWidenCastSC:
+  case VPWidenCastEVLSC:
   case VPWidenGEPSC:
   case VPWidenIntOrFpInductionSC:
   case VPWidenPHISC:
@@ -1554,6 +1557,40 @@ InstructionCost VPWidenCastRecipe::computeCost(ElementCount VF,
       dyn_cast_if_present<Instruction>(getUnderlyingValue()));
 }
 
+void VPWidenCastEVLRecipe::execute(VPTransformState &State) {
+  unsigned Opcode = getOpcode();
+  State.setDebugLocFrom(getDebugLoc());
+  assert(State.UF == 1 && "Expected only UF == 1 when vectorizing with "
+                          "explicit vector length.");
+
+  assert(State.get(getOperand(0), 0)->getType()->isVectorTy() &&
+         "VPWidenCastEVLRecipe should not be used for scalars");
+
+  // TODO: add more cast instruction, eg: fptoint/inttofp/inttoptr/fptofp
+  if (Opcode == Instruction::SExt || Opcode == Instruction::ZExt ||
+      Opcode == Instruction::Trunc) {
+    Value *SrcVal = State.get(getOperand(0), 0);
+    VectorType *SrcTy = cast<VectorType>(SrcVal->getType());
+    VectorType *DsType =
+        VectorType::get(getResultType(), SrcTy->getElementCount());
+
+    IRBuilderBase &BuilderIR = State.Builder;
+    VectorBuilder Builder(BuilderIR);
+    Value *Mask = BuilderIR.CreateVectorSplat(State.VF, BuilderIR.getTrue());
+
+    Builder.setMask(Mask).setEVL(State.get(getEVL(), 0, /*NeedsScalar=*/true));
+    Value *VPInst =
+        Builder.createVectorInstruction(Opcode, DsType, {SrcVal}, "vp.cast");
+    if (VPInst) {
+      if (auto *VecOp = dyn_cast<CastInst>(VPInst))
+        VecOp->copyIRFlags(getUnderlyingInstr());
+    }
+    State.set(this, VPInst, 0);
+    State.addMetadata(VPInst,
+                      dyn_cast_or_null<Instruction>(getUnderlyingValue()));
+  }
+}
+
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
 void VPWidenCastRecipe::print(raw_ostream &O, const Twine &Indent,
                               VPSlotTracker &SlotTracker) const {
@@ -1564,6 +1601,16 @@ void VPWidenCastRecipe::print(raw_ostream &O, const Twine &Indent,
   printOperands(O, SlotTracker);
   O << " to " << *getResultType();
 }
+
+void VPWidenCastEVLRecipe::print(raw_ostream &O, const Twine &Indent,
+                                 VPSlotTracker &SlotTracker) const {
+  O << Indent << "WIDEN-VP ";
+  printAsOperand(O, SlotTracker);
+  O << " = vp." << Instruction::getOpcodeName(getOpcode()) << " ";
+  printFlags(O);
+  printOperands(O, SlotTracker);
+  O << " to " << *getResultType();
+}
 #endif
 
 InstructionCost VPHeaderPHIRecipe::computeCost(ElementCount VF,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index ba791c679b4ae4..57801b90739a3f 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1477,6 +1477,15 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                   return nullptr;
                 return new VPWidenEVLRecipe(*W, EVL);
               })
+              .Case<VPWidenCastRecipe>(
+                  [&](VPWidenCastRecipe *W) -> VPRecipeBase * {
+                    unsigned Opcode = W->getOpcode();
+                    if (Opcode != Instruction::SExt &&
+                        Opcode != Instruction::ZExt &&
+                        Opcode != Instruction::Trunc)
+                      return nullptr;
+                    return new VPWidenCastEVLRecipe(*W, EVL);
+                  })
               .Case<VPReductionRecipe>([&](VPReductionRecipe *Red) {
                 VPValue *NewMask = GetNewMask(Red->getCondOp());
                 return new VPReductionEVLRecipe(*Red, EVL, NewMask);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h
index 691b0d40823cfb..97fc2e88f412d4 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h
@@ -337,6 +337,7 @@ class VPDef {
     VPWidenCallSC,
     VPWidenCanonicalIVSC,
     VPWidenCastSC,
+    VPWidenCastEVLSC,
     VPWidenGEPSC,
     VPWidenIntrinsicSC,
     VPWidenLoadEVLSC,
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
index 8368022bceed92..9ee1708f175ab0 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
@@ -194,16 +194,16 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-INLOOP-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-INLOOP-NEXT:    [[EVL_BASED_IV:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
-; IF-EVL-INLOOP-NEXT:    [[AVL:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
-; IF-EVL-INLOOP-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 8, i1 true)
-; IF-EVL-INLOOP-NEXT:    [[TMP6:%.*]] = add i32 [[EVL_BASED_IV]], 0
-; IF-EVL-INLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP6]]
-; IF-EVL-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP8]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP5]])
-; IF-EVL-INLOOP-NEXT:    [[TMP9:%.*]] = sext <vscale x 8 x i16> [[VP_OP_LOAD]] to <vscale x 8 x i32>
-; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[TMP9]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-INLOOP-NEXT:    [[TMP5:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-INLOOP-NEXT:    [[TMP6:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[TMP5]], i32 8, i1 true)
+; IF-EVL-INLOOP-NEXT:    [[TMP7:%.*]] = add i32 [[EVL_BASED_IV]], 0
+; IF-EVL-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP7]]
+; IF-EVL-INLOOP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP9]], <vscale x 8 x i1> 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), i32 [[TMP6]])
+; IF-EVL-INLOOP-NEXT:    [[VP_CAST:%.*]] = call <vscale x 8 x i32> @llvm.vp.sext.nxv8i32.nxv8i16(<vscale x 8 x i16> [[VP_OP_LOAD]], <vscale x 8 x i1> 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), i32 [[TMP6]])
+; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[VP_CAST]], <vscale x 8 x i1> 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), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP11]] = add i32 [[TMP10]], [[VEC_PHI]]
-; IF-EVL-INLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP5]], [[EVL_BASED_IV]]
+; IF-EVL-INLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP6]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP4]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; IF-EVL-INLOOP-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
new file mode 100644
index 00000000000000..f1f3d2c88301e1
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -0,0 +1,227 @@
+; REQUIRES: asserts
+
+; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
+; RUN: -force-tail-folding-style=data-with-evl \
+; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s
+
+; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
+; RUN: -force-tail-folding-style=none \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=NO-VP %s
+
+define void @vp_sext(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<%0> = VF * UF
+; IF-EVL-NEXT: Live-in vp<%1> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:  vector.body:
+; IF-EVL-NEXT:    EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%12>
+; IF-EVL-NEXT:    EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%4> = phi ir<0>, vp<%11>
+; IF-EVL-NEXT:    EMIT vp<%5> = EXPLICIT-VECTOR-LENGTH vp<%4>, ir<%N>
+; IF-EVL-NEXT:    vp<%6> = SCALAR-STEPS vp<%4>, ir<1>
+; IF-EVL-NEXT:    CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%6>
+; IF-EVL-NEXT:    vp<%7> = vector-pointer ir<%arrayidx>
+; IF-EVL-NEXT:    WIDEN ir<%0> = vp.load vp<%7>, vp<%5>
+; IF-EVL-NEXT:    WIDEN-VP vp<%8> = vp.sext  ir<%0>, vp<%5> to i64
+; IF-EVL-NEXT:    CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%6>
+; IF-EVL-NEXT:    vp<%9> = vector-pointer ir<%arrayidx4>
+; IF-EVL-NEXT:    WIDEN vp.store vp<%9>, vp<%8>, vp<%5>
+; IF-EVL-NEXT:    SCALAR-CAST vp<%10> = zext vp<%5> to i64
+; IF-EVL-NEXT:    EMIT vp<%11> = add vp<%10>, vp<%4>
+; IF-EVL-NEXT:    EMIT vp<%12> = add vp<%3>, vp<%0>
+; IF-EVL-NEXT:    EMIT branch-on-count vp<%12>, vp<%1>
+; IF-EVL-NEXT:  No successors
+; IF-EVL-NEXT: }
+
+; NO-VP: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF>=1' {
+; NO-VP-NEXT: Live-in vp<%0> = VF * UF
+; NO-VP-NEXT: Live-in vp<%1> = vector-trip-count
+; NO-VP-NEXT: Live-in ir<%N> = original trip-count
+ 
+; NO-VP: vector.ph:
+; NO-VP-NEXT: Successor(s): vector loop
+
+; NO-VP: <x1> vector loop: {
+; NO-VP-NEXT:   vector.body:
+; NO-VP-NEXT:     EMIT vp<%2> = CANONICAL-INDUCTION ir<0>, vp<%6>
+; NO-VP-NEXT:     vp<%3> = SCALAR-STEPS vp<%2>, ir<1>
+; NO-VP-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%3>
+; NO-VP-NEXT:     vp<%4> = vector-pointer ir<%arrayidx>
+; NO-VP-NEXT:     WIDEN ir<%0> = load vp<%4>
+; NO-VP-NEXT:     WIDEN-CAST ir<%conv2> = sext  ir<%0> to i64
+; NO-VP-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%3>
+; NO-VP-NEXT:     vp<%5> = vector-pointer ir<%arrayidx4>
+; NO-VP-NEXT:     WIDEN store vp<%5>, ir<%conv2>
+; NO-VP-NEXT:     EMIT vp<%6> = add nuw vp<%2>, vp<%0>
+; NO-VP-NEXT:     EMIT branch-on-count vp<%6>, vp<%1>
+; NO-VP-NEXT:   No successors
+; NO-VP-NEXT: }
+
+entry:
+  %cmp8 = icmp sgt i64 %N, 0
+  br i1 %cmp8, label %for.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+
+for.body:
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+  %arrayidx = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
+  %0 = load i32, ptr %arrayidx, align 4
+  %conv2 = sext i32 %0 to i64
+  %arrayidx4 = getelementptr inbounds i64, ptr %a, i64 %indvars.iv
+  store i64 %conv2, ptr %arrayidx4, align 8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %N
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+}
+
+define void @vp_zext(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<%0> = VF * UF
+; IF-EVL-NEXT: Live-in vp<%1> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%12>
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%4> = phi ir<0>, vp<%11>
+; IF-EVL-NEXT:     EMIT vp<%5> = EXPLICIT-VECTOR-LENGTH vp<%4>, ir<%N>
+; IF-EVL-NEXT:     vp<%6> = SCALAR-STEPS vp<%4>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%6>
+; IF-EVL-NEXT:     vp<%7> = vector-pointer ir<%arrayidx>
+; IF-EVL-NEXT:     WIDEN ir<%0> = vp.load vp<%7>, vp<%5>
+; IF-EVL-NEXT:     WIDEN-VP vp<%8> = vp.zext  ir<%0>, vp<%5> to i64
+; IF-EVL-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%6>
+; IF-EVL-NEXT:     vp<%9> = vector-pointer ir<%arrayidx4>
+; IF-EVL-NEXT:     WIDEN vp.store vp<%9>, vp<%8>, vp<%5>
+; IF-EVL-NEXT:     SCALAR-CAST vp<%10> = zext vp<%5> to i64
+; IF-EVL-NEXT:     EMIT vp<%11> = add vp<%10>, vp<%4>
+; IF-EVL-NEXT:     EMIT vp<%12> = add vp<%3>, vp<%0>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<%12>, vp<%1>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
+
+; NO-VP: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF>=1' {
+; NO-VP-NEXT: Live-in vp<%0> = VF * UF
+; NO-VP-NEXT: Live-in vp<%1> = vector-trip-count
+; NO-VP-NEXT: Live-in ir<%N> = original trip-count
+
+; NO-VP: vector.ph:
+; NO-VP-NEXT: Successor(s): vector loop
+
+; NO-VP: <x1> vector loop: {
+; NO-VP-NEXT:   vector.body:
+; NO-VP-NEXT:     EMIT vp<%2> = CANONICAL-INDUCTION ir<0>, vp<%6>
+; NO-VP-NEXT:     vp<%3> = SCALAR-STEPS vp<%2>, ir<1>
+; NO-VP-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%3>
+; NO-VP-NEXT:     vp<%4> = vector-pointer ir<%arrayidx>
+; NO-VP-NEXT:     WIDEN ir<%0> = load vp<%4>
+; NO-VP-NEXT:     WIDEN-CAST ir<%conv2> = zext  ir<%0> to i64
+; NO-VP-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%3>
+; NO-VP-NEXT:     vp<%5> = vector-pointer ir<%arrayidx4>
+; NO-VP-NEXT:     WIDEN store vp<%5>, ir<%conv2>
+; NO-VP-NEXT:     EMIT vp<%6> = add nuw vp<%2>, vp<%0>
+; NO-VP-NEXT:     EMIT branch-on-count vp<%6>, vp<%1>
+; NO-VP-NEXT:   No successors
+; NO-VP-NEXT: }
+
+entry:
+  %cmp8 = icmp sgt i64 %N, 0
+  br i1 %cmp8, label %for.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+
+for.body:
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+  %arrayidx = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
+  %0 = load i32, ptr %arrayidx, align 4
+  %conv2 = zext i32 %0 to i64
+  %arrayidx4 = getelementptr inbounds i64, ptr %a, i64 %indvars.iv
+  store i64 %conv2, ptr %arrayidx4, align 8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %N
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+}
+
+define void @vp_truncate(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL : VPlan 'Initial VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
+; IF-EVL-NEXT : Live-in vp<%0> = VF * UF
+; IF-EVL-NEXT : Live-in vp<%1> = vector-trip-count
+; IF-EVL-NEXT : Live-in ir<%N> = original trip-count
+
+; IF-EVL : vector.ph:
+; IF-EVL-NEXT : Successor(s): vector loop
+
+; IF-EVL : <x1> vector loop: {
+; IF-EVL-NEXT :   vector.body:
+; IF-EVL-NEXT :     EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%12>
+; IF-EVL-NEXT :     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%4> = phi ir<0>, vp<%11>
+; IF-EVL-NEXT :     EMIT vp<%5> = EXPLICIT-VECTOR-LENGTH vp<%4>, ir<%N>
+; IF-EVL-NEXT :     vp<%6> = SCALAR-STEPS vp<%4>, ir<1>
+; IF-EVL-NEXT :     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%6>
+; IF-EVL-NEXT :     vp<%7> = vector-pointer ir<%arrayidx>
+; IF-EVL-NEXT :     WIDEN ir<%0> = vp.load vp<%7>, vp<%5>
+; IF-EVL-NEXT :     WIDEN-VP vp<%8> = vp.trunc  ir<%0>, vp<%5> to i16
+; IF-EVL-NEXT :     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%6>
+; IF-EVL-NEXT :     vp<%9> = vector-pointer ir<%arrayidx4>
+; IF-EVL-NEXT :     WIDEN vp.store vp<%9>, vp<%8>, vp<%5>
+; IF-EVL-NEXT :     SCALAR-CAST vp<%10> = zext vp<%5> to i64
+; IF-EVL-NEXT :     EMIT vp<%11> = add vp<%10>, vp<%4>
+; IF-EVL-NEXT :     EMIT vp<%12> = add vp<%3>, vp<%0>
+; IF-EVL-NEXT :     EMIT branch-on-count vp<%12>, vp<%1>
+; IF-EVL-NEXT :   No successors
+; IF-EVL-NEXT : }
+
+; NO-VP: Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF>=1' {
+; NO-VP-NEXT: Live-in vp<%0> = VF * UF
+; NO-VP-NEXT: Live-in vp<%1> = vector-trip-count
+; NO-VP-NEXT: Live-in ir<%N> = original trip-count
+
+; NO-VP: vector.ph:
+; NO-VP-NEXT: Successor(s): vector loop
+
+; NO-VP: <x1> vector loop: {
+; NO-VP:   vector.body:
+; NO-VP-NEXT:     EMIT vp<%2> = CANONICAL-INDUCTION ir<0>, vp<%6>
+; NO-VP-NEXT:     vp<%3> = SCALAR-STEPS vp<%2>, ir<1>
+; NO-VP-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%3>
+; NO-VP-NEXT:     vp<%4> = vector-pointer ir<%arrayidx>
+; NO-VP-NEXT:     WIDEN ir<%0> = load vp<%4>
+; NO-VP-NEXT:     WIDEN-CAST ir<%conv2> = trunc  ir<%0> to i16
+; NO-VP-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%3>
+; NO-VP-NEXT:     vp<%5> = vector-pointer ir<%arrayidx4>
+; NO-VP-NEXT:     WIDEN store vp<%5>, ir<%conv2>
+; NO-VP-NEXT:     EMIT vp<%6> = add nuw vp<%2>, vp<%0>
+; NO-VP-NEXT:     EMIT branch-on-count vp<%6>, vp<%1>
+; NO-VP-NEXT:   No successors
+; NO-VP-NEXT: }
+
+entry:
+  %cmp8 = icmp sgt i64 %N, 0
+  br i1 %cmp8, label %for.body, label %for.cond.cleanup
+
+for.cond.cleanup:
+  ret void
+
+for.body:
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+  %arrayidx = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
+  %0 = load i32, ptr %arrayidx, align 4
+  %conv2 = trunc i32 %0 to i16
+  %arrayidx4 = getelementptr inbounds i16, ptr %a, i64 %indvars.iv
+  store i16 %conv2, ptr %arrayidx4, align 2
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %N
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+}

>From 4d3be159fabe574c31902dd077e6eb9316d27053 Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Fri, 18 Oct 2024 14:02:54 +0800
Subject: [PATCH 2/9] Rebase and Revert the implement of castEVLRecipe

---
 .../Transforms/Vectorize/LoopVectorize.cpp    |   1 -
 llvm/lib/Transforms/Vectorize/VPlan.h         |  77 +--
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp |  47 --
 .../Transforms/Vectorize/VPlanTransforms.cpp  |   9 -
 llvm/lib/Transforms/Vectorize/VPlanValue.h    |   1 -
 .../LoopVectorize/RISCV/inloop-reduction.ll   |   2 +-
 .../RISCV/vplan-vp-cast-intrinsics.ll         | 526 ++++++++++++------
 7 files changed, 351 insertions(+), 312 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 6b6d338cdf8ebf..d68a26251ac9d4 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -4516,7 +4516,6 @@ static bool willGenerateVectors(VPlan &Plan, ElementCount VF,
       case VPDef::VPWidenCallSC:
       case VPDef::VPWidenCanonicalIVSC:
       case VPDef::VPWidenCastSC:
-      case VPDef::VPWidenCastEVLSC:
       case VPDef::VPWidenGEPSC:
       case VPDef::VPWidenIntrinsicSC:
       case VPDef::VPWidenSC:
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index e5035a319f3441..1b1630ebc6c23a 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -866,7 +866,6 @@ class VPSingleDefRecipe : public VPRecipeBase, public VPValue {
     case VPRecipeBase::VPWidenCallSC:
     case VPRecipeBase::VPWidenCanonicalIVSC:
     case VPRecipeBase::VPWidenCastSC:
-    case VPRecipeBase::VPWidenCastEVLSC:
     case VPRecipeBase::VPWidenGEPSC:
     case VPRecipeBase::VPWidenIntrinsicSC:
     case VPRecipeBase::VPWidenSC:
@@ -1064,7 +1063,6 @@ class VPRecipeWithIRFlags : public VPSingleDefRecipe {
            R->getVPDefID() == VPRecipeBase::VPWidenEVLSC ||
            R->getVPDefID() == VPRecipeBase::VPWidenGEPSC ||
            R->getVPDefID() == VPRecipeBase::VPWidenCastSC ||
-           R->getVPDefID() == VPRecipeBase::VPWidenCastEVLSC ||
            R->getVPDefID() == VPRecipeBase::VPReplicateSC ||
            R->getVPDefID() == VPRecipeBase::VPReverseVectorPointerSC ||
            R->getVPDefID() == VPRecipeBase::VPVectorPointerSC;
@@ -1544,28 +1542,19 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
   /// Result type for the cast.
   Type *ResultTy;
 
-protected:
-  VPWidenCastRecipe(unsigned VPDefOpcode, Instruction::CastOps Opcode,
-                    VPValue *Op, Type *ResultTy, CastInst &UI)
+public:
+  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy,
+                    CastInst &UI)
       : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op, UI), Opcode(Opcode),
         ResultTy(ResultTy) {
     assert(UI.getOpcode() == Opcode &&
            "opcode of underlying cast doesn't match");
   }
 
-  VPWidenCastRecipe(unsigned VPDefOpcode, Instruction::CastOps Opcode,
-                    VPValue *Op, Type *ResultTy)
+  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy)
       : VPRecipeWithIRFlags(VPDef::VPWidenCastSC, Op), Opcode(Opcode),
         ResultTy(ResultTy) {}
 
-public:
-  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy,
-                    CastInst &UI)
-      : VPWidenCastRecipe(VPDef::VPWidenCastSC, Opcode, Op, ResultTy, UI) {}
-
-  VPWidenCastRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy)
-      : VPWidenCastRecipe(VPDef::VPWidenCastSC, Opcode, Op, ResultTy) {}
-
   ~VPWidenCastRecipe() override = default;
 
   VPWidenCastRecipe *clone() override {
@@ -1576,15 +1565,7 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
     return new VPWidenCastRecipe(Opcode, getOperand(0), ResultTy);
   }
 
-  static inline bool classof(const VPRecipeBase *R) {
-    return R->getVPDefID() == VPRecipeBase::VPWidenCastSC ||
-           R->getVPDefID() == VPRecipeBase::VPWidenCastEVLSC;
-  }
-
-  static inline bool classof(const VPUser *U) {
-    auto *R = dyn_cast<VPRecipeBase>(U);
-    return R && classof(R);
-  }
+  VP_CLASSOF_IMPL(VPDef::VPWidenCastSC)
 
   /// Produce widened copies of the cast.
   void execute(VPTransformState &State) override;
@@ -1605,54 +1586,6 @@ class VPWidenCastRecipe : public VPRecipeWithIRFlags {
   Type *getResultType() const { return ResultTy; }
 };
 
-// A recipe for widening cast operation with vector-predication intrinsics with
-/// explicit vector length (EVL).
-class VPWidenCastEVLRecipe : public VPWidenCastRecipe {
-  using VPRecipeWithIRFlags::transferFlags;
-
-public:
-  VPWidenCastEVLRecipe(Instruction::CastOps Opcode, VPValue *Op, Type *ResultTy,
-                       VPValue &EVL)
-      : VPWidenCastRecipe(VPDef::VPWidenCastEVLSC, Opcode, Op, ResultTy) {
-    addOperand(&EVL);
-  }
-  VPWidenCastEVLRecipe(VPWidenCastRecipe &W, VPValue &EVL)
-      : VPWidenCastEVLRecipe(W.getOpcode(), W.getOperand(0), W.getResultType(),
-                             EVL) {
-    transferFlags(W);
-  }
-
-  ~VPWidenCastEVLRecipe() override = default;
-
-  VPWidenCastEVLRecipe *clone() final {
-    llvm_unreachable("VPWidenEVLRecipe cannot be cloned");
-    return nullptr;
-  }
-
-  VP_CLASSOF_IMPL(VPDef::VPWidenCastEVLSC)
-
-  VPValue *getEVL() { return getOperand(getNumOperands() - 1); }
-  const VPValue *getEVL() const { return getOperand(getNumOperands() - 1); }
-
-  /// Produce a vp-intrinsic copies of the cast.
-  void execute(VPTransformState &State) final;
-
-  /// Returns true if the recipe only uses the first lane of operand \p Op.
-  bool onlyFirstLaneUsed(const VPValue *Op) const override {
-    assert(is_contained(operands(), Op) &&
-           "Op must be an operand of the recipe");
-    // EVL in that recipe is always the last operand, thus any use before means
-    // the VPValue should be vectorized.
-    return getEVL() == Op;
-  }
-
-#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
-  /// Print the recipe.
-  void print(raw_ostream &O, const Twine &Indent,
-             VPSlotTracker &SlotTracker) const final;
-#endif
-};
-
 /// VPScalarCastRecipe is a recipe to create scalar cast instructions.
 class VPScalarCastRecipe : public VPSingleDefRecipe {
   Instruction::CastOps Opcode;
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index f8303a64ca14ef..3b11ec77813f3e 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -92,7 +92,6 @@ bool VPRecipeBase::mayWriteToMemory() const {
   case VPVectorPointerSC:
   case VPWidenCanonicalIVSC:
   case VPWidenCastSC:
-  case VPWidenCastEVLSC:
   case VPWidenGEPSC:
   case VPWidenIntOrFpInductionSC:
   case VPWidenLoadEVLSC:
@@ -139,7 +138,6 @@ bool VPRecipeBase::mayReadFromMemory() const {
   case VPVectorPointerSC:
   case VPWidenCanonicalIVSC:
   case VPWidenCastSC:
-  case VPWidenCastEVLSC:
   case VPWidenGEPSC:
   case VPWidenIntOrFpInductionSC:
   case VPWidenPHISC:
@@ -180,7 +178,6 @@ bool VPRecipeBase::mayHaveSideEffects() const {
   case VPVectorPointerSC:
   case VPWidenCanonicalIVSC:
   case VPWidenCastSC:
-  case VPWidenCastEVLSC:
   case VPWidenGEPSC:
   case VPWidenIntOrFpInductionSC:
   case VPWidenPHISC:
@@ -1557,40 +1554,6 @@ InstructionCost VPWidenCastRecipe::computeCost(ElementCount VF,
       dyn_cast_if_present<Instruction>(getUnderlyingValue()));
 }
 
-void VPWidenCastEVLRecipe::execute(VPTransformState &State) {
-  unsigned Opcode = getOpcode();
-  State.setDebugLocFrom(getDebugLoc());
-  assert(State.UF == 1 && "Expected only UF == 1 when vectorizing with "
-                          "explicit vector length.");
-
-  assert(State.get(getOperand(0), 0)->getType()->isVectorTy() &&
-         "VPWidenCastEVLRecipe should not be used for scalars");
-
-  // TODO: add more cast instruction, eg: fptoint/inttofp/inttoptr/fptofp
-  if (Opcode == Instruction::SExt || Opcode == Instruction::ZExt ||
-      Opcode == Instruction::Trunc) {
-    Value *SrcVal = State.get(getOperand(0), 0);
-    VectorType *SrcTy = cast<VectorType>(SrcVal->getType());
-    VectorType *DsType =
-        VectorType::get(getResultType(), SrcTy->getElementCount());
-
-    IRBuilderBase &BuilderIR = State.Builder;
-    VectorBuilder Builder(BuilderIR);
-    Value *Mask = BuilderIR.CreateVectorSplat(State.VF, BuilderIR.getTrue());
-
-    Builder.setMask(Mask).setEVL(State.get(getEVL(), 0, /*NeedsScalar=*/true));
-    Value *VPInst =
-        Builder.createVectorInstruction(Opcode, DsType, {SrcVal}, "vp.cast");
-    if (VPInst) {
-      if (auto *VecOp = dyn_cast<CastInst>(VPInst))
-        VecOp->copyIRFlags(getUnderlyingInstr());
-    }
-    State.set(this, VPInst, 0);
-    State.addMetadata(VPInst,
-                      dyn_cast_or_null<Instruction>(getUnderlyingValue()));
-  }
-}
-
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
 void VPWidenCastRecipe::print(raw_ostream &O, const Twine &Indent,
                               VPSlotTracker &SlotTracker) const {
@@ -1601,16 +1564,6 @@ void VPWidenCastRecipe::print(raw_ostream &O, const Twine &Indent,
   printOperands(O, SlotTracker);
   O << " to " << *getResultType();
 }
-
-void VPWidenCastEVLRecipe::print(raw_ostream &O, const Twine &Indent,
-                                 VPSlotTracker &SlotTracker) const {
-  O << Indent << "WIDEN-VP ";
-  printAsOperand(O, SlotTracker);
-  O << " = vp." << Instruction::getOpcodeName(getOpcode()) << " ";
-  printFlags(O);
-  printOperands(O, SlotTracker);
-  O << " to " << *getResultType();
-}
 #endif
 
 InstructionCost VPHeaderPHIRecipe::computeCost(ElementCount VF,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 57801b90739a3f..ba791c679b4ae4 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1477,15 +1477,6 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                   return nullptr;
                 return new VPWidenEVLRecipe(*W, EVL);
               })
-              .Case<VPWidenCastRecipe>(
-                  [&](VPWidenCastRecipe *W) -> VPRecipeBase * {
-                    unsigned Opcode = W->getOpcode();
-                    if (Opcode != Instruction::SExt &&
-                        Opcode != Instruction::ZExt &&
-                        Opcode != Instruction::Trunc)
-                      return nullptr;
-                    return new VPWidenCastEVLRecipe(*W, EVL);
-                  })
               .Case<VPReductionRecipe>([&](VPReductionRecipe *Red) {
                 VPValue *NewMask = GetNewMask(Red->getCondOp());
                 return new VPReductionEVLRecipe(*Red, EVL, NewMask);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanValue.h b/llvm/lib/Transforms/Vectorize/VPlanValue.h
index 97fc2e88f412d4..691b0d40823cfb 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanValue.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanValue.h
@@ -337,7 +337,6 @@ class VPDef {
     VPWidenCallSC,
     VPWidenCanonicalIVSC,
     VPWidenCastSC,
-    VPWidenCastEVLSC,
     VPWidenGEPSC,
     VPWidenIntrinsicSC,
     VPWidenLoadEVLSC,
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
index 9ee1708f175ab0..d273a88a84feaa 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
@@ -200,7 +200,7 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP7]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
 ; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP9]], <vscale x 8 x i1> 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), i32 [[TMP6]])
-; IF-EVL-INLOOP-NEXT:    [[VP_CAST:%.*]] = call <vscale x 8 x i32> @llvm.vp.sext.nxv8i32.nxv8i16(<vscale x 8 x i16> [[VP_OP_LOAD]], <vscale x 8 x i1> 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), i32 [[TMP6]])
+; IF-EVL-INLOOP-NEXT:    [[VP_CAST:%.*]] = sext <vscale x 8 x i16> [[VP_OP_LOAD]] to <vscale x 8 x i32>
 ; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[VP_CAST]], <vscale x 8 x i1> 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), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP11]] = add i32 [[TMP10]], [[VEC_PHI]]
 ; IF-EVL-INLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP6]], [[EVL_BASED_IV]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
index f1f3d2c88301e1..6894a30f30efe4 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -1,91 +1,63 @@
 ; REQUIRES: asserts
-
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
 ; RUN: -force-tail-folding-style=data-with-evl \
 ; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s
 
-; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
-; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=NO-VP %s
-
 define void @vp_sext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
-; IF-EVL-NEXT: Live-in vp<%0> = VF * UF
-; IF-EVL-NEXT: Live-in vp<%1> = vector-trip-count
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
 ; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
 
 ; IF-EVL: vector.ph:
 ; IF-EVL-NEXT: Successor(s): vector loop
 
-; IF-EVL: <x1> vector loop: {
-; IF-EVL-NEXT:  vector.body:
-; IF-EVL-NEXT:    EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%12>
-; IF-EVL-NEXT:    EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%4> = phi ir<0>, vp<%11>
-; IF-EVL-NEXT:    EMIT vp<%5> = EXPLICIT-VECTOR-LENGTH vp<%4>, ir<%N>
-; IF-EVL-NEXT:    vp<%6> = SCALAR-STEPS vp<%4>, ir<1>
-; IF-EVL-NEXT:    CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%6>
-; IF-EVL-NEXT:    vp<%7> = vector-pointer ir<%arrayidx>
-; IF-EVL-NEXT:    WIDEN ir<%0> = vp.load vp<%7>, vp<%5>
-; IF-EVL-NEXT:    WIDEN-VP vp<%8> = vp.sext  ir<%0>, vp<%5> to i64
-; IF-EVL-NEXT:    CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%6>
-; IF-EVL-NEXT:    vp<%9> = vector-pointer ir<%arrayidx4>
-; IF-EVL-NEXT:    WIDEN vp.store vp<%9>, vp<%8>, vp<%5>
-; IF-EVL-NEXT:    SCALAR-CAST vp<%10> = zext vp<%5> to i64
-; IF-EVL-NEXT:    EMIT vp<%11> = add vp<%10>, vp<%4>
-; IF-EVL-NEXT:    EMIT vp<%12> = add vp<%3>, vp<%0>
-; IF-EVL-NEXT:    EMIT branch-on-count vp<%12>, vp<%1>
-; IF-EVL-NEXT:  No successors
+; IF-EVL:   <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[SEXT:%.+]]> = sext  ir<[[LD1]]> to i64
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
 ; IF-EVL-NEXT: }
+; IF-EVL-NEXT: Successor(s): middle.block
 
-; NO-VP: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF>=1' {
-; NO-VP-NEXT: Live-in vp<%0> = VF * UF
-; NO-VP-NEXT: Live-in vp<%1> = vector-trip-count
-; NO-VP-NEXT: Live-in ir<%N> = original trip-count
- 
-; NO-VP: vector.ph:
-; NO-VP-NEXT: Successor(s): vector loop
-
-; NO-VP: <x1> vector loop: {
-; NO-VP-NEXT:   vector.body:
-; NO-VP-NEXT:     EMIT vp<%2> = CANONICAL-INDUCTION ir<0>, vp<%6>
-; NO-VP-NEXT:     vp<%3> = SCALAR-STEPS vp<%2>, ir<1>
-; NO-VP-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%3>
-; NO-VP-NEXT:     vp<%4> = vector-pointer ir<%arrayidx>
-; NO-VP-NEXT:     WIDEN ir<%0> = load vp<%4>
-; NO-VP-NEXT:     WIDEN-CAST ir<%conv2> = sext  ir<%0> to i64
-; NO-VP-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%3>
-; NO-VP-NEXT:     vp<%5> = vector-pointer ir<%arrayidx4>
-; NO-VP-NEXT:     WIDEN store vp<%5>, ir<%conv2>
-; NO-VP-NEXT:     EMIT vp<%6> = add nuw vp<%2>, vp<%0>
-; NO-VP-NEXT:     EMIT branch-on-count vp<%6>, vp<%1>
-; NO-VP-NEXT:   No successors
-; NO-VP-NEXT: }
 
 entry:
-  %cmp8 = icmp sgt i64 %N, 0
-  br i1 %cmp8, label %for.body, label %for.cond.cleanup
-
-for.cond.cleanup:
-  ret void
+  br label %loop
 
-for.body:
-  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
-  %arrayidx = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
-  %0 = load i32, ptr %arrayidx, align 4
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
   %conv2 = sext i32 %0 to i64
-  %arrayidx4 = getelementptr inbounds i64, ptr %a, i64 %indvars.iv
-  store i64 %conv2, ptr %arrayidx4, align 8
-  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
-  %exitcond.not = icmp eq i64 %indvars.iv.next, %N
-  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+  %gep4 = getelementptr inbounds i64, ptr %a, i64 %iv
+  store i64 %conv2, ptr %gep4, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
 }
 
 define void @vp_zext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
-; IF-EVL-NEXT: Live-in vp<%0> = VF * UF
-; IF-EVL-NEXT: Live-in vp<%1> = vector-trip-count
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
 ; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
 
 ; IF-EVL: vector.ph:
@@ -93,135 +65,327 @@ define void @vp_zext(ptr noalias %a, ptr noalias %b, i64 %N) {
 
 ; IF-EVL: <x1> vector loop: {
 ; IF-EVL-NEXT:   vector.body:
-; IF-EVL-NEXT:     EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%12>
-; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%4> = phi ir<0>, vp<%11>
-; IF-EVL-NEXT:     EMIT vp<%5> = EXPLICIT-VECTOR-LENGTH vp<%4>, ir<%N>
-; IF-EVL-NEXT:     vp<%6> = SCALAR-STEPS vp<%4>, ir<1>
-; IF-EVL-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%6>
-; IF-EVL-NEXT:     vp<%7> = vector-pointer ir<%arrayidx>
-; IF-EVL-NEXT:     WIDEN ir<%0> = vp.load vp<%7>, vp<%5>
-; IF-EVL-NEXT:     WIDEN-VP vp<%8> = vp.zext  ir<%0>, vp<%5> to i64
-; IF-EVL-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%6>
-; IF-EVL-NEXT:     vp<%9> = vector-pointer ir<%arrayidx4>
-; IF-EVL-NEXT:     WIDEN vp.store vp<%9>, vp<%8>, vp<%5>
-; IF-EVL-NEXT:     SCALAR-CAST vp<%10> = zext vp<%5> to i64
-; IF-EVL-NEXT:     EMIT vp<%11> = add vp<%10>, vp<%4>
-; IF-EVL-NEXT:     EMIT vp<%12> = add vp<%3>, vp<%0>
-; IF-EVL-NEXT:     EMIT branch-on-count vp<%12>, vp<%1>
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[ZEXT:%.+]]> = zext ir<[[LD1]]> to i64
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[ZEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
 ; IF-EVL-NEXT:   No successors
 ; IF-EVL-NEXT: }
 
-; NO-VP: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF>=1' {
-; NO-VP-NEXT: Live-in vp<%0> = VF * UF
-; NO-VP-NEXT: Live-in vp<%1> = vector-trip-count
-; NO-VP-NEXT: Live-in ir<%N> = original trip-count
-
-; NO-VP: vector.ph:
-; NO-VP-NEXT: Successor(s): vector loop
-
-; NO-VP: <x1> vector loop: {
-; NO-VP-NEXT:   vector.body:
-; NO-VP-NEXT:     EMIT vp<%2> = CANONICAL-INDUCTION ir<0>, vp<%6>
-; NO-VP-NEXT:     vp<%3> = SCALAR-STEPS vp<%2>, ir<1>
-; NO-VP-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%3>
-; NO-VP-NEXT:     vp<%4> = vector-pointer ir<%arrayidx>
-; NO-VP-NEXT:     WIDEN ir<%0> = load vp<%4>
-; NO-VP-NEXT:     WIDEN-CAST ir<%conv2> = zext  ir<%0> to i64
-; NO-VP-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%3>
-; NO-VP-NEXT:     vp<%5> = vector-pointer ir<%arrayidx4>
-; NO-VP-NEXT:     WIDEN store vp<%5>, ir<%conv2>
-; NO-VP-NEXT:     EMIT vp<%6> = add nuw vp<%2>, vp<%0>
-; NO-VP-NEXT:     EMIT branch-on-count vp<%6>, vp<%1>
-; NO-VP-NEXT:   No successors
-; NO-VP-NEXT: }
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv2 = zext i32 %0 to i64
+  %gep4 = getelementptr inbounds i64, ptr %a, i64 %iv
+  store i64 %conv2, ptr %gep4, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_truncate(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Initial VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[TRUNC:%.+]]> = trunc ir<[[LD1]]> to i16
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[TRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>, vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
 
 entry:
-  %cmp8 = icmp sgt i64 %N, 0
-  br i1 %cmp8, label %for.body, label %for.cond.cleanup
+  br label %loop
 
-for.cond.cleanup:
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv2 = trunc i32 %0 to i16
+  %gep4 = getelementptr inbounds i16, ptr %a, i64 %iv
+  store i16 %conv2, ptr %gep4, align 2
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
   ret void
+}
 
-for.body:
-  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
-  %arrayidx = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
-  %0 = load i32, ptr %arrayidx, align 4
-  %conv2 = zext i32 %0 to i64
-  %arrayidx4 = getelementptr inbounds i64, ptr %a, i64 %indvars.iv
-  store i64 %conv2, ptr %arrayidx4, align 8
-  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
-  %exitcond.not = icmp eq i64 %indvars.iv.next, %N
-  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+define void @vp_fpext(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPEXT:%.+]]> = fpext ir<[[LD1]]> to 
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds float, ptr %b, i64 %iv
+  %0 = load float, ptr %gep, align 4
+  %conv2 = fpext float %0 to double
+  %gep4 = getelementptr inbounds double, ptr %a, i64 %iv
+  store double %conv2, ptr %gep4, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
 }
 
-define void @vp_truncate(ptr noalias %a, ptr noalias %b, i64 %N) {
-; IF-EVL : VPlan 'Initial VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
-; IF-EVL-NEXT : Live-in vp<%0> = VF * UF
-; IF-EVL-NEXT : Live-in vp<%1> = vector-trip-count
-; IF-EVL-NEXT : Live-in ir<%N> = original trip-count
-
-; IF-EVL : vector.ph:
-; IF-EVL-NEXT : Successor(s): vector loop
-
-; IF-EVL : <x1> vector loop: {
-; IF-EVL-NEXT :   vector.body:
-; IF-EVL-NEXT :     EMIT vp<%3> = CANONICAL-INDUCTION ir<0>, vp<%12>
-; IF-EVL-NEXT :     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<%4> = phi ir<0>, vp<%11>
-; IF-EVL-NEXT :     EMIT vp<%5> = EXPLICIT-VECTOR-LENGTH vp<%4>, ir<%N>
-; IF-EVL-NEXT :     vp<%6> = SCALAR-STEPS vp<%4>, ir<1>
-; IF-EVL-NEXT :     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%6>
-; IF-EVL-NEXT :     vp<%7> = vector-pointer ir<%arrayidx>
-; IF-EVL-NEXT :     WIDEN ir<%0> = vp.load vp<%7>, vp<%5>
-; IF-EVL-NEXT :     WIDEN-VP vp<%8> = vp.trunc  ir<%0>, vp<%5> to i16
-; IF-EVL-NEXT :     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%6>
-; IF-EVL-NEXT :     vp<%9> = vector-pointer ir<%arrayidx4>
-; IF-EVL-NEXT :     WIDEN vp.store vp<%9>, vp<%8>, vp<%5>
-; IF-EVL-NEXT :     SCALAR-CAST vp<%10> = zext vp<%5> to i64
-; IF-EVL-NEXT :     EMIT vp<%11> = add vp<%10>, vp<%4>
-; IF-EVL-NEXT :     EMIT vp<%12> = add vp<%3>, vp<%0>
-; IF-EVL-NEXT :     EMIT branch-on-count vp<%12>, vp<%1>
-; IF-EVL-NEXT :   No successors
-; IF-EVL-NEXT : }
-
-; NO-VP: Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF>=1' {
-; NO-VP-NEXT: Live-in vp<%0> = VF * UF
-; NO-VP-NEXT: Live-in vp<%1> = vector-trip-count
-; NO-VP-NEXT: Live-in ir<%N> = original trip-count
-
-; NO-VP: vector.ph:
-; NO-VP-NEXT: Successor(s): vector loop
-
-; NO-VP: <x1> vector loop: {
-; NO-VP:   vector.body:
-; NO-VP-NEXT:     EMIT vp<%2> = CANONICAL-INDUCTION ir<0>, vp<%6>
-; NO-VP-NEXT:     vp<%3> = SCALAR-STEPS vp<%2>, ir<1>
-; NO-VP-NEXT:     CLONE ir<%arrayidx> = getelementptr inbounds ir<%b>, vp<%3>
-; NO-VP-NEXT:     vp<%4> = vector-pointer ir<%arrayidx>
-; NO-VP-NEXT:     WIDEN ir<%0> = load vp<%4>
-; NO-VP-NEXT:     WIDEN-CAST ir<%conv2> = trunc  ir<%0> to i16
-; NO-VP-NEXT:     CLONE ir<%arrayidx4> = getelementptr inbounds ir<%a>, vp<%3>
-; NO-VP-NEXT:     vp<%5> = vector-pointer ir<%arrayidx4>
-; NO-VP-NEXT:     WIDEN store vp<%5>, ir<%conv2>
-; NO-VP-NEXT:     EMIT vp<%6> = add nuw vp<%2>, vp<%0>
-; NO-VP-NEXT:     EMIT branch-on-count vp<%6>, vp<%1>
-; NO-VP-NEXT:   No successors
-; NO-VP-NEXT: }
+define void @vp_fptrunct(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPTRUNC:%.+]]> = fptrunc ir<[[LD1]]> to float
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
 
 entry:
-  %cmp8 = icmp sgt i64 %N, 0
-  br i1 %cmp8, label %for.body, label %for.cond.cleanup
+  br label %loop
 
-for.cond.cleanup:
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds double, ptr %b, i64 %iv
+  %0 = load double, ptr %gep, align 8
+  %conv2 = fptrunc double %0 to float
+  %gep4 = getelementptr inbounds float, ptr %a, i64 %iv
+  store float %conv2, ptr %gep4, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
   ret void
+}
 
-for.body:
-  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
-  %arrayidx = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
-  %0 = load i32, ptr %arrayidx, align 4
-  %conv2 = trunc i32 %0 to i16
-  %arrayidx4 = getelementptr inbounds i16, ptr %a, i64 %indvars.iv
-  store i16 %conv2, ptr %arrayidx4, align 2
-  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
-  %exitcond.not = icmp eq i64 %indvars.iv.next, %N
-  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+define void @vp_fptosi(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPTOSI:%.+]]> = fptoui ir<[[LD1]]> to i32
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTOSI]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds float, ptr %b, i64 %iv
+  %0 = load float, ptr %gep, align 4
+  %conv2 = fptoui float %0 to i32
+  %gep4 = getelementptr inbounds i32, ptr %a, i64 %iv
+  store i32 %conv2, ptr %gep4, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_inttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[SITOFP:%.+]]> = sitofp ir<[[LD1]]> to float
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv2 = sitofp i32 %0 to float
+  %gep4 = getelementptr inbounds float, ptr %a, i64 %iv
+  store float %conv2, ptr %gep4, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
 }
+
+define void @vp_uinttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-CAST ir<[[UITOFP:%.+]]> = uitofp ir<[[LD1]]> to float
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[UITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv2 = uitofp i32 %0 to float
+  %gep4 = getelementptr inbounds float, ptr %a, i64 %iv
+  store float %conv2, ptr %gep4, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
\ No newline at end of file

>From 3f4e0e56aa3890d8516cdd2a2c7a7a3fd5001c26 Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Fri, 18 Oct 2024 17:20:53 +0800
Subject: [PATCH 3/9] [VPlan] Use VPWidenIntrinsicRecipe to vp.cast

---
 llvm/include/llvm/IR/VectorBuilder.h          |  4 +-
 llvm/lib/IR/VectorBuilder.cpp                 |  9 ++--
 llvm/lib/Transforms/Utils/LoopUtils.cpp       |  4 +-
 llvm/lib/Transforms/Vectorize/VPlan.h         |  9 ++++
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp | 49 ++++++++++++-------
 .../Transforms/Vectorize/VPlanTransforms.cpp  | 18 +++++++
 .../LoopVectorize/RISCV/inloop-reduction.ll   |  2 +-
 .../RISCV/vplan-vp-cast-intrinsics.ll         | 32 ++++++------
 8 files changed, 83 insertions(+), 44 deletions(-)

diff --git a/llvm/include/llvm/IR/VectorBuilder.h b/llvm/include/llvm/IR/VectorBuilder.h
index b0277c2b52595e..830163984e37ba 100644
--- a/llvm/include/llvm/IR/VectorBuilder.h
+++ b/llvm/include/llvm/IR/VectorBuilder.h
@@ -99,11 +99,11 @@ class VectorBuilder {
                                  const Twine &Name = Twine());
 
   /// Emit a VP reduction intrinsic call for recurrence kind.
-  /// \param RdxID       The intrinsic ID of llvm.vector.reduce.*
+  /// \param ID          The intrinsic ID of call Intrinsic
   /// \param ValTy       The type of operand which the reduction operation is
   ///                    performed.
   /// \param VecOpArray  The operand list.
-  Value *createSimpleReduction(Intrinsic::ID RdxID, Type *ValTy,
+  Value *createSimpleIntrinsic(Intrinsic::ID ID, Type *ValTy,
                                ArrayRef<Value *> VecOpArray,
                                const Twine &Name = Twine());
 };
diff --git a/llvm/lib/IR/VectorBuilder.cpp b/llvm/lib/IR/VectorBuilder.cpp
index 737f49b1334d76..d629a2fb6af7b3 100644
--- a/llvm/lib/IR/VectorBuilder.cpp
+++ b/llvm/lib/IR/VectorBuilder.cpp
@@ -60,13 +60,12 @@ Value *VectorBuilder::createVectorInstruction(unsigned Opcode, Type *ReturnTy,
   return createVectorInstructionImpl(VPID, ReturnTy, InstOpArray, Name);
 }
 
-Value *VectorBuilder::createSimpleReduction(Intrinsic::ID RdxID,
-                                            Type *ValTy,
+Value *VectorBuilder::createSimpleIntrinsic(Intrinsic::ID ID, Type *ValTy,
                                             ArrayRef<Value *> InstOpArray,
                                             const Twine &Name) {
-  auto VPID = VPIntrinsic::getForIntrinsic(RdxID);
-  assert(VPReductionIntrinsic::isVPReduction(VPID) &&
-         "No VPIntrinsic for this reduction");
+  auto VPID = VPIntrinsic::getForIntrinsic(ID);
+  assert(VPIntrinsic::isVPIntrinsic(VPID) &&
+         "No VPIntrinsic for this Intrinsic");
   return createVectorInstructionImpl(VPID, ValTy, InstOpArray, Name);
 }
 
diff --git a/llvm/lib/Transforms/Utils/LoopUtils.cpp b/llvm/lib/Transforms/Utils/LoopUtils.cpp
index 70047273c3b9af..2dac2d43f7f3a3 100644
--- a/llvm/lib/Transforms/Utils/LoopUtils.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUtils.cpp
@@ -1300,7 +1300,7 @@ Value *llvm::createSimpleReduction(VectorBuilder &VBuilder, Value *Src,
   Type *SrcEltTy = SrcTy->getElementType();
   Value *Iden = getRecurrenceIdentity(Kind, SrcEltTy, Desc.getFastMathFlags());
   Value *Ops[] = {Iden, Src};
-  return VBuilder.createSimpleReduction(Id, SrcTy, Ops);
+  return VBuilder.createSimpleIntrinsic(Id, SrcTy, Ops);
 }
 
 Value *llvm::createReduction(IRBuilderBase &B,
@@ -1343,7 +1343,7 @@ Value *llvm::createOrderedReduction(VectorBuilder &VBuilder,
   Intrinsic::ID Id = getReductionIntrinsicID(RecurKind::FAdd);
   auto *SrcTy = cast<VectorType>(Src->getType());
   Value *Ops[] = {Start, Src};
-  return VBuilder.createSimpleReduction(Id, SrcTy, Ops);
+  return VBuilder.createSimpleIntrinsic(Id, SrcTy, Ops);
 }
 
 void llvm::propagateIRFlags(Value *I, ArrayRef<Value *> VL, Value *OpValue,
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 1b1630ebc6c23a..136d9b11907598 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1659,6 +1659,15 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
         MayWriteToMemory(CI.mayWriteToMemory()),
         MayHaveSideEffects(CI.mayHaveSideEffects()) {}
 
+  // VPWidenIntrinsicRecipe(CastInst &CI, Intrinsic::ID VectorIntrinsicID,
+  //                        ArrayRef<VPValue *> CallArguments, Type *Ty,
+  //                        DebugLoc DL = {})
+  //     : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, CI),
+  //       VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty),
+  //       MayReadFromMemory(CI.mayReadFromMemory()),
+  //       MayWriteToMemory(CI.mayWriteToMemory()),
+  //       MayHaveSideEffects(CI.mayHaveSideEffects()) {}
+
   VPWidenIntrinsicRecipe(Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
                          DebugLoc DL = {})
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 3b11ec77813f3e..04fc61241eb538 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -960,24 +960,37 @@ void VPWidenIntrinsicRecipe::execute(VPTransformState &State) {
     Args.push_back(Arg);
   }
 
-  // Use vector version of the intrinsic.
-  Module *M = State.Builder.GetInsertBlock()->getModule();
-  Function *VectorF =
-      Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
-  assert(VectorF && "Can't retrieve vector intrinsic.");
-
-  auto *CI = cast_or_null<CallInst>(getUnderlyingValue());
-  SmallVector<OperandBundleDef, 1> OpBundles;
-  if (CI)
-    CI->getOperandBundlesAsDefs(OpBundles);
-
-  CallInst *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
-
-  setFlags(V);
-
-  if (!V->getType()->isVoidTy())
-    State.set(this, V);
-  State.addMetadata(V, CI);
+  if (VPIntrinsic::isVPIntrinsic(VectorIntrinsicID)) {
+    // Use vector version of the vector predicate Intrinsic
+    IRBuilderBase &BuilderIR = State.Builder;
+    VectorBuilder VBuilder(BuilderIR);
+    Value *Mask = BuilderIR.CreateVectorSplat(State.VF, BuilderIR.getTrue());
+    VBuilder.setMask(Mask).setEVL(Args.back());
+    Args.pop_back();
+    Value *VPInst = VBuilder.createSimpleIntrinsic(
+        VectorIntrinsicID, TysForDecl[0], Args, "vp.call");
+    if (!VPInst->getType()->isVoidTy())
+      State.set(this, VPInst);
+    State.addMetadata(VPInst,
+                      dyn_cast_or_null<Instruction>(getUnderlyingValue()));
+  } else {
+    // Use vector version of the intrinsic.
+    Module *M = State.Builder.GetInsertBlock()->getModule();
+    Function *VectorF =
+        Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
+    assert(VectorF && "Can't retrieve vector intrinsic.");
+
+    auto *CI = cast_or_null<CallInst>(getUnderlyingValue());
+    SmallVector<OperandBundleDef, 1> OpBundles;
+    if (CI)
+      CI->getOperandBundlesAsDefs(OpBundles);
+
+    CallInst *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
+    setFlags(V);
+    if (!V->getType()->isVoidTy())
+      State.set(this, V);
+    State.addMetadata(V, CI);
+  }
 }
 
 InstructionCost VPWidenIntrinsicRecipe::computeCost(ElementCount VF,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index ba791c679b4ae4..7d4e64b05aa7c4 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1501,6 +1501,24 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                         *CI, VPID, Ops, TypeInfo.inferScalarType(CInst),
                         CInst->getDebugLoc());
                   })
+              .Case<VPWidenCastRecipe>(
+                  [&](VPWidenCastRecipe *CInst) -> VPRecipeBase * {
+                    auto *CI = cast<CastInst>(CInst->getUnderlyingInstr());
+                    SmallVector<VPValue *> Ops(CInst->operands());
+                    Ops.push_back(&EVL);
+                    Intrinsic::ID VPID =
+                        VPIntrinsic::getForOpcode(CI->getOpcode());
+                    if (VPID == Intrinsic::not_intrinsic)
+                      return nullptr;
+                    // FIXME: In fact, can we really not pass the
+                    // underlyingInstr? In this case, how to set the Flag and
+                    // add metadata in VPWidenIntrinsicRecipe::execute?
+                    // return new VPWidenIntrinsicRecipe(
+                    //     *CI, VPID, Ops, CI->getType(), CI->getDebugLoc());
+                    return new VPWidenIntrinsicRecipe(
+                        VPID, Ops, TypeInfo.inferScalarType(CInst), false,
+                        false, false);
+                  })
               .Case<VPWidenSelectRecipe>([&](VPWidenSelectRecipe *Sel) {
                 SmallVector<VPValue *> Ops(Sel->operands());
                 Ops.push_back(&EVL);
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
index d273a88a84feaa..9ee1708f175ab0 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
@@ -200,7 +200,7 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP7]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
 ; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP9]], <vscale x 8 x i1> 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), i32 [[TMP6]])
-; IF-EVL-INLOOP-NEXT:    [[VP_CAST:%.*]] = sext <vscale x 8 x i16> [[VP_OP_LOAD]] to <vscale x 8 x i32>
+; IF-EVL-INLOOP-NEXT:    [[VP_CAST:%.*]] = call <vscale x 8 x i32> @llvm.vp.sext.nxv8i32.nxv8i16(<vscale x 8 x i16> [[VP_OP_LOAD]], <vscale x 8 x i1> 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), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[VP_CAST]], <vscale x 8 x i1> 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), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP11]] = add i32 [[TMP10]], [[VEC_PHI]]
 ; IF-EVL-INLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP6]], [[EVL_BASED_IV]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
index 6894a30f30efe4..8692f6c802e559 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -23,10 +23,10 @@ define void @vp_sext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[SEXT:%.+]]> = sext  ir<[[LD1]]> to i64
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SEXT:%.+]]> = call llvm.vp.sext(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -73,10 +73,10 @@ define void @vp_zext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[ZEXT:%.+]]> = zext ir<[[LD1]]> to i64
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[ZEXT:%.+]]> = call llvm.vp.zext(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[ZEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[ZEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -121,10 +121,10 @@ define void @vp_truncate(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[TRUNC:%.+]]> = trunc ir<[[LD1]]> to i16
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[TRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[TRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -169,10 +169,10 @@ define void @vp_fpext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPEXT:%.+]]> = fpext ir<[[LD1]]> to 
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -217,10 +217,10 @@ define void @vp_fptrunct(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPTRUNC:%.+]]> = fptrunc ir<[[LD1]]> to float
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTRUNC:%.+]]> = call llvm.vp.fptrunc(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -265,10 +265,10 @@ define void @vp_fptosi(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPTOSI:%.+]]> = fptoui ir<[[LD1]]> to i32
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTOSI]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTOSI]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -313,10 +313,10 @@ define void @vp_inttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[SITOFP:%.+]]> = sitofp ir<[[LD1]]> to float
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -361,10 +361,10 @@ define void @vp_uinttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[UITOFP:%.+]]> = uitofp ir<[[LD1]]> to float
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[UITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[UITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>

>From cc04719ce8b34060d2e6a1af112ec982300ecee1 Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Thu, 31 Oct 2024 18:37:56 +0800
Subject: [PATCH 4/9] fix the comments

---
 llvm/lib/Transforms/Vectorize/VPlan.h         | 16 +++++-----
 .../Transforms/Vectorize/VPlanTransforms.cpp  |  8 +----
 .../RISCV/vplan-vp-cast-intrinsics.ll         | 32 +++++++++----------
 3 files changed, 25 insertions(+), 31 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 136d9b11907598..e84c3fb0a43af7 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1659,14 +1659,14 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
         MayWriteToMemory(CI.mayWriteToMemory()),
         MayHaveSideEffects(CI.mayHaveSideEffects()) {}
 
-  // VPWidenIntrinsicRecipe(CastInst &CI, Intrinsic::ID VectorIntrinsicID,
-  //                        ArrayRef<VPValue *> CallArguments, Type *Ty,
-  //                        DebugLoc DL = {})
-  //     : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, CI),
-  //       VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty),
-  //       MayReadFromMemory(CI.mayReadFromMemory()),
-  //       MayWriteToMemory(CI.mayWriteToMemory()),
-  //       MayHaveSideEffects(CI.mayHaveSideEffects()) {}
+  VPWidenIntrinsicRecipe(CastInst &CI, Intrinsic::ID VectorIntrinsicID,
+                         ArrayRef<VPValue *> CallArguments, Type *Ty,
+                         DebugLoc DL = {})
+      : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, CI),
+        VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty),
+        MayReadFromMemory(CI.mayReadFromMemory()),
+        MayWriteToMemory(CI.mayWriteToMemory()),
+        MayHaveSideEffects(CI.mayHaveSideEffects()) {}
 
   VPWidenIntrinsicRecipe(Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 7d4e64b05aa7c4..a1ec7f6fc248f8 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1510,14 +1510,8 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                         VPIntrinsic::getForOpcode(CI->getOpcode());
                     if (VPID == Intrinsic::not_intrinsic)
                       return nullptr;
-                    // FIXME: In fact, can we really not pass the
-                    // underlyingInstr? In this case, how to set the Flag and
-                    // add metadata in VPWidenIntrinsicRecipe::execute?
-                    // return new VPWidenIntrinsicRecipe(
-                    //     *CI, VPID, Ops, CI->getType(), CI->getDebugLoc());
                     return new VPWidenIntrinsicRecipe(
-                        VPID, Ops, TypeInfo.inferScalarType(CInst), false,
-                        false, false);
+                        *CI, VPID, Ops, CI->getType(), CI->getDebugLoc());
                   })
               .Case<VPWidenSelectRecipe>([&](VPWidenSelectRecipe *Sel) {
                 SmallVector<VPValue *> Ops(Sel->operands());
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
index 8692f6c802e559..0d517407b5326d 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -23,10 +23,10 @@ define void @vp_sext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SEXT:%.+]]> = call llvm.vp.sext(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SEXT:%.+]]> = call llvm.vp.sext(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -73,10 +73,10 @@ define void @vp_zext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[ZEXT:%.+]]> = call llvm.vp.zext(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[ZEXT:%.+]]> = call llvm.vp.zext(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[ZEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[ZEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -121,10 +121,10 @@ define void @vp_truncate(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[TRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[TRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -169,10 +169,10 @@ define void @vp_fpext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -217,10 +217,10 @@ define void @vp_fptrunct(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTRUNC:%.+]]> = call llvm.vp.fptrunc(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTRUNC:%.+]]> = call llvm.vp.fptrunc(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -265,10 +265,10 @@ define void @vp_fptosi(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTOSI]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTOSI]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -313,10 +313,10 @@ define void @vp_inttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -361,10 +361,10 @@ define void @vp_uinttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[UITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[UITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>

>From 363d625238acff5a70852c60a69a76226de1c57b Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Wed, 6 Nov 2024 14:05:52 +0800
Subject: [PATCH 5/9] fix the comments

---
 llvm/include/llvm/IR/VectorBuilder.h          |  4 +-
 llvm/lib/IR/VectorBuilder.cpp                 |  9 ++--
 llvm/lib/Transforms/Utils/LoopUtils.cpp       |  4 +-
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp | 49 +++++++------------
 .../Transforms/Vectorize/VPlanTransforms.cpp  | 11 ++++-
 5 files changed, 36 insertions(+), 41 deletions(-)

diff --git a/llvm/include/llvm/IR/VectorBuilder.h b/llvm/include/llvm/IR/VectorBuilder.h
index 830163984e37ba..b0277c2b52595e 100644
--- a/llvm/include/llvm/IR/VectorBuilder.h
+++ b/llvm/include/llvm/IR/VectorBuilder.h
@@ -99,11 +99,11 @@ class VectorBuilder {
                                  const Twine &Name = Twine());
 
   /// Emit a VP reduction intrinsic call for recurrence kind.
-  /// \param ID          The intrinsic ID of call Intrinsic
+  /// \param RdxID       The intrinsic ID of llvm.vector.reduce.*
   /// \param ValTy       The type of operand which the reduction operation is
   ///                    performed.
   /// \param VecOpArray  The operand list.
-  Value *createSimpleIntrinsic(Intrinsic::ID ID, Type *ValTy,
+  Value *createSimpleReduction(Intrinsic::ID RdxID, Type *ValTy,
                                ArrayRef<Value *> VecOpArray,
                                const Twine &Name = Twine());
 };
diff --git a/llvm/lib/IR/VectorBuilder.cpp b/llvm/lib/IR/VectorBuilder.cpp
index d629a2fb6af7b3..737f49b1334d76 100644
--- a/llvm/lib/IR/VectorBuilder.cpp
+++ b/llvm/lib/IR/VectorBuilder.cpp
@@ -60,12 +60,13 @@ Value *VectorBuilder::createVectorInstruction(unsigned Opcode, Type *ReturnTy,
   return createVectorInstructionImpl(VPID, ReturnTy, InstOpArray, Name);
 }
 
-Value *VectorBuilder::createSimpleIntrinsic(Intrinsic::ID ID, Type *ValTy,
+Value *VectorBuilder::createSimpleReduction(Intrinsic::ID RdxID,
+                                            Type *ValTy,
                                             ArrayRef<Value *> InstOpArray,
                                             const Twine &Name) {
-  auto VPID = VPIntrinsic::getForIntrinsic(ID);
-  assert(VPIntrinsic::isVPIntrinsic(VPID) &&
-         "No VPIntrinsic for this Intrinsic");
+  auto VPID = VPIntrinsic::getForIntrinsic(RdxID);
+  assert(VPReductionIntrinsic::isVPReduction(VPID) &&
+         "No VPIntrinsic for this reduction");
   return createVectorInstructionImpl(VPID, ValTy, InstOpArray, Name);
 }
 
diff --git a/llvm/lib/Transforms/Utils/LoopUtils.cpp b/llvm/lib/Transforms/Utils/LoopUtils.cpp
index 2dac2d43f7f3a3..70047273c3b9af 100644
--- a/llvm/lib/Transforms/Utils/LoopUtils.cpp
+++ b/llvm/lib/Transforms/Utils/LoopUtils.cpp
@@ -1300,7 +1300,7 @@ Value *llvm::createSimpleReduction(VectorBuilder &VBuilder, Value *Src,
   Type *SrcEltTy = SrcTy->getElementType();
   Value *Iden = getRecurrenceIdentity(Kind, SrcEltTy, Desc.getFastMathFlags());
   Value *Ops[] = {Iden, Src};
-  return VBuilder.createSimpleIntrinsic(Id, SrcTy, Ops);
+  return VBuilder.createSimpleReduction(Id, SrcTy, Ops);
 }
 
 Value *llvm::createReduction(IRBuilderBase &B,
@@ -1343,7 +1343,7 @@ Value *llvm::createOrderedReduction(VectorBuilder &VBuilder,
   Intrinsic::ID Id = getReductionIntrinsicID(RecurKind::FAdd);
   auto *SrcTy = cast<VectorType>(Src->getType());
   Value *Ops[] = {Start, Src};
-  return VBuilder.createSimpleIntrinsic(Id, SrcTy, Ops);
+  return VBuilder.createSimpleReduction(Id, SrcTy, Ops);
 }
 
 void llvm::propagateIRFlags(Value *I, ArrayRef<Value *> VL, Value *OpValue,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 04fc61241eb538..d1922b4e0755ef 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -960,37 +960,24 @@ void VPWidenIntrinsicRecipe::execute(VPTransformState &State) {
     Args.push_back(Arg);
   }
 
-  if (VPIntrinsic::isVPIntrinsic(VectorIntrinsicID)) {
-    // Use vector version of the vector predicate Intrinsic
-    IRBuilderBase &BuilderIR = State.Builder;
-    VectorBuilder VBuilder(BuilderIR);
-    Value *Mask = BuilderIR.CreateVectorSplat(State.VF, BuilderIR.getTrue());
-    VBuilder.setMask(Mask).setEVL(Args.back());
-    Args.pop_back();
-    Value *VPInst = VBuilder.createSimpleIntrinsic(
-        VectorIntrinsicID, TysForDecl[0], Args, "vp.call");
-    if (!VPInst->getType()->isVoidTy())
-      State.set(this, VPInst);
-    State.addMetadata(VPInst,
-                      dyn_cast_or_null<Instruction>(getUnderlyingValue()));
-  } else {
-    // Use vector version of the intrinsic.
-    Module *M = State.Builder.GetInsertBlock()->getModule();
-    Function *VectorF =
-        Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
-    assert(VectorF && "Can't retrieve vector intrinsic.");
-
-    auto *CI = cast_or_null<CallInst>(getUnderlyingValue());
-    SmallVector<OperandBundleDef, 1> OpBundles;
-    if (CI)
-      CI->getOperandBundlesAsDefs(OpBundles);
-
-    CallInst *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
-    setFlags(V);
-    if (!V->getType()->isVoidTy())
-      State.set(this, V);
-    State.addMetadata(V, CI);
-  }
+  // Use vector version of the intrinsic.
+  Module *M = State.Builder.GetInsertBlock()->getModule();
+  Function *VectorF =
+      Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
+  assert(VectorF && "Can't retrieve vector intrinsic.");
+
+  auto *CI = cast_or_null<CallInst>(getUnderlyingValue());
+  SmallVector<OperandBundleDef, 1> OpBundles;
+  if (CI)
+    CI->getOperandBundlesAsDefs(OpBundles);
+
+  CallInst *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
+
+  setFlags(V);
+
+  if (!V->getType()->isVoidTy())
+    State.set(this, V);
+  State.addMetadata(V, dyn_cast_or_null<Instruction>(getUnderlyingValue()));
 }
 
 InstructionCost VPWidenIntrinsicRecipe::computeCost(ElementCount VF,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index a1ec7f6fc248f8..56e5c67590f1db 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1504,12 +1504,19 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
               .Case<VPWidenCastRecipe>(
                   [&](VPWidenCastRecipe *CInst) -> VPRecipeBase * {
                     auto *CI = cast<CastInst>(CInst->getUnderlyingInstr());
-                    SmallVector<VPValue *> Ops(CInst->operands());
-                    Ops.push_back(&EVL);
                     Intrinsic::ID VPID =
                         VPIntrinsic::getForOpcode(CI->getOpcode());
                     if (VPID == Intrinsic::not_intrinsic)
                       return nullptr;
+                    SmallVector<VPValue *> Ops(CInst->operands());
+                    if (VPIntrinsic::getMaskParamPos(VPID)) {
+                      VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
+                          IntegerType::getInt1Ty(CI->getContext())));
+                      Ops.push_back(Mask);
+                    }
+                    if (VPIntrinsic::getVectorLengthParamPos(VPID)) {
+                      Ops.push_back(&EVL);
+                    }
                     return new VPWidenIntrinsicRecipe(
                         *CI, VPID, Ops, CI->getType(), CI->getDebugLoc());
                   })

>From 0e45dc495417e3c53a62d6e27b90dac981a9ad2b Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Thu, 14 Nov 2024 17:24:01 +0800
Subject: [PATCH 6/9] fix the commets

---
 llvm/lib/Transforms/Vectorize/VPlan.h         | 19 +++--------
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp | 25 +++++++++++----
 .../Transforms/Vectorize/VPlanTransforms.cpp  | 21 ++++++------
 .../RISCV/vplan-vp-cast-intrinsics.ll         | 32 +++++++++----------
 4 files changed, 51 insertions(+), 46 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index e84c3fb0a43af7..65863fbd276dc4 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1650,23 +1650,14 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
   bool MayHaveSideEffects;
 
 public:
-  VPWidenIntrinsicRecipe(CallInst &CI, Intrinsic::ID VectorIntrinsicID,
+  VPWidenIntrinsicRecipe(Instruction &I, Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
                          DebugLoc DL = {})
-      : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, CI),
+      : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, I),
         VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty),
-        MayReadFromMemory(CI.mayReadFromMemory()),
-        MayWriteToMemory(CI.mayWriteToMemory()),
-        MayHaveSideEffects(CI.mayHaveSideEffects()) {}
-
-  VPWidenIntrinsicRecipe(CastInst &CI, Intrinsic::ID VectorIntrinsicID,
-                         ArrayRef<VPValue *> CallArguments, Type *Ty,
-                         DebugLoc DL = {})
-      : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, CI),
-        VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty),
-        MayReadFromMemory(CI.mayReadFromMemory()),
-        MayWriteToMemory(CI.mayWriteToMemory()),
-        MayHaveSideEffects(CI.mayHaveSideEffects()) {}
+        MayReadFromMemory(I.mayReadFromMemory()),
+        MayWriteToMemory(I.mayWriteToMemory()),
+        MayHaveSideEffects(I.mayHaveSideEffects()) {}
 
   VPWidenIntrinsicRecipe(Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index d1922b4e0755ef..312590093f97fa 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -962,18 +962,31 @@ void VPWidenIntrinsicRecipe::execute(VPTransformState &State) {
 
   // Use vector version of the intrinsic.
   Module *M = State.Builder.GetInsertBlock()->getModule();
+  bool IsVPIntrinsic = VPIntrinsic::isVPIntrinsic(VectorIntrinsicID);
   Function *VectorF =
-      Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
+      IsVPIntrinsic
+          ? VPIntrinsic::getOrInsertDeclarationForParams(M, VectorIntrinsicID,
+                                                         TysForDecl[0], Args)
+          : Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
   assert(VectorF && "Can't retrieve vector intrinsic.");
 
-  auto *CI = cast_or_null<CallInst>(getUnderlyingValue());
   SmallVector<OperandBundleDef, 1> OpBundles;
-  if (CI)
-    CI->getOperandBundlesAsDefs(OpBundles);
+  if (!IsVPIntrinsic) {
+    if (auto *CI = cast_or_null<CallInst>(getUnderlyingValue()))
+      CI->getOperandBundlesAsDefs(OpBundles);
+  }
 
-  CallInst *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
+  Value *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
 
-  setFlags(V);
+  if (IsVPIntrinsic) {
+    // Currently vp-intrinsics only accept FMF flags.
+    // TODO: Enable other flags when support is added.
+    // vp_uitofp will get OperationType::NonNegOp
+    if (isa<FPMathOperator>(V) && VectorIntrinsicID != Intrinsic::vp_uitofp)
+      setFlags(cast<Instruction>(V));
+  } else {
+    setFlags(cast<Instruction>(V));
+  }
 
   if (!V->getType()->isVoidTy())
     State.set(this, V);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 56e5c67590f1db..aa9b0ec35546ff 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1503,22 +1503,23 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                   })
               .Case<VPWidenCastRecipe>(
                   [&](VPWidenCastRecipe *CInst) -> VPRecipeBase * {
-                    auto *CI = cast<CastInst>(CInst->getUnderlyingInstr());
+                    auto *CI = dyn_cast<CastInst>(CInst->getUnderlyingInstr());
                     Intrinsic::ID VPID =
                         VPIntrinsic::getForOpcode(CI->getOpcode());
                     if (VPID == Intrinsic::not_intrinsic)
                       return nullptr;
+
                     SmallVector<VPValue *> Ops(CInst->operands());
-                    if (VPIntrinsic::getMaskParamPos(VPID)) {
-                      VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
-                          IntegerType::getInt1Ty(CI->getContext())));
-                      Ops.push_back(Mask);
-                    }
-                    if (VPIntrinsic::getVectorLengthParamPos(VPID)) {
-                      Ops.push_back(&EVL);
-                    }
+                    assert(VPIntrinsic::getMaskParamPos(VPID) &&
+                           VPIntrinsic::getVectorLengthParamPos(VPID) &&
+                           "Expected VP intrinsic");
+                    VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
+                        IntegerType::getInt1Ty(CI->getContext())));
+                    Ops.push_back(Mask);
+                    Ops.push_back(&EVL);
                     return new VPWidenIntrinsicRecipe(
-                        *CI, VPID, Ops, CI->getType(), CI->getDebugLoc());
+                        *CI, VPID, Ops, TypeInfo.inferScalarType(CInst),
+                        CInst->getDebugLoc());
                   })
               .Case<VPWidenSelectRecipe>([&](VPWidenSelectRecipe *Sel) {
                 SmallVector<VPValue *> Ops(Sel->operands());
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
index 0d517407b5326d..e9e5f2b2c4d2f2 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -4,7 +4,7 @@
 ; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s
 
-define void @vp_sext(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_sext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -23,7 +23,7 @@ define void @vp_sext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SEXT:%.+]]> = call llvm.vp.sext(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SEXT:%.+]]> = call llvm.vp.sext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SEXT]]>, vp<[[EVL]]>
@@ -54,7 +54,7 @@ exit:
   ret void
 }
 
-define void @vp_zext(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_zext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -73,7 +73,7 @@ define void @vp_zext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[ZEXT:%.+]]> = call llvm.vp.zext(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[ZEXT:%.+]]> = call llvm.vp.zext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[ZEXT]]>, vp<[[EVL]]>
@@ -102,7 +102,7 @@ exit:
   ret void
 }
 
-define void @vp_truncate(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_truncate(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Initial VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -121,7 +121,7 @@ define void @vp_truncate(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[TRUNC]]>, vp<[[EVL]]>
@@ -150,7 +150,7 @@ exit:
   ret void
 }
 
-define void @vp_fpext(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_fpext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -169,7 +169,7 @@ define void @vp_fpext(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPEXT]]>, vp<[[EVL]]>
@@ -198,7 +198,7 @@ exit:
   ret void
 }
 
-define void @vp_fptrunct(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_fptrunct(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -217,7 +217,7 @@ define void @vp_fptrunct(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTRUNC:%.+]]> = call llvm.vp.fptrunc(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTRUNC:%.+]]> = call llvm.vp.fptrunc(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTRUNC]]>, vp<[[EVL]]>
@@ -246,7 +246,7 @@ exit:
   ret void
 }
 
-define void @vp_fptosi(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -265,7 +265,7 @@ define void @vp_fptosi(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTOSI]]>, vp<[[EVL]]>
@@ -294,7 +294,7 @@ exit:
   ret void
 }
 
-define void @vp_inttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_inttofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -313,7 +313,7 @@ define void @vp_inttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SITOFP]]>, vp<[[EVL]]>
@@ -342,7 +342,7 @@ exit:
   ret void
 }
 
-define void @vp_uinttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
+define void @vp_uinttofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -361,7 +361,7 @@ define void @vp_uinttofp(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
 ; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[UITOFP]]>, vp<[[EVL]]>

>From 1d8f27cdb840238130328bde2fcfa325aaf6ed8a Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Sun, 17 Nov 2024 15:12:05 +0800
Subject: [PATCH 7/9] fix the comments

---
 llvm/lib/Transforms/Vectorize/VPlan.h         | 10 +++---
 .../Transforms/Vectorize/VPlanTransforms.cpp  |  2 +-
 .../LoopVectorize/RISCV/inloop-reduction.ll   |  8 ++---
 .../RISCV/vplan-vp-cast-intrinsics.ll         | 34 +++++++++----------
 4 files changed, 27 insertions(+), 27 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 65863fbd276dc4..1b1630ebc6c23a 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1650,14 +1650,14 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
   bool MayHaveSideEffects;
 
 public:
-  VPWidenIntrinsicRecipe(Instruction &I, Intrinsic::ID VectorIntrinsicID,
+  VPWidenIntrinsicRecipe(CallInst &CI, Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
                          DebugLoc DL = {})
-      : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, I),
+      : VPRecipeWithIRFlags(VPDef::VPWidenIntrinsicSC, CallArguments, CI),
         VectorIntrinsicID(VectorIntrinsicID), ResultTy(Ty),
-        MayReadFromMemory(I.mayReadFromMemory()),
-        MayWriteToMemory(I.mayWriteToMemory()),
-        MayHaveSideEffects(I.mayHaveSideEffects()) {}
+        MayReadFromMemory(CI.mayReadFromMemory()),
+        MayWriteToMemory(CI.mayWriteToMemory()),
+        MayHaveSideEffects(CI.mayHaveSideEffects()) {}
 
   VPWidenIntrinsicRecipe(Intrinsic::ID VectorIntrinsicID,
                          ArrayRef<VPValue *> CallArguments, Type *Ty,
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index aa9b0ec35546ff..4d8dc44cf8a67e 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1518,7 +1518,7 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                     Ops.push_back(Mask);
                     Ops.push_back(&EVL);
                     return new VPWidenIntrinsicRecipe(
-                        *CI, VPID, Ops, TypeInfo.inferScalarType(CInst),
+                        VPID, Ops, TypeInfo.inferScalarType(CInst),
                         CInst->getDebugLoc());
                   })
               .Case<VPWidenSelectRecipe>([&](VPWidenSelectRecipe *Sel) {
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
index 9ee1708f175ab0..85f3b8066fc3df 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
@@ -142,10 +142,10 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP6:%.*]] = add i32 [[EVL_BASED_IV]], 0
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP6]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.vp.load.nxv4i16.p0(ptr align 2 [[TMP8]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
-; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = sext <vscale x 4 x i16> [[VP_OP_LOAD]] to <vscale x 4 x i32>
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP:%.*]] = call <vscale x 4 x i32> @llvm.vp.add.nxv4i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP9]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
-; IF-EVL-OUTLOOP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.vp.load.nxv4i16.p0(ptr align 2 [[TMP8]], <vscale x 4 x i1> 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), i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = call <vscale x 4 x i32> @llvm.vp.sext.nxv4i32.nxv4i16(<vscale x 4 x i16> [[VP_OP_LOAD]], <vscale x 4 x i1> 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), i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP:%.*]] = call <vscale x 4 x i32> @llvm.vp.add.nxv4i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP9]], <vscale x 4 x i1> 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), i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> 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), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP5]])
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP5]], [[EVL_BASED_IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP4]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
index e9e5f2b2c4d2f2..05e1040df6cd87 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -23,10 +23,10 @@ define void @vp_sext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SEXT:%.+]]> = call llvm.vp.sext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SEXT:%.+]]> = call llvm.vp.sext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -73,10 +73,10 @@ define void @vp_zext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[ZEXT:%.+]]> = call llvm.vp.zext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[ZEXT:%.+]]> = call llvm.vp.zext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[ZEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[ZEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -121,10 +121,10 @@ define void @vp_truncate(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[TRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[TRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -169,10 +169,10 @@ define void @vp_fpext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPEXT]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPEXT]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -217,10 +217,10 @@ define void @vp_fptrunct(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTRUNC:%.+]]> = call llvm.vp.fptrunc(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTRUNC:%.+]]> = call llvm.vp.fptrunc(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -265,10 +265,10 @@ define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[FPTOSI]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTOSI]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -313,10 +313,10 @@ define void @vp_inttofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[SITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -361,10 +361,10 @@ define void @vp_uinttofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[UITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[UITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -388,4 +388,4 @@ loop:
 
 exit:
   ret void
-}
\ No newline at end of file
+}

>From 5bd40b9d801c72c22330d88dde183389ad7b1679 Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Thu, 28 Nov 2024 12:18:48 +0800
Subject: [PATCH 8/9] Rebase && add ir test && address the failed tests

---
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp |   14 +-
 .../Transforms/Vectorize/VPlanTransforms.cpp  |   11 +-
 .../LoopVectorize/RISCV/inloop-reduction.ll   |   14 +-
 ...ize-force-tail-with-evl-call-intrinsics.ll |    8 +-
 ...ize-force-tail-with-evl-cast-intrinsics.ll | 1109 +++++++++++++++++
 .../RISCV/vplan-vp-call-intrinsics.ll         |   16 +-
 .../RISCV/vplan-vp-cast-intrinsics.ll         |  160 ++-
 7 files changed, 1277 insertions(+), 55 deletions(-)
 create mode 100644 llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll

diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 312590093f97fa..d9429da8dfa93b 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -968,7 +968,8 @@ void VPWidenIntrinsicRecipe::execute(VPTransformState &State) {
           ? VPIntrinsic::getOrInsertDeclarationForParams(M, VectorIntrinsicID,
                                                          TysForDecl[0], Args)
           : Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
-  assert(VectorF && "Can't retrieve vector intrinsic.");
+  assert(VectorF &&
+         "Can't retrieve vector intrinsic or vector-predication intrinsics.");
 
   SmallVector<OperandBundleDef, 1> OpBundles;
   if (!IsVPIntrinsic) {
@@ -976,16 +977,15 @@ void VPWidenIntrinsicRecipe::execute(VPTransformState &State) {
       CI->getOperandBundlesAsDefs(OpBundles);
   }
 
-  Value *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
+  Instruction *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
 
   if (IsVPIntrinsic) {
-    // Currently vp-intrinsics only accept FMF flags.
-    // TODO: Enable other flags when support is added.
-    // vp_uitofp will get OperationType::NonNegOp
+    // Currently vp-intrinsics only accept FMF flags. llvm.vp.uitofp will get
+    // Flags of OperationType::NonNegOp && OperationType::FPMathOp.
     if (isa<FPMathOperator>(V) && VectorIntrinsicID != Intrinsic::vp_uitofp)
-      setFlags(cast<Instruction>(V));
+      setFlags(V);
   } else {
-    setFlags(cast<Instruction>(V));
+    setFlags(V);
   }
 
   if (!V->getType()->isVoidTy())
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 4d8dc44cf8a67e..f349313feec6ee 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1446,7 +1446,8 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
   VPTypeAnalysis TypeInfo(CanonicalIVType);
   LLVMContext &Ctx = CanonicalIVType->getContext();
   SmallVector<VPValue *> HeaderMasks = collectAllHeaderMasks(Plan);
-
+  VPValue *AllOneMask =
+      Plan.getOrAddLiveIn(ConstantInt::getTrue(IntegerType::getInt1Ty(Ctx)));
   for (VPUser *U : Plan.getVF().users()) {
     if (auto *R = dyn_cast<VPReverseVectorPointerRecipe>(U))
       R->setOperand(1, &EVL);
@@ -1493,9 +1494,7 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                     assert(VPIntrinsic::getMaskParamPos(VPID) &&
                            VPIntrinsic::getVectorLengthParamPos(VPID) &&
                            "Expected VP intrinsic");
-                    VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
-                        IntegerType::getInt1Ty(CI->getContext())));
-                    Ops.push_back(Mask);
+                    Ops.push_back(AllOneMask);
                     Ops.push_back(&EVL);
                     return new VPWidenIntrinsicRecipe(
                         *CI, VPID, Ops, TypeInfo.inferScalarType(CInst),
@@ -1513,9 +1512,7 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                     assert(VPIntrinsic::getMaskParamPos(VPID) &&
                            VPIntrinsic::getVectorLengthParamPos(VPID) &&
                            "Expected VP intrinsic");
-                    VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
-                        IntegerType::getInt1Ty(CI->getContext())));
-                    Ops.push_back(Mask);
+                    Ops.push_back(AllOneMask);
                     Ops.push_back(&EVL);
                     return new VPWidenIntrinsicRecipe(
                         VPID, Ops, TypeInfo.inferScalarType(CInst),
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
index 85f3b8066fc3df..ec1796fae03d95 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
@@ -142,10 +142,10 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP6:%.*]] = add i32 [[EVL_BASED_IV]], 0
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP6]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.vp.load.nxv4i16.p0(ptr align 2 [[TMP8]], <vscale x 4 x i1> 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), i32 [[TMP5]])
-; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = call <vscale x 4 x i32> @llvm.vp.sext.nxv4i32.nxv4i16(<vscale x 4 x i16> [[VP_OP_LOAD]], <vscale x 4 x i1> 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), i32 [[TMP5]])
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP:%.*]] = call <vscale x 4 x i32> @llvm.vp.add.nxv4i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP9]], <vscale x 4 x i1> 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), i32 [[TMP5]])
-; IF-EVL-OUTLOOP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> 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), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.vp.load.nxv4i16.p0(ptr align 2 [[TMP8]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = call <vscale x 4 x i32> @llvm.vp.sext.nxv4i32.nxv4i16(<vscale x 4 x i16> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP:%.*]] = call <vscale x 4 x i32> @llvm.vp.add.nxv4i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP9]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP5]])
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP5]], [[EVL_BASED_IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP4]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -199,9 +199,9 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-INLOOP-NEXT:    [[TMP7:%.*]] = add i32 [[EVL_BASED_IV]], 0
 ; IF-EVL-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP7]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP9]], <vscale x 8 x i1> 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), i32 [[TMP6]])
-; IF-EVL-INLOOP-NEXT:    [[VP_CAST:%.*]] = call <vscale x 8 x i32> @llvm.vp.sext.nxv8i32.nxv8i16(<vscale x 8 x i16> [[VP_OP_LOAD]], <vscale x 8 x i1> 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), i32 [[TMP6]])
-; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[VP_CAST]], <vscale x 8 x i1> 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), i32 [[TMP6]])
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP9]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
+; IF-EVL-INLOOP-NEXT:    [[TMP14:%.*]] = call <vscale x 8 x i32> @llvm.vp.sext.nxv8i32.nxv8i16(<vscale x 8 x i16> [[VP_OP_LOAD]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
+; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[TMP14]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP11]] = add i32 [[TMP10]], [[VEC_PHI]]
 ; IF-EVL-INLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP6]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP4]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll
index 64f86c6cdeed6c..66a5e086a01171 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll
@@ -708,9 +708,9 @@ define void @vp_lrint(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP10]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
 ; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
-; IF-EVL-NEXT:    [[TMP13:%.*]] = fpext <vscale x 4 x float> [[VP_OP_LOAD]] to <vscale x 4 x double>
+; IF-EVL-NEXT:    [[TMP13:%.*]] = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call <vscale x 4 x i64> @llvm.vp.lrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
-; IF-EVL-NEXT:    [[TMP15:%.*]] = trunc <vscale x 4 x i64> [[TMP14]] to <vscale x 4 x i32>
+; IF-EVL-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.vp.trunc.nxv4i32.nxv4i64(<vscale x 4 x i64> [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP10]]
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
 ; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP15]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
@@ -817,9 +817,9 @@ define void @vp_llrint(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP10]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
 ; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
-; IF-EVL-NEXT:    [[TMP13:%.*]] = fpext <vscale x 4 x float> [[VP_OP_LOAD]] to <vscale x 4 x double>
+; IF-EVL-NEXT:    [[TMP13:%.*]] = call <vscale x 4 x double> @llvm.vp.fpext.nxv4f64.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call <vscale x 4 x i64> @llvm.vp.llrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
-; IF-EVL-NEXT:    [[TMP15:%.*]] = trunc <vscale x 4 x i64> [[TMP14]] to <vscale x 4 x i32>
+; IF-EVL-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.vp.trunc.nxv4i32.nxv4i64(<vscale x 4 x i64> [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP10]]
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
 ; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP15]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll
new file mode 100644
index 00000000000000..4557e95f1e1b6a
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll
@@ -0,0 +1,1109 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -passes=loop-vectorize \
+; RUN: -force-tail-folding-style=data-with-evl \
+; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL
+
+; RUN: opt -passes=loop-vectorize \
+; RUN: -force-tail-folding-style=none \
+; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP
+
+define void @vp_sext(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_sext(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 18, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = shl i64 [[N]], 3
+; IF-EVL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]]
+; IF-EVL-NEXT:    [[TMP6:%.*]] = shl i64 [[N]], 2
+; IF-EVL-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP6]]
+; IF-EVL-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; IF-EVL-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; IF-EVL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; IF-EVL-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
+; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP8]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP9]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP8]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; IF-EVL-NEXT:    [[TMP13:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0(ptr align 4 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META0:![0-9]+]]
+; IF-EVL-NEXT:    [[TMP16:%.*]] = call <vscale x 2 x i64> @llvm.vp.sext.nxv2i64.nxv2i32(<vscale x 2 x i32> [[VP_OP_LOAD]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP17]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP16]], ptr align 8 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
+; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP21:%.*]] = load i32, ptr [[GEP]], align 4
+; IF-EVL-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP21]] to i64
+; IF-EVL-NEXT:    [[GEP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store i64 [[CONV2]], ptr [[GEP4]], align 8
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP8:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_sext(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP0]] to i64
+; NO-VP-NEXT:    [[GEP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store i64 [[CONV2]], ptr [[GEP4]], align 8
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv2 = sext i32 %0 to i64
+  %gep4 = getelementptr inbounds i64, ptr %a, i64 %iv
+  store i64 %conv2, ptr %gep4, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_zext(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_zext(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 18, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = shl i64 [[N]], 3
+; IF-EVL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]]
+; IF-EVL-NEXT:    [[TMP6:%.*]] = shl i64 [[N]], 2
+; IF-EVL-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP6]]
+; IF-EVL-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; IF-EVL-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; IF-EVL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; IF-EVL-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
+; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP8]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP9]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP8]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; IF-EVL-NEXT:    [[TMP13:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0(ptr align 4 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META9:![0-9]+]]
+; IF-EVL-NEXT:    [[TMP16:%.*]] = call <vscale x 2 x i64> @llvm.vp.zext.nxv2i64.nxv2i32(<vscale x 2 x i32> [[VP_OP_LOAD]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP17]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP16]], ptr align 8 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META12:![0-9]+]], !noalias [[META9]]
+; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP21:%.*]] = load i32, ptr [[GEP]], align 4
+; IF-EVL-NEXT:    [[CONV:%.*]] = zext i32 [[TMP21]] to i64
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store i64 [[CONV]], ptr [[GEP2]], align 8
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_zext(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[CONV:%.*]] = zext i32 [[TMP0]] to i64
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store i64 [[CONV]], ptr [[GEP2]], align 8
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv = zext i32 %0 to i64
+  %gep2 = getelementptr inbounds i64, ptr %a, i64 %iv
+  store i64 %conv, ptr %gep2, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_trunc(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_trunc(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 18, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = shl i64 [[N]], 2
+; IF-EVL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]]
+; IF-EVL-NEXT:    [[TMP6:%.*]] = shl i64 [[N]], 3
+; IF-EVL-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP6]]
+; IF-EVL-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; IF-EVL-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; IF-EVL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; IF-EVL-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
+; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP8]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP9]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP8]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; IF-EVL-NEXT:    [[TMP13:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[TMP14]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META16:![0-9]+]]
+; IF-EVL-NEXT:    [[TMP16:%.*]] = call <vscale x 2 x i32> @llvm.vp.trunc.nxv2i32.nxv2i64(<vscale x 2 x i64> [[VP_OP_LOAD]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i32.p0(<vscale x 2 x i32> [[TMP16]], ptr align 4 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META19:![0-9]+]], !noalias [[META16]]
+; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP21:%.*]] = load i64, ptr [[GEP]], align 8
+; IF-EVL-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP21]] to i32
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP22:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_trunc(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[GEP]], align 8
+; NO-VP-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i64, ptr %b, i64 %iv
+  %0 = load i64, ptr %gep, align 8
+  %conv = trunc i64 %0 to i32
+  %gep2 = getelementptr inbounds i32, ptr %a, i64 %iv
+  store i32 %conv, ptr %gep2, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_fpext(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_fpext(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 14, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = shl i64 [[N]], 3
+; IF-EVL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]]
+; IF-EVL-NEXT:    [[TMP6:%.*]] = shl i64 [[N]], 2
+; IF-EVL-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP6]]
+; IF-EVL-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; IF-EVL-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; IF-EVL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; IF-EVL-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
+; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP8]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP9]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP8]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; IF-EVL-NEXT:    [[TMP13:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0(ptr align 4 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META23:![0-9]+]]
+; IF-EVL-NEXT:    [[TMP16:%.*]] = call <vscale x 2 x double> @llvm.vp.fpext.nxv2f64.nxv2f32(<vscale x 2 x float> [[VP_OP_LOAD]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, ptr [[TMP17]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2f64.p0(<vscale x 2 x double> [[TMP16]], ptr align 8 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META26:![0-9]+]], !noalias [[META23]]
+; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP21:%.*]] = load float, ptr [[GEP]], align 4
+; IF-EVL-NEXT:    [[CONV:%.*]] = fpext float [[TMP21]] to double
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store double [[CONV]], ptr [[GEP2]], align 8
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_fpext(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[CONV:%.*]] = fpext float [[TMP0]] to double
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store double [[CONV]], ptr [[GEP2]], align 8
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds float, ptr %b, i64 %iv
+  %0 = load float, ptr %gep, align 4
+  %conv = fpext float %0 to double
+  %gep2 = getelementptr inbounds double, ptr %a, i64 %iv
+  store double %conv, ptr %gep2, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_fptrunc(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_fptrunc(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 14, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = shl i64 [[N]], 2
+; IF-EVL-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]]
+; IF-EVL-NEXT:    [[TMP6:%.*]] = shl i64 [[N]], 3
+; IF-EVL-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP6]]
+; IF-EVL-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; IF-EVL-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; IF-EVL-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; IF-EVL-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 2
+; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP8]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP9]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP8]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; IF-EVL-NEXT:    [[TMP13:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds double, ptr [[TMP14]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x double> @llvm.vp.load.nxv2f64.p0(ptr align 8 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META30:![0-9]+]]
+; IF-EVL-NEXT:    [[TMP16:%.*]] = call <vscale x 2 x float> @llvm.vp.fptrunc.nxv2f32.nxv2f64(<vscale x 2 x double> [[VP_OP_LOAD]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP13]]
+; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2f32.p0(<vscale x 2 x float> [[TMP16]], ptr align 4 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META33:![0-9]+]], !noalias [[META30]]
+; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP21:%.*]] = load double, ptr [[GEP]], align 8
+; IF-EVL-NEXT:    [[CONV:%.*]] = fptrunc double [[TMP21]] to float
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP36:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_fptrunc(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load double, ptr [[GEP]], align 8
+; NO-VP-NEXT:    [[CONV:%.*]] = fptrunc double [[TMP0]] to float
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds double, ptr %b, i64 %iv
+  %0 = load double, ptr %gep, align 8
+  %conv = fptrunc double %0 to float
+  %gep2 = getelementptr inbounds float, ptr %a, i64 %iv
+  store float %conv, ptr %gep2, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_sitofp(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_sitofp(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; IF-EVL-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 10, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
+; IF-EVL-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 4
+; IF-EVL-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[B2]]
+; IF-EVL-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; IF-EVL-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 4
+; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP10]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP11]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP10]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
+; IF-EVL-NEXT:    [[TMP15:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP18:%.*]] = call <vscale x 4 x float> @llvm.vp.sitofp.nxv4f32.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP19]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
+; IF-EVL-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP22]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP23:%.*]] = load i32, ptr [[GEP]], align 4
+; IF-EVL-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP23]] to float
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP38:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_sitofp(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv = sitofp i32 %0 to float
+  %gep2 = getelementptr inbounds float, ptr %a, i64 %iv
+  store float %conv, ptr %gep2, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_uitofp(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_uitofp(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; IF-EVL-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 10, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
+; IF-EVL-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 4
+; IF-EVL-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[B2]]
+; IF-EVL-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; IF-EVL-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 4
+; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP10]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP11]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP10]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
+; IF-EVL-NEXT:    [[TMP15:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP18:%.*]] = call <vscale x 4 x float> @llvm.vp.uitofp.nxv4f32.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP19]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
+; IF-EVL-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP22]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP23:%.*]] = load i32, ptr [[GEP]], align 4
+; IF-EVL-NEXT:    [[CONV:%.*]] = uitofp i32 [[TMP23]] to float
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP40:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_uitofp(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[CONV:%.*]] = uitofp i32 [[TMP0]] to float
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv = uitofp i32 %0 to float
+  %gep2 = getelementptr inbounds float, ptr %a, i64 %iv
+  store float %conv, ptr %gep2, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_fptosi(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; IF-EVL-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 10, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
+; IF-EVL-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 4
+; IF-EVL-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[B2]]
+; IF-EVL-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; IF-EVL-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 4
+; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP10]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP11]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP10]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
+; IF-EVL-NEXT:    [[TMP15:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP18:%.*]] = call <vscale x 4 x i32> @llvm.vp.fptosi.nxv4i32.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
+; IF-EVL-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP22]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP23:%.*]] = load float, ptr [[GEP]], align 4
+; IF-EVL-NEXT:    [[CONV:%.*]] = fptosi float [[TMP23]] to i32
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP42:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_fptosi(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[CONV:%.*]] = fptosi float [[TMP0]] to i32
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds float, ptr %b, i64 %iv
+  %0 = load float, ptr %gep, align 4
+  %conv = fptosi float %0 to i32
+  %gep2 = getelementptr inbounds i32, ptr %a, i64 %iv
+  store i32 %conv, ptr %gep2, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_fptoui(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_fptoui(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; IF-EVL-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 4
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 10, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 4
+; IF-EVL-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 4
+; IF-EVL-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[B2]]
+; IF-EVL-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; IF-EVL-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 4
+; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP10]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP11]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP10]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 4
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
+; IF-EVL-NEXT:    [[TMP15:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP18:%.*]] = call <vscale x 4 x i32> @llvm.vp.fptoui.nxv4i32.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
+; IF-EVL-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP22]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP23:%.*]] = load float, ptr [[GEP]], align 4
+; IF-EVL-NEXT:    [[CONV:%.*]] = fptoui float [[TMP23]] to i32
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP44:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_fptoui(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[CONV:%.*]] = fptoui float [[TMP0]] to i32
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds float, ptr %b, i64 %iv
+  %0 = load float, ptr %gep, align 4
+  %conv = fptoui float %0 to i32
+  %gep2 = getelementptr inbounds i32, ptr %a, i64 %iv
+  store i32 %conv, ptr %gep2, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_inttoptr(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_inttoptr(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; IF-EVL-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; IF-EVL-NEXT:    [[TMP3:%.*]] = call i64 @llvm.umax.i64(i64 13, i64 [[TMP2]])
+; IF-EVL-NEXT:    [[TMP4:%.*]] = icmp ult i64 [[TMP0]], [[TMP3]]
+; IF-EVL-NEXT:    br i1 [[TMP4]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; IF-EVL:       [[VECTOR_MEMCHECK]]:
+; IF-EVL-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP6:%.*]] = mul i64 [[TMP5]], 2
+; IF-EVL-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 8
+; IF-EVL-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[B2]]
+; IF-EVL-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; IF-EVL-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 2
+; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP10]], 1
+; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP11]]
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP10]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 2
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; IF-EVL-NEXT:    [[TMP15:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[TMP16]], i32 0
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP18:%.*]] = call <vscale x 2 x ptr> @llvm.vp.inttoptr.nxv2p0.nxv2i64(<vscale x 2 x i64> [[VP_OP_LOAD]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[TMP15]]
+; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds ptr, ptr [[TMP19]], i32 0
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2p0.p0(<vscale x 2 x ptr> [[TMP18]], ptr align 8 [[TMP20]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
+; IF-EVL-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP22]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP23:%.*]] = load i64, ptr [[GEP]], align 8
+; IF-EVL-NEXT:    [[TMP24:%.*]] = inttoptr i64 [[TMP23]] to ptr
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store ptr [[TMP24]], ptr [[GEP2]], align 8
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP46:![0-9]+]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_inttoptr(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[GEP]], align 8
+; NO-VP-NEXT:    [[TMP1:%.*]] = inttoptr i64 [[TMP0]] to ptr
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store ptr [[TMP1]], ptr [[GEP2]], align 8
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i64, ptr %b, i64 %iv
+  %0 = load i64, ptr %gep, align 8
+  %1 = inttoptr i64 %0 to ptr
+  %gep2 = getelementptr inbounds ptr, ptr %a, i64 %iv
+  store ptr %1, ptr %gep2, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_ptrtoint(ptr %a, ptr %b, i64 %N) {
+; IF-EVL-LABEL: define void @vp_ptrtoint(
+; IF-EVL-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    br label %[[LOOP:.*]]
+; IF-EVL:       [[LOOP]]:
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; IF-EVL-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP0:%.*]] = ptrtoint ptr [[GEP]] to i64
+; IF-EVL-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    store i64 [[TMP0]], ptr [[GEP2]], align 8
+; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; IF-EVL:       [[EXIT]]:
+; IF-EVL-NEXT:    ret void
+;
+; NO-VP-LABEL: define void @vp_ptrtoint(
+; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    br label %[[LOOP:.*]]
+; NO-VP:       [[LOOP]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = ptrtoint ptr [[GEP]] to i64
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store i64 [[TMP0]], ptr [[GEP2]], align 8
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP:       [[EXIT]]:
+; NO-VP-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = ptrtoint ptr %gep to i64
+  %gep2 = getelementptr inbounds i64, ptr %a, i64 %iv
+  store i64 %0, ptr %gep2, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+;.
+; IF-EVL: [[META0]] = !{[[META1:![0-9]+]]}
+; IF-EVL: [[META1]] = distinct !{[[META1]], [[META2:![0-9]+]]}
+; IF-EVL: [[META2]] = distinct !{[[META2]], !"LVerDomain"}
+; IF-EVL: [[META3]] = !{[[META4:![0-9]+]]}
+; IF-EVL: [[META4]] = distinct !{[[META4]], [[META2]]}
+; IF-EVL: [[LOOP5]] = distinct !{[[LOOP5]], [[META6:![0-9]+]], [[META7:![0-9]+]]}
+; IF-EVL: [[META6]] = !{!"llvm.loop.isvectorized", i32 1}
+; IF-EVL: [[META7]] = !{!"llvm.loop.unroll.runtime.disable"}
+; IF-EVL: [[LOOP8]] = distinct !{[[LOOP8]], [[META6]]}
+; IF-EVL: [[META9]] = !{[[META10:![0-9]+]]}
+; IF-EVL: [[META10]] = distinct !{[[META10]], [[META11:![0-9]+]]}
+; IF-EVL: [[META11]] = distinct !{[[META11]], !"LVerDomain"}
+; IF-EVL: [[META12]] = !{[[META13:![0-9]+]]}
+; IF-EVL: [[META13]] = distinct !{[[META13]], [[META11]]}
+; IF-EVL: [[LOOP14]] = distinct !{[[LOOP14]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP15]] = distinct !{[[LOOP15]], [[META6]]}
+; IF-EVL: [[META16]] = !{[[META17:![0-9]+]]}
+; IF-EVL: [[META17]] = distinct !{[[META17]], [[META18:![0-9]+]]}
+; IF-EVL: [[META18]] = distinct !{[[META18]], !"LVerDomain"}
+; IF-EVL: [[META19]] = !{[[META20:![0-9]+]]}
+; IF-EVL: [[META20]] = distinct !{[[META20]], [[META18]]}
+; IF-EVL: [[LOOP21]] = distinct !{[[LOOP21]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP22]] = distinct !{[[LOOP22]], [[META6]]}
+; IF-EVL: [[META23]] = !{[[META24:![0-9]+]]}
+; IF-EVL: [[META24]] = distinct !{[[META24]], [[META25:![0-9]+]]}
+; IF-EVL: [[META25]] = distinct !{[[META25]], !"LVerDomain"}
+; IF-EVL: [[META26]] = !{[[META27:![0-9]+]]}
+; IF-EVL: [[META27]] = distinct !{[[META27]], [[META25]]}
+; IF-EVL: [[LOOP28]] = distinct !{[[LOOP28]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP29]] = distinct !{[[LOOP29]], [[META6]]}
+; IF-EVL: [[META30]] = !{[[META31:![0-9]+]]}
+; IF-EVL: [[META31]] = distinct !{[[META31]], [[META32:![0-9]+]]}
+; IF-EVL: [[META32]] = distinct !{[[META32]], !"LVerDomain"}
+; IF-EVL: [[META33]] = !{[[META34:![0-9]+]]}
+; IF-EVL: [[META34]] = distinct !{[[META34]], [[META32]]}
+; IF-EVL: [[LOOP35]] = distinct !{[[LOOP35]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP36]] = distinct !{[[LOOP36]], [[META6]]}
+; IF-EVL: [[LOOP37]] = distinct !{[[LOOP37]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP38]] = distinct !{[[LOOP38]], [[META6]]}
+; IF-EVL: [[LOOP39]] = distinct !{[[LOOP39]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP40]] = distinct !{[[LOOP40]], [[META6]]}
+; IF-EVL: [[LOOP41]] = distinct !{[[LOOP41]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP42]] = distinct !{[[LOOP42]], [[META6]]}
+; IF-EVL: [[LOOP43]] = distinct !{[[LOOP43]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP44]] = distinct !{[[LOOP44]], [[META6]]}
+; IF-EVL: [[LOOP45]] = distinct !{[[LOOP45]], [[META6]], [[META7]]}
+; IF-EVL: [[LOOP46]] = distinct !{[[LOOP46]], [[META6]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
index d64ae380799bfe..5c47f07f4436c3 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
@@ -332,12 +332,12 @@ define void @vp_lrint(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPEXT:%.+]]> = fpext  ir<[[LD1]]> to double
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[LRINT:%.+]]> = call llvm.vp.lrint(ir<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[TRUNC:%.+]]> = trunc ir<[[LRINT]]> to i32
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[LRINT:%.+]]> = call llvm.vp.lrint(vp<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LRINT]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[TRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[TRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -384,12 +384,12 @@ define void @vp_llrint(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[FPEXT:%.+]]> = fpext  ir<[[LD1]]> to double
-; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[LLRINT:%.+]]> = call llvm.vp.llrint(ir<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
-; IF-EVL-NEXT:     WIDEN-CAST ir<[[TRUNC:%.+]]> = trunc ir<[[LLRINT]]> to i32
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC ir<[[LLRINT:%.+]]> = call llvm.vp.llrint(vp<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LLRINT]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, ir<[[TRUNC]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[TRUNC]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
index 05e1040df6cd87..c8611c60966e30 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -102,7 +102,7 @@ exit:
   ret void
 }
 
-define void @vp_truncate(ptr %a, ptr %b, i64 %N) {
+define void @vp_trunc(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Initial VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -198,7 +198,7 @@ exit:
   ret void
 }
 
-define void @vp_fptrunct(ptr %a, ptr %b, i64 %N) {
+define void @vp_fptrunc(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -246,7 +246,7 @@ exit:
   ret void
 }
 
-define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
+define void @vp_sitofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -265,10 +265,10 @@ define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTOSI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTOSI]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -281,11 +281,11 @@ entry:
 
 loop:
   %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
-  %gep = getelementptr inbounds float, ptr %b, i64 %iv
-  %0 = load float, ptr %gep, align 4
-  %conv2 = fptoui float %0 to i32
-  %gep4 = getelementptr inbounds i32, ptr %a, i64 %iv
-  store i32 %conv2, ptr %gep4, align 4
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = load i32, ptr %gep, align 4
+  %conv2 = sitofp i32 %0 to float
+  %gep4 = getelementptr inbounds float, ptr %a, i64 %iv
+  store float %conv2, ptr %gep4, align 4
   %iv.next = add nuw nsw i64 %iv, 1
   %exitcond.not = icmp eq i64 %iv.next, %N
   br i1 %exitcond.not, label %exit, label %loop
@@ -294,7 +294,7 @@ exit:
   ret void
 }
 
-define void @vp_inttofp(ptr %a, ptr %b, i64 %N) {
+define void @vp_uitofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -313,10 +313,10 @@ define void @vp_inttofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[SITOFP:%.+]]> = call llvm.vp.sitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[SITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[UITOFP]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -331,7 +331,7 @@ loop:
   %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
   %gep = getelementptr inbounds i32, ptr %b, i64 %iv
   %0 = load i32, ptr %gep, align 4
-  %conv2 = sitofp i32 %0 to float
+  %conv2 = uitofp i32 %0 to float
   %gep4 = getelementptr inbounds float, ptr %a, i64 %iv
   store float %conv2, ptr %gep4, align 4
   %iv.next = add nuw nsw i64 %iv, 1
@@ -342,7 +342,7 @@ exit:
   ret void
 }
 
-define void @vp_uinttofp(ptr %a, ptr %b, i64 %N) {
+define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
 ; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
 ; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
@@ -361,10 +361,10 @@ define void @vp_uinttofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
 ; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[UITOFP:%.+]]> = call llvm.vp.uitofp(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTOSI:%.+]]> = call llvm.vp.fptosi(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
 ; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[UITOFP]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTOSI]]>, vp<[[EVL]]>
 ; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
 ; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -377,11 +377,59 @@ entry:
 
 loop:
   %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
-  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
-  %0 = load i32, ptr %gep, align 4
-  %conv2 = uitofp i32 %0 to float
-  %gep4 = getelementptr inbounds float, ptr %a, i64 %iv
-  store float %conv2, ptr %gep4, align 4
+  %gep = getelementptr inbounds float, ptr %b, i64 %iv
+  %0 = load float, ptr %gep, align 4
+  %conv2 = fptosi float %0 to i32
+  %gep4 = getelementptr inbounds i32, ptr %a, i64 %iv
+  store i32 %conv2, ptr %gep4, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_fptoui(ptr %a, ptr %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2,vscale x 4},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[FPTOUI:%.+]]> = call llvm.vp.fptoui(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[FPTOUI]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.next, %loop ], [ 0, %entry ]
+  %gep = getelementptr inbounds float, ptr %b, i64 %iv
+  %0 = load float, ptr %gep, align 4
+  %conv2 = fptoui float %0 to i32
+  %gep4 = getelementptr inbounds i32, ptr %a, i64 %iv
+  store i32 %conv2, ptr %gep4, align 4
   %iv.next = add nuw nsw i64 %iv, 1
   %exitcond.not = icmp eq i64 %iv.next, %N
   br i1 %exitcond.not, label %exit, label %loop
@@ -389,3 +437,71 @@ loop:
 exit:
   ret void
 }
+
+define void @vp_inttoptr(ptr %a, ptr %b, i64 %N) {
+; IF-EVL: VPlan 'Final VPlan for VF={vscale x 1,vscale x 2},UF={1}' {
+; IF-EVL-NEXT: Live-in vp<[[VFUF:%[0-9]+]]> = VF * UF
+; IF-EVL-NEXT: Live-in vp<[[VTC:%[0-9]+]]> = vector-trip-count
+; IF-EVL-NEXT: Live-in ir<%N> = original trip-count
+
+; IF-EVL: vector.ph:
+; IF-EVL-NEXT: Successor(s): vector loop
+
+; IF-EVL: <x1> vector loop: {
+; IF-EVL-NEXT:   vector.body:
+; IF-EVL-NEXT:     EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
+; IF-EVL-NEXT:     EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]>  = phi ir<0>, vp<[[IV_NEXT:%.+]]>
+; IF-EVL-NEXT:     EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
+; IF-EVL-NEXT:     vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
+; IF-EVL-NEXT:     CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
+; IF-EVL-NEXT:     WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     WIDEN-INTRINSIC vp<[[INTTOPTR:%.+]]> = call llvm.vp.inttoptr(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT:     CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
+; IF-EVL-NEXT:     vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
+; IF-EVL-NEXT:     WIDEN vp.store vp<[[PTR2]]>, vp<[[INTTOPTR]]>, vp<[[EVL]]>
+; IF-EVL-NEXT:     SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:     EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
+; IF-EVL-NEXT:     EMIT branch-on-count vp<[[IV_NEXT_EXIT]]>,  vp<[[VTC]]>
+; IF-EVL-NEXT:   No successors
+; IF-EVL-NEXT: }
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i64, ptr %b, i64 %iv
+  %0 = load i64, ptr %gep, align 8
+  %1 = inttoptr i64 %0 to ptr
+  %gep2 = getelementptr inbounds ptr, ptr %a, i64 %iv
+  store ptr %1, ptr %gep2, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
+
+define void @vp_ptrtoint(ptr %a, ptr %b, i64 %N) {
+; IF-EVL: LV: Vectorization is possible but not beneficial.
+
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
+  %gep = getelementptr inbounds i32, ptr %b, i64 %iv
+  %0 = ptrtoint ptr %gep to i64
+  %gep2 = getelementptr inbounds i64, ptr %a, i64 %iv
+  store i64 %0, ptr %gep2, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %exit, label %loop
+
+exit:
+  ret void
+}
\ No newline at end of file

>From 6b8fb807603d6a72ee4da943a29e10003f7cf492 Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Tue, 3 Dec 2024 11:53:36 +0800
Subject: [PATCH 9/9] fix the comments

---
 llvm/lib/Analysis/VectorUtils.cpp             |  4 +++
 .../lib/Transforms/Vectorize/VPlanRecipes.cpp | 26 ++++++-------------
 .../Transforms/Vectorize/VPlanTransforms.cpp  | 20 ++++++++------
 .../LoopVectorize/RISCV/inloop-reduction.ll   |  2 +-
 4 files changed, 25 insertions(+), 27 deletions(-)

diff --git a/llvm/lib/Analysis/VectorUtils.cpp b/llvm/lib/Analysis/VectorUtils.cpp
index 989090b80e1c87..6d03071c5026f6 100644
--- a/llvm/lib/Analysis/VectorUtils.cpp
+++ b/llvm/lib/Analysis/VectorUtils.cpp
@@ -144,6 +144,10 @@ bool llvm::isVectorIntrinsicWithOverloadTypeAtArg(
   if (TTI && Intrinsic::isTargetIntrinsic(ID))
     return TTI->isVectorIntrinsicWithOverloadTypeAtArg(ID, OpdIdx);
 
+  if (VPCastIntrinsic::isVPCast(ID)) {
+    return OpdIdx == -1 || OpdIdx == 0;
+  }
+
   switch (ID) {
   case Intrinsic::fptosi_sat:
   case Intrinsic::fptoui_sat:
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index aabf6243f236bb..f729c93c476242 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -962,31 +962,21 @@ void VPWidenIntrinsicRecipe::execute(VPTransformState &State) {
 
   // Use vector version of the intrinsic.
   Module *M = State.Builder.GetInsertBlock()->getModule();
-  bool IsVPIntrinsic = VPIntrinsic::isVPIntrinsic(VectorIntrinsicID);
   Function *VectorF =
-      IsVPIntrinsic
-          ? VPIntrinsic::getOrInsertDeclarationForParams(M, VectorIntrinsicID,
-                                                         TysForDecl[0], Args)
-          : Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
+      Intrinsic::getOrInsertDeclaration(M, VectorIntrinsicID, TysForDecl);
   assert(VectorF &&
          "Can't retrieve vector intrinsic or vector-predication intrinsics.");
 
+  auto *CI = cast_or_null<CallInst>(getUnderlyingValue());
   SmallVector<OperandBundleDef, 1> OpBundles;
-  if (!IsVPIntrinsic) {
-    if (auto *CI = cast_or_null<CallInst>(getUnderlyingValue()))
-      CI->getOperandBundlesAsDefs(OpBundles);
-  }
+  if (CI)
+    CI->getOperandBundlesAsDefs(OpBundles);
 
-  Instruction *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
+  CallInst *V = State.Builder.CreateCall(VectorF, Args, OpBundles);
 
-  if (IsVPIntrinsic) {
-    // Currently vp-intrinsics only accept FMF flags. llvm.vp.uitofp will get
-    // Flags of OperationType::NonNegOp && OperationType::FPMathOp.
-    if (isa<FPMathOperator>(V) && VectorIntrinsicID != Intrinsic::vp_uitofp)
-      setFlags(V);
-  } else {
-    setFlags(V);
-  }
+  // FIXME: vp.cast and vp.select dont pass the underlying instruction into the
+  // recipe, which set the flags and metadata be needed.
+  setFlags(V);
 
   if (!V->getType()->isVoidTy())
     State.set(this, V);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 1092e10104dad1..604e0fa18779ed 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1445,8 +1445,7 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
   VPTypeAnalysis TypeInfo(CanonicalIVType);
   LLVMContext &Ctx = CanonicalIVType->getContext();
   SmallVector<VPValue *> HeaderMasks = collectAllHeaderMasks(Plan);
-  VPValue *AllOneMask =
-      Plan.getOrAddLiveIn(ConstantInt::getTrue(IntegerType::getInt1Ty(Ctx)));
+
   for (VPUser *U : Plan.getVF().users()) {
     if (auto *R = dyn_cast<VPReverseVectorPointerRecipe>(U))
       R->setOperand(1, &EVL);
@@ -1486,14 +1485,17 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                     auto *CI = cast<CallInst>(CInst->getUnderlyingInstr());
                     Intrinsic::ID VPID = VPIntrinsic::getForIntrinsic(
                         CI->getCalledFunction()->getIntrinsicID());
-                    if (VPID == Intrinsic::not_intrinsic)
-                      return nullptr;
+                    assert(VPID != Intrinsic::not_intrinsic &&
+                           "Expected VP Instrinsic");
 
                     SmallVector<VPValue *> Ops(CInst->operands());
                     assert(VPIntrinsic::getMaskParamPos(VPID) &&
                            VPIntrinsic::getVectorLengthParamPos(VPID) &&
                            "Expected VP intrinsic");
-                    Ops.push_back(AllOneMask);
+                    VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
+                        IntegerType::getInt1Ty(CI->getContext())));
+                    Ops.push_back(Mask);
+                    ;
                     Ops.push_back(&EVL);
                     return new VPWidenIntrinsicRecipe(
                         *CI, VPID, Ops, TypeInfo.inferScalarType(CInst),
@@ -1504,14 +1506,16 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
                     auto *CI = dyn_cast<CastInst>(CInst->getUnderlyingInstr());
                     Intrinsic::ID VPID =
                         VPIntrinsic::getForOpcode(CI->getOpcode());
-                    if (VPID == Intrinsic::not_intrinsic)
-                      return nullptr;
+                    assert(VPID != Intrinsic::not_intrinsic &&
+                           "Expected vp.casts Instrinsic");
 
                     SmallVector<VPValue *> Ops(CInst->operands());
                     assert(VPIntrinsic::getMaskParamPos(VPID) &&
                            VPIntrinsic::getVectorLengthParamPos(VPID) &&
                            "Expected VP intrinsic");
-                    Ops.push_back(AllOneMask);
+                    VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
+                        IntegerType::getInt1Ty(CI->getContext())));
+                    Ops.push_back(Mask);
                     Ops.push_back(&EVL);
                     return new VPWidenIntrinsicRecipe(
                         VPID, Ops, TypeInfo.inferScalarType(CInst),
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
index 80c58afb746f3d..14818199072c28 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
@@ -203,7 +203,7 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-INLOOP-NEXT:    [[TMP14:%.*]] = call <vscale x 8 x i32> @llvm.vp.sext.nxv8i32.nxv8i16(<vscale x 8 x i16> [[VP_OP_LOAD]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[TMP14]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP11]] = add i32 [[TMP10]], [[VEC_PHI]]
-; IF-EVL-INLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i32 [[TMP5]], [[EVL_BASED_IV]]
+; IF-EVL-INLOOP-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i32 [[TMP6]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP4]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; IF-EVL-INLOOP-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]



More information about the llvm-commits mailing list