[RISCV] Add lowering for vp.fptosi and vp.sitofp.

[craig.topper]

Summary

This as an alternative version of D120641. Starting from the code here
https://repo.hca.bsc.es/gitlab/rferrer/llvm-epi/-/raw/EPI/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
but with some modifications to how the interim types are calculated,
and adding support for f16.

Still need to add fptosi for mask vectors.

Lots of masked isel patterns added so we can pass the mask through
the type changes.

[craig.topper]

Fix typo in comment

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

6175

Source repo had a separate else for DstTypeSize == SrcTypeSize. I merged them since the only difference was that else didn't check the pre-widening conditions. But the pre-widening conditions won't match for the single-width case so we can share the code.

[arcbbb]

llvm/test/CodeGen/RISCV/rvv/vfptosi-vp.ll

27

%m is not used here

[craig.topper]

llvm/test/CodeGen/RISCV/rvv/vfptosi-vp.ll

27

Oops. Thanks. I moved to a ConstantExpr and forgot to delete it.

[craig.topper]

Remove dead instructions from tests

[arcbbb]

LGTM

[frasercrmck]

LGTM otherwise

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

6171

I feel like the Ops[0], Ops[1] style isn't very common for us. Could we perhaps just use SDValue Src, Mask, VL?

6181

Just to double-check - it's not possible for this to end up in an illegal integer type, is it? We can never have f64 legal but i64 not - I guess that'd be Zve32d which doesn't exist.

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fptosi-vp.ll

3

Is there a reason we're not testing RV32 anymore?

[craig.topper]

Add rv32 tests.
Use Src, Mask, VL
Rename some variables.

llvm/lib/Target/RISCV/RISCVISelLowering.h

private:
   SDValue lowerFixedLengthVectorShiftToRVV(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerFixedLengthVectorSelectToRVV(SDValue Op,
                                             SelectionDAG &DAG) const;
   SDValue lowerToScalableOp(SDValue Op, SelectionDAG &DAG, unsigned NewOpc,
                             bool HasMask = true) const;
   SDValue lowerVPOp(SDValue Op, SelectionDAG &DAG, unsigned RISCVISDOpc) const;
   SDValue lowerLogicVPOp(SDValue Op, SelectionDAG &DAG, unsigned MaskOpc,
                          unsigned VecOpc) const;
+  SDValue lowerVPFPIntConvOp(SDValue Op, SelectionDAG &DAG,
+                             unsigned RISCVISDOpc) const;
   SDValue lowerFixedLengthVectorExtendToRVV(SDValue Op, SelectionDAG &DAG,
                                             unsigned ExtendOpc) const;
   SDValue lowerGET_ROUNDING(SDValue Op, SelectionDAG &DAG) const;
   SDValue lowerSET_ROUNDING(SDValue Op, SelectionDAG &DAG) const;
 
   SDValue expandUnalignedRVVLoad(SDValue Op, SelectionDAG &DAG) const;
   SDValue expandUnalignedRVVStore(SDValue Op, SelectionDAG &DAG) const;
 

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

if (Subtarget.hasVInstructions()) {
     static const unsigned IntegerVPOps[] = {
         ISD::VP_ADD,         ISD::VP_SUB,         ISD::VP_MUL,
         ISD::VP_SDIV,        ISD::VP_UDIV,        ISD::VP_SREM,
         ISD::VP_UREM,        ISD::VP_AND,         ISD::VP_OR,
         ISD::VP_XOR,         ISD::VP_ASHR,        ISD::VP_LSHR,
         ISD::VP_SHL,         ISD::VP_REDUCE_ADD,  ISD::VP_REDUCE_AND,
         ISD::VP_REDUCE_OR,   ISD::VP_REDUCE_XOR,  ISD::VP_REDUCE_SMAX,
         ISD::VP_REDUCE_SMIN, ISD::VP_REDUCE_UMAX, ISD::VP_REDUCE_UMIN,
-        ISD::VP_MERGE,       ISD::VP_SELECT};
+        ISD::VP_MERGE,       ISD::VP_SELECT,      ISD::VP_FPTOSI};
 
     static const unsigned FloatingPointVPOps[] = {
         ISD::VP_FADD,        ISD::VP_FSUB,        ISD::VP_FMUL,
         ISD::VP_FDIV,        ISD::VP_FNEG,        ISD::VP_FMA,
         ISD::VP_REDUCE_FADD, ISD::VP_REDUCE_SEQ_FADD, ISD::VP_REDUCE_FMIN,
-        ISD::VP_REDUCE_FMAX, ISD::VP_MERGE,       ISD::VP_SELECT};
+        ISD::VP_REDUCE_FMAX, ISD::VP_MERGE,       ISD::VP_SELECT,
+        ISD::VP_SITOFP};
 
     if (!Subtarget.is64Bit()) {
       // We must custom-lower certain vXi64 operations on RV32 due to the vector
       // element type being illegal.
       setOperationAction(ISD::INSERT_VECTOR_ELT, MVT::i64, Custom);
       setOperationAction(ISD::EXTRACT_VECTOR_ELT, MVT::i64, Custom);
 
       setOperationAction(ISD::VECREDUCE_ADD, MVT::i64, Custom);
if (Subtarget.useRVVForFixedLengthVectors()) {
         // Operations below are different for between masks and other vectors.
         if (VT.getVectorElementType() == MVT::i1) {
           setOperationAction(ISD::VP_AND, VT, Custom);
           setOperationAction(ISD::VP_OR, VT, Custom);
           setOperationAction(ISD::VP_XOR, VT, Custom);
           setOperationAction(ISD::AND, VT, Custom);
           setOperationAction(ISD::OR, VT, Custom);
           setOperationAction(ISD::XOR, VT, Custom);
+
+          setOperationAction(ISD::VP_FPTOSI, VT, Custom);
           continue;
         }
 
         // Make SPLAT_VECTOR Legal so DAGCombine will convert splat vectors to
         // it before type legalization for i64 vectors on RV32. It will then be
         // type legalized to SPLAT_VECTOR_PARTS which we need to Custom handle.
         // FIXME: Use SPLAT_VECTOR for all types? DAGCombine probably needs
         // improvements first.
SDValue RISCVTargetLowering::LowerOperation(SDValue Op,
   case ISD::VP_FMUL:
     return lowerVPOp(Op, DAG, RISCVISD::FMUL_VL);
   case ISD::VP_FDIV:
     return lowerVPOp(Op, DAG, RISCVISD::FDIV_VL);
   case ISD::VP_FNEG:
     return lowerVPOp(Op, DAG, RISCVISD::FNEG_VL);
   case ISD::VP_FMA:
     return lowerVPOp(Op, DAG, RISCVISD::FMA_VL);
+  case ISD::VP_FPTOSI:
+    return lowerVPFPIntConvOp(Op, DAG, RISCVISD::FP_TO_SINT_VL);
+  case ISD::VP_SITOFP:
+    return lowerVPFPIntConvOp(Op, DAG, RISCVISD::SINT_TO_FP_VL);
   }
 }
 
 static SDValue getTargetNode(GlobalAddressSDNode *N, SDLoc DL, EVT Ty,
                              SelectionDAG &DAG, unsigned Flags) {
   return DAG.getTargetGlobalAddress(N->getGlobal(), DL, Ty, 0, Flags);
 }
 
SDValue RISCVTargetLowering::lowerVPOp(SDValue Op, SelectionDAG &DAG,
 
   MVT ContainerVT = getContainerForFixedLengthVector(VT);
 
   SDValue VPOp = DAG.getNode(RISCVISDOpc, DL, ContainerVT, Ops);
 
   return convertFromScalableVector(VT, VPOp, DAG, Subtarget);
 }
 
+// Lower Floating-Point/Integer Type-Convert VP SDNodes
+SDValue RISCVTargetLowering::lowerVPFPIntConvOp(SDValue Op, SelectionDAG &DAG,
+                                                unsigned RISCVISDOpc) const {
+  SDLoc DL(Op);
+
+  SDValue Src = Op.getOperand(0);
+  SDValue Mask = Op.getOperand(1);
+  SDValue VL = Op.getOperand(2);
+
+  MVT DstVT = Op.getSimpleValueType();
+  MVT SrcVT = Src.getSimpleValueType();
+  if (DstVT.isFixedLengthVector()) {
+    DstVT = getContainerForFixedLengthVector(DstVT);
+    SrcVT = getContainerForFixedLengthVector(SrcVT);
+    Src = convertToScalableVector(SrcVT, Src, DAG, Subtarget);
+    MVT MaskVT = MVT::getVectorVT(MVT::i1, DstVT.getVectorElementCount());
+    Mask = convertToScalableVector(MaskVT, Mask, DAG, Subtarget);
+  }
+
+  unsigned RISCVISDExtOpc = (RISCVISDOpc == RISCVISD::SINT_TO_FP_VL ||
+                             RISCVISDOpc == RISCVISD::FP_TO_SINT_VL)
+                                ? RISCVISD::VSEXT_VL
+                                : RISCVISD::VZEXT_VL;
+
+  unsigned DstEltSize = DstVT.getScalarSizeInBits();
+  unsigned SrcEltSize = SrcVT.getScalarSizeInBits();
+
+  SDValue Result;
+  if (DstEltSize >= SrcEltSize) { // Single-width and widening conversion.
+    if (SrcVT.isInteger()) {
+      assert(DstVT.isFloatingPoint() && "Wrong input/output vector types");
+
+      // Do we need to do any pre-widening before converting?
+      if (SrcEltSize == 1) {
+        MVT IntVT = DstVT.changeVectorElementTypeToInteger();
+        MVT XLenVT = Subtarget.getXLenVT();
+        SDValue Zero = DAG.getConstant(0, DL, XLenVT);
+        SDValue ZeroSplat = DAG.getNode(RISCVISD::VMV_V_X_VL, DL, IntVT,
+                                        DAG.getUNDEF(IntVT), Zero, VL);
+        SDValue One = DAG.getConstant(
+            RISCVISDExtOpc == RISCVISD::VZEXT_VL ? 1 : -1, DL, XLenVT);
+        SDValue OneSplat = DAG.getNode(RISCVISD::VMV_V_X_VL, DL, IntVT,
+                                       DAG.getUNDEF(IntVT), One, VL);
+        Src = DAG.getNode(RISCVISD::VSELECT_VL, DL, IntVT, Src, OneSplat,
+                          ZeroSplat, VL);
+      } else if (DstEltSize > (2 * SrcEltSize)) {
+        // Widen before converting.
+        MVT IntVT = MVT::getVectorVT(MVT::getIntegerVT(DstEltSize / 2),
+                                     DstVT.getVectorElementCount());
+        Src = DAG.getNode(RISCVISDExtOpc, DL, IntVT, {Src, Mask, VL});
+      }
+
+      Result = DAG.getNode(RISCVISDOpc, DL, DstVT, {Src, Mask, VL});
+    } else {
+      assert(SrcVT.isFloatingPoint() && DstVT.isInteger() &&
+             "Wrong input/output vector types");
+
+      // Convert f16 to f32 then convert f32 to i64.
+      if (DstEltSize > (2 * SrcEltSize)) {
+        assert(SrcVT.getVectorElementType() == MVT::f16 && "Unexpected type!");
+        MVT InterimFVT =
+            MVT::getVectorVT(MVT::f32, DstVT.getVectorElementCount());
+        Src = DAG.getNode(RISCVISD::FP_EXTEND_VL, DL, InterimFVT,
+                          {Src, Mask, VL});
+      }
+
+      Result = DAG.getNode(RISCVISDOpc, DL, DstVT, {Src, Mask, VL});
+    }
+  } else { // Narrowing + Conversion
+    if (SrcVT.isInteger()) {
+      assert(DstVT.isFloatingPoint() && "Wrong input/output vector types");
+      // First do a narrowing convert to an FP type half the size, then round
+      // the FP type to a small FP type if needed.
+
+      MVT InterimFVT = DstVT;
+      if (SrcEltSize > (2 * DstEltSize)) {
+        assert(SrcEltSize == (4 * DstEltSize) && "Unexpected types!");
+        assert(DstVT.getVectorElementType() == MVT::f16 && "Unexpected type!");
+        InterimFVT = MVT::getVectorVT(MVT::f32, DstVT.getVectorElementCount());
+      }
+
+      Result = DAG.getNode(RISCVISDOpc, DL, InterimFVT, {Src, Mask, VL});
+
+      if (InterimFVT != DstVT) {
+        Src = Result;
+        Result = DAG.getNode(RISCVISD::FP_ROUND_VL, DL, DstVT, {Src, Mask, VL});
+      }
+    } else {
+      assert(SrcVT.isFloatingPoint() && DstVT.isInteger() &&
+             "Wrong input/output vector types");
+      // First do a narrowing conversion to an integer half the size, then
+      // truncate if needed.
+
+      // TODO: Handle mask vectors
+      assert(DstVT.getVectorElementType() != MVT::i1 &&
+             "Don't know how to handle masks yet!");
+      MVT InterimIVT = MVT::getVectorVT(MVT::getIntegerVT(SrcEltSize / 2),
+                                        DstVT.getVectorElementCount());
+
+      Result = DAG.getNode(RISCVISDOpc, DL, InterimIVT, {Src, Mask, VL});
+
+      while (InterimIVT != DstVT) {
+        SrcEltSize /= 2;
+        Src = Result;
+        InterimIVT = MVT::getVectorVT(MVT::getIntegerVT(SrcEltSize / 2),
+                                      DstVT.getVectorElementCount());
+        Result = DAG.getNode(RISCVISD::TRUNCATE_VECTOR_VL, DL, InterimIVT,
+                             {Src, Mask, VL});
+      }
+    }
+  }
+
+  MVT VT = Op.getSimpleValueType();
+  if (!VT.isFixedLengthVector())
+    return Result;
+  return convertFromScalableVector(VT, Result, DAG, Subtarget);
+}
+
 SDValue RISCVTargetLowering::lowerLogicVPOp(SDValue Op, SelectionDAG &DAG,
                                             unsigned MaskOpc,
                                             unsigned VecOpc) const {
   MVT VT = Op.getSimpleValueType();
   if (VT.getVectorElementType() != MVT::i1)
     return lowerVPOp(Op, DAG, VecOpc);
 
   // It is safe to drop mask parameter as masked-off elements are undef.

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

multiclass VPatExtendSDNode_V_VL<SDNode vop, string inst_name, string suffix,
                                  list <VTypeInfoToFraction> fraction_list> {
   foreach vtiTofti = fraction_list in {
     defvar vti = vtiTofti.Vti;
     defvar fti = vtiTofti.Fti;
     def : Pat<(vti.Vector (vop (fti.Vector fti.RegClass:$rs2),
                                true_mask, VLOpFrag)),
               (!cast<Instruction>(inst_name#"_"#suffix#"_"#vti.LMul.MX)
                   fti.RegClass:$rs2, GPR:$vl, vti.Log2SEW)>;
+    def : Pat<(vti.Vector (vop (fti.Vector fti.RegClass:$rs2),
+                               (fti.Mask V0), VLOpFrag)),
+              (!cast<Instruction>(inst_name#"_"#suffix#"_"#vti.LMul.MX#"_MASK")
+                  (vti.Vector (IMPLICIT_DEF)),
+                  fti.RegClass:$rs2,
+                  (fti.Mask V0), GPR:$vl, vti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 }
 
 multiclass VPatConvertFP2ISDNode_V_VL<SDNode vop, string instruction_name> {
   foreach fvti = AllFloatVectors in {
     defvar ivti = GetIntVTypeInfo<fvti>.Vti;
     def : Pat<(ivti.Vector (vop (fvti.Vector fvti.RegClass:$rs1),
                                 (fvti.Mask true_mask),
                                 VLOpFrag)),
               (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX)
                   fvti.RegClass:$rs1, GPR:$vl, ivti.Log2SEW)>;
+    def : Pat<(ivti.Vector (vop (fvti.Vector fvti.RegClass:$rs1),
+                                (fvti.Mask V0),
+                                VLOpFrag)),
+              (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX#"_MASK")
+                  (ivti.Vector (IMPLICIT_DEF)), fvti.RegClass:$rs1,
+                  (fvti.Mask V0), GPR:$vl, ivti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 }
 
 multiclass VPatConvertI2FPSDNode_V_VL<SDNode vop, string instruction_name> {
   foreach fvti = AllFloatVectors in {
     defvar ivti = GetIntVTypeInfo<fvti>.Vti;
     def : Pat<(fvti.Vector (vop (ivti.Vector ivti.RegClass:$rs1),
                                 (ivti.Mask true_mask),
                                 VLOpFrag)),
               (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX)
                   ivti.RegClass:$rs1, GPR:$vl, fvti.Log2SEW)>;
+    def : Pat<(fvti.Vector (vop (ivti.Vector ivti.RegClass:$rs1),
+                                (ivti.Mask V0),
+                                VLOpFrag)),
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_MASK")
+                  (fvti.Vector (IMPLICIT_DEF)), ivti.RegClass:$rs1,
+                  (ivti.Mask V0), GPR:$vl, fvti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 }
 
 multiclass VPatWConvertFP2ISDNode_V_VL<SDNode vop, string instruction_name> {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar iwti = GetIntVTypeInfo<fvtiToFWti.Wti>.Vti;
     def : Pat<(iwti.Vector (vop (fvti.Vector fvti.RegClass:$rs1),
                                 (fvti.Mask true_mask),
                                 VLOpFrag)),
               (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX)
                   fvti.RegClass:$rs1, GPR:$vl, fvti.Log2SEW)>;
+    def : Pat<(iwti.Vector (vop (fvti.Vector fvti.RegClass:$rs1),
+                                (fvti.Mask V0),
+                                VLOpFrag)),
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_MASK")
+                  (iwti.Vector (IMPLICIT_DEF)), fvti.RegClass:$rs1,
+                  (fvti.Mask V0), GPR:$vl, fvti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 }
 
 multiclass VPatWConvertI2FPSDNode_V_VL<SDNode vop, string instruction_name> {
   foreach vtiToWti = AllWidenableIntToFloatVectors in {
     defvar ivti = vtiToWti.Vti;
     defvar fwti = vtiToWti.Wti;
     def : Pat<(fwti.Vector (vop (ivti.Vector ivti.RegClass:$rs1),
                                 (ivti.Mask true_mask),
                                 VLOpFrag)),
               (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX)
                   ivti.RegClass:$rs1, GPR:$vl, ivti.Log2SEW)>;
+    def : Pat<(fwti.Vector (vop (ivti.Vector ivti.RegClass:$rs1),
+                                (ivti.Mask V0),
+                                VLOpFrag)),
+              (!cast<Instruction>(instruction_name#"_"#ivti.LMul.MX#"_MASK")
+                  (fwti.Vector (IMPLICIT_DEF)), ivti.RegClass:$rs1,
+                  (ivti.Mask V0), GPR:$vl, ivti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 }
 
 multiclass VPatNConvertFP2ISDNode_V_VL<SDNode vop, string instruction_name> {
   foreach vtiToWti = AllWidenableIntToFloatVectors in {
     defvar vti = vtiToWti.Vti;
     defvar fwti = vtiToWti.Wti;
     def : Pat<(vti.Vector (vop (fwti.Vector fwti.RegClass:$rs1),
                                (fwti.Mask true_mask),
                                VLOpFrag)),
               (!cast<Instruction>(instruction_name#"_"#vti.LMul.MX)
                   fwti.RegClass:$rs1, GPR:$vl, vti.Log2SEW)>;
+    def : Pat<(vti.Vector (vop (fwti.Vector fwti.RegClass:$rs1),
+                               (fwti.Mask V0),
+                               VLOpFrag)),
+              (!cast<Instruction>(instruction_name#"_"#vti.LMul.MX#"_MASK")
+                  (vti.Vector (IMPLICIT_DEF)), fwti.RegClass:$rs1,
+                  (fwti.Mask V0), GPR:$vl, vti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 }
 
 multiclass VPatNConvertI2FPSDNode_V_VL<SDNode vop, string instruction_name> {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar iwti = GetIntVTypeInfo<fvtiToFWti.Wti>.Vti;
     def : Pat<(fvti.Vector (vop (iwti.Vector iwti.RegClass:$rs1),
                                 (iwti.Mask true_mask),
                                 VLOpFrag)),
               (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX)
                   iwti.RegClass:$rs1, GPR:$vl, fvti.Log2SEW)>;
+    def : Pat<(fvti.Vector (vop (iwti.Vector iwti.RegClass:$rs1),
+                                (iwti.Mask V0),
+                                VLOpFrag)),
+              (!cast<Instruction>(instruction_name#"_"#fvti.LMul.MX#"_MASK")
+                  (fvti.Vector (IMPLICIT_DEF)), iwti.RegClass:$rs1,
+                  (iwti.Mask V0), GPR:$vl, fvti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 }
 
 multiclass VPatReductionVL<SDNode vop, string instruction_name, bit is_float> {
   foreach vti = !if(is_float, AllFloatVectors, AllIntegerVectors) in {
     defvar vti_m1 = !cast<VTypeInfo>(!if(is_float, "VF", "VI") # vti.SEW # "M1");
     def: Pat<(vti_m1.Vector (vop (vti_m1.Vector VR:$merge), (vti.Vector vti.RegClass:$rs1), VR:$rs2,
                                  (vti.Mask true_mask),

 foreach vtiTowti = AllWidenableIntVectors in {
   defvar vti = vtiTowti.Vti;
   defvar wti = vtiTowti.Wti;
   def : Pat<(vti.Vector (riscv_trunc_vector_vl (wti.Vector wti.RegClass:$rs1),
                                                (vti.Mask true_mask),
                                                VLOpFrag)),
             (!cast<Instruction>("PseudoVNSRL_WX_"#vti.LMul.MX)
                 wti.RegClass:$rs1, X0, GPR:$vl, vti.Log2SEW)>;
+  def : Pat<(vti.Vector (riscv_trunc_vector_vl (wti.Vector wti.RegClass:$rs1),
+                                               (vti.Mask V0),
+                                               VLOpFrag)),
+            (!cast<Instruction>("PseudoVNSRL_WX_"#vti.LMul.MX#"_MASK")
+                (vti.Vector (IMPLICIT_DEF)), wti.RegClass:$rs1, X0,
+                (vti.Mask V0), GPR:$vl, vti.Log2SEW, TAIL_AGNOSTIC)>;
 
   def : Pat<(vti.Vector
              (riscv_trunc_vector_vl
               (wti.Vector
                (riscv_sra_vl wti.RegClass:$rs1, (SplatPat XLenVT:$rs2),
                              true_mask, VLOpFrag)), true_mask, VLOpFrag)),
             (!cast<Instruction>("PseudoVNSRA_WX_"#vti.LMul.MX)
                 wti.RegClass:$rs1, GPR:$rs2, GPR:$vl, vti.Log2SEW)>;
foreach fvti = AllFloatVectors in {
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar fwti = fvtiToFWti.Wti;
     def : Pat<(fwti.Vector (riscv_fpextend_vl (fvti.Vector fvti.RegClass:$rs1),
                                               (fvti.Mask true_mask),
                                               VLOpFrag)),
               (!cast<Instruction>("PseudoVFWCVT_F_F_V_"#fvti.LMul.MX)
                   fvti.RegClass:$rs1, GPR:$vl, fvti.Log2SEW)>;
+    def : Pat<(fwti.Vector (riscv_fpextend_vl (fvti.Vector fvti.RegClass:$rs1),
+                                              (fvti.Mask V0),
+                                              VLOpFrag)),
+              (!cast<Instruction>("PseudoVFWCVT_F_F_V_"#fvti.LMul.MX#"_MASK")
+                  (fwti.Vector (IMPLICIT_DEF)), fvti.RegClass:$rs1,
+                  (fvti.Mask V0), GPR:$vl, fvti.Log2SEW, TAIL_AGNOSTIC)>;
   }
 
   // 14.19 Narrowing Floating-Point/Integer Type-Convert Instructions
   defm : VPatNConvertFP2ISDNode_V_VL<riscv_fp_to_sint_vl, "PseudoVFNCVT_RTZ_X_F_W">;
   defm : VPatNConvertFP2ISDNode_V_VL<riscv_fp_to_uint_vl, "PseudoVFNCVT_RTZ_XU_F_W">;
   defm : VPatNConvertI2FPSDNode_V_VL<riscv_sint_to_fp_vl, "PseudoVFNCVT_F_X_W">;
   defm : VPatNConvertI2FPSDNode_V_VL<riscv_uint_to_fp_vl, "PseudoVFNCVT_F_XU_W">;
   foreach fvtiToFWti = AllWidenableFloatVectors in {
     defvar fvti = fvtiToFWti.Vti;
     defvar fwti = fvtiToFWti.Wti;
     def : Pat<(fvti.Vector (riscv_fpround_vl (fwti.Vector fwti.RegClass:$rs1),
                                              (fwti.Mask true_mask),
                                              VLOpFrag)),
               (!cast<Instruction>("PseudoVFNCVT_F_F_W_"#fvti.LMul.MX)
                   fwti.RegClass:$rs1, GPR:$vl, fvti.Log2SEW)>;
+    def : Pat<(fvti.Vector (riscv_fpround_vl (fwti.Vector fwti.RegClass:$rs1),
+                                             (fwti.Mask V0),
+                                             VLOpFrag)),
+              (!cast<Instruction>("PseudoVFNCVT_F_F_W_"#fvti.LMul.MX#"_MASK")
+                  (fvti.Vector (IMPLICIT_DEF)), fwti.RegClass:$rs1,
+                  (fwti.Mask V0), GPR:$vl, fvti.Log2SEW, TAIL_AGNOSTIC)>;
 
     def : Pat<(fvti.Vector (riscv_fncvt_rod_vl (fwti.Vector fwti.RegClass:$rs1),
                                                (fwti.Mask true_mask),
                                                VLOpFrag)),
               (!cast<Instruction>("PseudoVFNCVT_ROD_F_F_W_"#fvti.LMul.MX)
                   fwti.RegClass:$rs1, GPR:$vl, fvti.Log2SEW)>;
   }
 }

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fptosi-vp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+zfh,+experimental-zvfh \
+; RUN:     -riscv-v-vector-bits-min=128 < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+zfh,+experimental-zvfh \
+; RUN:     -riscv-v-vector-bits-min=128 < %s | FileCheck %s
+
+declare <4 x i8> @llvm.vp.fptosi.v4i8.v4f16(<4 x half>, <4 x i1>, i32)
+
+define <4 x i8> @vfptosi_v4i8_v4f16(<4 x half> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i8_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x i8> @llvm.vp.fptosi.v4i8.v4f16(<4 x half> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i8> %v
+}
+
+define <4 x i8> @vfptosi_v4i8_v4f16_unmasked(<4 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i8_v4f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x i8> @llvm.vp.fptosi.v4i8.v4f16(<4 x half> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i8> %v
+}
+
+declare <4 x i16> @llvm.vp.fptosi.v4i16.v4f16(<4 x half>, <4 x i1>, i32)
+
+define <4 x i16> @vfptosi_v4i16_v4f16(<4 x half> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i16_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x i16> @llvm.vp.fptosi.v4i16.v4f16(<4 x half> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i16> %v
+}
+
+define <4 x i16> @vfptosi_v4i16_v4f16_unmasked(<4 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i16_v4f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x i16> @llvm.vp.fptosi.v4i16.v4f16(<4 x half> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i16> %v
+}
+
+declare <4 x i32> @llvm.vp.fptosi.v4i32.v4f16(<4 x half>, <4 x i1>, i32)
+
+define <4 x i32> @vfptosi_v4i32_v4f16(<4 x half> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i32_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x i32> @llvm.vp.fptosi.v4i32.v4f16(<4 x half> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i32> %v
+}
+
+define <4 x i32> @vfptosi_v4i32_v4f16_unmasked(<4 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i32_v4f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x i32> @llvm.vp.fptosi.v4i32.v4f16(<4 x half> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i32> %v
+}
+
+declare <4 x i64> @llvm.vp.fptosi.v4i64.v4f16(<4 x half>, <4 x i1>, i32)
+
+define <4 x i64> @vfptosi_v4i64_v4f16(<4 x half> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i64_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.f.v v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x i64> @llvm.vp.fptosi.v4i64.v4f16(<4 x half> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i64> %v
+}
+
+define <4 x i64> @vfptosi_v4i64_v4f16_unmasked(<4 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i64_v4f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.f.v v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x i64> @llvm.vp.fptosi.v4i64.v4f16(<4 x half> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i64> %v
+}
+
+declare <4 x i8> @llvm.vp.fptosi.v4i8.v4f32(<4 x float>, <4 x i1>, i32)
+
+define <4 x i8> @vfptosi_v4i8_v4f32(<4 x float> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i8_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x i8> @llvm.vp.fptosi.v4i8.v4f32(<4 x float> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i8> %v
+}
+
+define <4 x i8> @vfptosi_v4i8_v4f32_unmasked(<4 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i8_v4f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x i8> @llvm.vp.fptosi.v4i8.v4f32(<4 x float> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i8> %v
+}
+
+declare <4 x i16> @llvm.vp.fptosi.v4i16.v4f32(<4 x float>, <4 x i1>, i32)
+
+define <4 x i16> @vfptosi_v4i16_v4f32(<4 x float> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i16_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x i16> @llvm.vp.fptosi.v4i16.v4f32(<4 x float> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i16> %v
+}
+
+define <4 x i16> @vfptosi_v4i16_v4f32_unmasked(<4 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i16_v4f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x i16> @llvm.vp.fptosi.v4i16.v4f32(<4 x float> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i16> %v
+}
+
+declare <4 x i32> @llvm.vp.fptosi.v4i32.v4f32(<4 x float>, <4 x i1>, i32)
+
+define <4 x i32> @vfptosi_v4i32_v4f32(<4 x float> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i32_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x i32> @llvm.vp.fptosi.v4i32.v4f32(<4 x float> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i32> %v
+}
+
+define <4 x i32> @vfptosi_v4i32_v4f32_unmasked(<4 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i32_v4f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x i32> @llvm.vp.fptosi.v4i32.v4f32(<4 x float> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i32> %v
+}
+
+declare <4 x i64> @llvm.vp.fptosi.v4i64.v4f32(<4 x float>, <4 x i1>, i32)
+
+define <4 x i64> @vfptosi_v4i64_v4f32(<4 x float> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i64_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x i64> @llvm.vp.fptosi.v4i64.v4f32(<4 x float> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i64> %v
+}
+
+define <4 x i64> @vfptosi_v4i64_v4f32_unmasked(<4 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i64_v4f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v10, v8
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x i64> @llvm.vp.fptosi.v4i64.v4f32(<4 x float> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i64> %v
+}
+
+declare <4 x i8> @llvm.vp.fptosi.v4i8.v4f64(<4 x double>, <4 x i1>, i32)
+
+define <4 x i8> @vfptosi_v4i8_v4f64(<4 x double> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i8_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x i8> @llvm.vp.fptosi.v4i8.v4f64(<4 x double> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i8> %v
+}
+
+define <4 x i8> @vfptosi_v4i8_v4f64_unmasked(<4 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i8_v4f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x i8> @llvm.vp.fptosi.v4i8.v4f64(<4 x double> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i8> %v
+}
+
+declare <4 x i16> @llvm.vp.fptosi.v4i16.v4f64(<4 x double>, <4 x i1>, i32)
+
+define <4 x i16> @vfptosi_v4i16_v4f64(<4 x double> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i16_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x i16> @llvm.vp.fptosi.v4i16.v4f64(<4 x double> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i16> %v
+}
+
+define <4 x i16> @vfptosi_v4i16_v4f64_unmasked(<4 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i16_v4f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x i16> @llvm.vp.fptosi.v4i16.v4f64(<4 x double> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i16> %v
+}
+
+declare <4 x i32> @llvm.vp.fptosi.v4i32.v4f64(<4 x double>, <4 x i1>, i32)
+
+define <4 x i32> @vfptosi_v4i32_v4f64(<4 x double> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i32_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x i32> @llvm.vp.fptosi.v4i32.v4f64(<4 x double> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i32> %v
+}
+
+define <4 x i32> @vfptosi_v4i32_v4f64_unmasked(<4 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i32_v4f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x i32> @llvm.vp.fptosi.v4i32.v4f64(<4 x double> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i32> %v
+}
+
+declare <4 x i64> @llvm.vp.fptosi.v4i64.v4f64(<4 x double>, <4 x i1>, i32)
+
+define <4 x i64> @vfptosi_v4i64_v4f64(<4 x double> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i64_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x i64> @llvm.vp.fptosi.v4i64.v4f64(<4 x double> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x i64> %v
+}
+
+define <4 x i64> @vfptosi_v4i64_v4f64_unmasked(<4 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_v4i64_v4f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x i64> @llvm.vp.fptosi.v4i64.v4f64(<4 x double> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x i64> %v
+}
+

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-sitofp-vp-mask.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+zfh,+experimental-zvfh \
+; RUN:     -riscv-v-vector-bits-min=128 < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+zfh,+experimental-zvfh \
+; RUN:     -riscv-v-vector-bits-min=128 < %s | FileCheck %s
+
+declare <4 x half> @llvm.vp.sitofp.v4f16.v4i1(<4 x i1>, <4 x i1>, i32)
+
+define <4 x half> @vsitofp_v4f16_v4i1(<4 x i1> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmv.v.i v9, 0
+; CHECK-NEXT:    vmerge.vim v9, v9, -1, v0
+; CHECK-NEXT:    vmv1r.v v0, v8
+; CHECK-NEXT:    vfcvt.f.x.v v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i1(<4 x i1> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x half> %v
+}
+
+define <4 x half> @vsitofp_v4f16_v4i1_unmasked(<4 x i1> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i1_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmerge.vim v8, v8, -1, v0
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i1(<4 x i1> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x half> %v
+}
+
+declare <4 x float> @llvm.vp.sitofp.v4f32.v4i1(<4 x i1>, <4 x i1>, i32)
+
+define <4 x float> @vsitofp_v4f32_v4i1(<4 x i1> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmv.v.i v9, 0
+; CHECK-NEXT:    vmerge.vim v9, v9, -1, v0
+; CHECK-NEXT:    vmv1r.v v0, v8
+; CHECK-NEXT:    vfcvt.f.x.v v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i1(<4 x i1> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x float> %v
+}
+
+define <4 x float> @vsitofp_v4f32_v4i1_unmasked(<4 x i1> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i1_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmerge.vim v8, v8, -1, v0
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i1(<4 x i1> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <4 x double> @llvm.vp.sitofp.v4f64.v4i1(<4 x i1>, <4 x i1>, i32)
+
+define <4 x double> @vsitofp_v4f64_v4i1(<4 x i1> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmv.v.i v10, 0
+; CHECK-NEXT:    vmerge.vim v10, v10, -1, v0
+; CHECK-NEXT:    vmv1r.v v0, v8
+; CHECK-NEXT:    vfcvt.f.x.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i1(<4 x i1> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x double> %v
+}
+
+define <4 x double> @vsitofp_v4f64_v4i1_unmasked(<4 x i1> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i1_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmerge.vim v8, v8, -1, v0
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i1(<4 x i1> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x double> %v
+}

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-sitofp-vp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+zfh,+experimental-zvfh \
+; RUN:     -riscv-v-vector-bits-min=128 < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+zfh,+experimental-zvfh \
+; RUN:     -riscv-v-vector-bits-min=128 < %s | FileCheck %s
+
+declare <4 x half> @llvm.vp.sitofp.v4f16.v4i8(<4 x i8>, <4 x i1>, i32)
+
+define <4 x half> @vsitofp_v4f16_v4i8(<4 x i8> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i8(<4 x i8> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x half> %v
+}
+
+define <4 x half> @vsitofp_v4f16_v4i8_unmasked(<4 x i8> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i8_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i8(<4 x i8> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x half> %v
+}
+
+declare <4 x half> @llvm.vp.sitofp.v4f16.v4i16(<4 x i16>, <4 x i1>, i32)
+
+define <4 x half> @vsitofp_v4f16_v4i16(<4 x i16> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i16(<4 x i16> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x half> %v
+}
+
+define <4 x half> @vsitofp_v4f16_v4i16_unmasked(<4 x i16> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i16(<4 x i16> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x half> %v
+}
+
+declare <4 x half> @llvm.vp.sitofp.v4f16.v4i32(<4 x i32>, <4 x i1>, i32)
+
+define <4 x half> @vsitofp_v4f16_v4i32(<4 x i32> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i32(<4 x i32> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x half> %v
+}
+
+define <4 x half> @vsitofp_v4f16_v4i32_unmasked(<4 x i32> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i32(<4 x i32> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x half> %v
+}
+
+declare <4 x half> @llvm.vp.sitofp.v4f16.v4i64(<4 x i64>, <4 x i1>, i32)
+
+define <4 x half> @vsitofp_v4f16_v4i64(<4 x i64> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.f.w v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i64(<4 x i64> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x half> %v
+}
+
+define <4 x half> @vsitofp_v4f16_v4i64_unmasked(<4 x i64> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f16_v4i64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.f.w v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x half> @llvm.vp.sitofp.v4f16.v4i64(<4 x i64> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x half> %v
+}
+
+declare <4 x float> @llvm.vp.sitofp.v4f32.v4i8(<4 x i8>, <4 x i1>, i32)
+
+define <4 x float> @vsitofp_v4f32_v4i8(<4 x i8> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vsext.vf2 v9, v8, v0.t
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i8(<4 x i8> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x float> %v
+}
+
+define <4 x float> @vsitofp_v4f32_v4i8_unmasked(<4 x i8> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i8_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vsext.vf2 v9, v8
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i8(<4 x i8> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <4 x float> @llvm.vp.sitofp.v4f32.v4i16(<4 x i16>, <4 x i1>, i32)
+
+define <4 x float> @vsitofp_v4f32_v4i16(<4 x i16> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i16(<4 x i16> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x float> %v
+}
+
+define <4 x float> @vsitofp_v4f32_v4i16_unmasked(<4 x i16> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i16(<4 x i16> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <4 x float> @llvm.vp.sitofp.v4f32.v4i32(<4 x i32>, <4 x i1>, i32)
+
+define <4 x float> @vsitofp_v4f32_v4i32(<4 x i32> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i32(<4 x i32> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x float> %v
+}
+
+define <4 x float> @vsitofp_v4f32_v4i32_unmasked(<4 x i32> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i32(<4 x i32> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <4 x float> @llvm.vp.sitofp.v4f32.v4i64(<4 x i64>, <4 x i1>, i32)
+
+define <4 x float> @vsitofp_v4f32_v4i64(<4 x i64> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i64(<4 x i64> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x float> %v
+}
+
+define <4 x float> @vsitofp_v4f32_v4i64_unmasked(<4 x i64> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f32_v4i64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x float> @llvm.vp.sitofp.v4f32.v4i64(<4 x i64> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x float> %v
+}
+
+declare <4 x double> @llvm.vp.sitofp.v4f64.v4i8(<4 x i8>, <4 x i1>, i32)
+
+define <4 x double> @vsitofp_v4f64_v4i8(<4 x i8> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf4 v10, v8, v0.t
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i8(<4 x i8> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x double> %v
+}
+
+define <4 x double> @vsitofp_v4f64_v4i8_unmasked(<4 x i8> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i8_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf4 v10, v8
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i8(<4 x i8> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x double> %v
+}
+
+declare <4 x double> @llvm.vp.sitofp.v4f64.v4i16(<4 x i16>, <4 x i1>, i32)
+
+define <4 x double> @vsitofp_v4f64_v4i16(<4 x i16> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf2 v10, v8, v0.t
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i16(<4 x i16> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x double> %v
+}
+
+define <4 x double> @vsitofp_v4f64_v4i16_unmasked(<4 x i16> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf2 v10, v8
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i16(<4 x i16> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x double> %v
+}
+
+declare <4 x double> @llvm.vp.sitofp.v4f64.v4i32(<4 x i32>, <4 x i1>, i32)
+
+define <4 x double> @vsitofp_v4f64_v4i32(<4 x i32> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v10, v8, v0.t
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i32(<4 x i32> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x double> %v
+}
+
+define <4 x double> @vsitofp_v4f64_v4i32_unmasked(<4 x i32> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v10, v8
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i32(<4 x i32> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x double> %v
+}
+
+declare <4 x double> @llvm.vp.sitofp.v4f64.v4i64(<4 x i64>, <4 x i1>, i32)
+
+define <4 x double> @vsitofp_v4f64_v4i64(<4 x i64> %va, <4 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i64(<4 x i64> %va, <4 x i1> %m, i32 %evl)
+  ret <4 x double> %v
+}
+
+define <4 x double> @vsitofp_v4f64_v4i64_unmasked(<4 x i64> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_v4f64_v4i64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <4 x double> @llvm.vp.sitofp.v4f64.v4i64(<4 x i64> %va, <4 x i1> shufflevector (<4 x i1> insertelement (<4 x i1> undef, i1 true, i32 0), <4 x i1> undef, <4 x i32> zeroinitializer), i32 %evl)
+  ret <4 x double> %v
+}

llvm/test/CodeGen/RISCV/rvv/vfptosi-vp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+zfh,+experimental-zvfh < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+zfh,+experimental-zvfh < %s | FileCheck %s
+
+declare <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i8> @vfptosi_nxv2i8_nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i8_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i8> %v
+}
+
+define <vscale x 2 x i8> @vfptosi_nxv2i8_nxv2f16_unmasked(<vscale x 2 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i8_nxv2f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i8> %v
+}
+
+declare <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i16> @vfptosi_nxv2i16_nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i16_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i16> %v
+}
+
+define <vscale x 2 x i16> @vfptosi_nxv2i16_nxv2f16_unmasked(<vscale x 2 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i16_nxv2f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i16> %v
+}
+
+declare <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i32> @vfptosi_nxv2i32_nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i32_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i32> %v
+}
+
+define <vscale x 2 x i32> @vfptosi_nxv2i32_nxv2f16_unmasked(<vscale x 2 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i32_nxv2f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i32> %v
+}
+
+declare <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f16(<vscale x 2 x half>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i64> @vfptosi_nxv2i64_nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i64_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.f.v v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i64> %v
+}
+
+define <vscale x 2 x i64> @vfptosi_nxv2i64_nxv2f16_unmasked(<vscale x 2 x half> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i64_nxv2f16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.f.v v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f16(<vscale x 2 x half> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i64> %v
+}
+
+declare <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i8> @vfptosi_nxv2i8_nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i8_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i8> %v
+}
+
+define <vscale x 2 x i8> @vfptosi_nxv2i8_nxv2f32_unmasked(<vscale x 2 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i8_nxv2f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i8> %v
+}
+
+declare <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i16> @vfptosi_nxv2i16_nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i16_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i16> %v
+}
+
+define <vscale x 2 x i16> @vfptosi_nxv2i16_nxv2f32_unmasked(<vscale x 2 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i16_nxv2f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i16> %v
+}
+
+declare <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i32> @vfptosi_nxv2i32_nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i32_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i32> %v
+}
+
+define <vscale x 2 x i32> @vfptosi_nxv2i32_nxv2f32_unmasked(<vscale x 2 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i32_nxv2f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i32> %v
+}
+
+declare <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f32(<vscale x 2 x float>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i64> @vfptosi_nxv2i64_nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i64_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i64> %v
+}
+
+define <vscale x 2 x i64> @vfptosi_nxv2i64_nxv2f32_unmasked(<vscale x 2 x float> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i64_nxv2f32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.rtz.x.f.v v10, v8
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f32(<vscale x 2 x float> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i64> %v
+}
+
+declare <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i8> @vfptosi_nxv2i8_nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i8_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i8> %v
+}
+
+define <vscale x 2 x i8> @vfptosi_nxv2i8_nxv2f64_unmasked(<vscale x 2 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i8_nxv2f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10
+; CHECK-NEXT:    vsetvli zero, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i8> @llvm.vp.fptosi.nxv2i8.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i8> %v
+}
+
+declare <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i16> @vfptosi_nxv2i16_nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i16_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i16> %v
+}
+
+define <vscale x 2 x i16> @vfptosi_nxv2i16_nxv2f64_unmasked(<vscale x 2 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i16_nxv2f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vncvt.x.x.w v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i16> @llvm.vp.fptosi.nxv2i16.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i16> %v
+}
+
+declare <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i32> @vfptosi_nxv2i32_nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i32_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i32> %v
+}
+
+define <vscale x 2 x i32> @vfptosi_nxv2i32_nxv2f64_unmasked(<vscale x 2 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i32_nxv2f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.rtz.x.f.w v10, v8
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i32> @llvm.vp.fptosi.nxv2i32.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i32> %v
+}
+
+declare <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f64(<vscale x 2 x double>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x i64> @vfptosi_nxv2i64_nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i64_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x i64> %v
+}
+
+define <vscale x 2 x i64> @vfptosi_nxv2i64_nxv2f64_unmasked(<vscale x 2 x double> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vfptosi_nxv2i64_nxv2f64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x i64> @llvm.vp.fptosi.nxv2i64.nxv2f64(<vscale x 2 x double> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x i64> %v
+}

llvm/test/CodeGen/RISCV/rvv/vsitofp-vp-mask.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+zfh,+experimental-zvfh < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+zfh,+experimental-zvfh < %s | FileCheck %s
+
+declare <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i1(<vscale x 2 x i1>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmv.v.i v9, 0
+; CHECK-NEXT:    vmerge.vim v9, v9, -1, v0
+; CHECK-NEXT:    vmv1r.v v0, v8
+; CHECK-NEXT:    vfcvt.f.x.v v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i1_unmasked(<vscale x 2 x i1> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i1_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmerge.vim v8, v8, -1, v0
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+declare <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i1(<vscale x 2 x i1>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmv.v.i v9, 0
+; CHECK-NEXT:    vmerge.vim v9, v9, -1, v0
+; CHECK-NEXT:    vmv1r.v v0, v8
+; CHECK-NEXT:    vfcvt.f.x.v v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i1_unmasked(<vscale x 2 x i1> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i1_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmerge.vim v8, v8, -1, v0
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i1(<vscale x 2 x i1>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmv.v.i v10, 0
+; CHECK-NEXT:    vmerge.vim v10, v10, -1, v0
+; CHECK-NEXT:    vmv1r.v v0, v8
+; CHECK-NEXT:    vfcvt.f.x.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i1_unmasked(<vscale x 2 x i1> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i1_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vmv.v.i v8, 0
+; CHECK-NEXT:    vmerge.vim v8, v8, -1, v0
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i1(<vscale x 2 x i1> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x double> %v
+}

llvm/test/CodeGen/RISCV/rvv/vsitofp-vp.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+m,+v,+zfh,+experimental-zvfh < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+m,+v,+zfh,+experimental-zvfh < %s | FileCheck %s
+
+declare <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i8(<vscale x 2 x i8>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i8_unmasked(<vscale x 2 x i8> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i8_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf4, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+declare <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i16(<vscale x 2 x i16>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i16_unmasked(<vscale x 2 x i16> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+declare <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i32_unmasked(<vscale x 2 x i32> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+declare <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.f.w v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+define <vscale x 2 x half> @vsitofp_nxv2f16_nxv2i64_unmasked(<vscale x 2 x i64> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f16_nxv2i64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfncvt.f.f.w v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x half> @llvm.vp.sitofp.nxv2f16.nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x half> %v
+}
+
+declare <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i8(<vscale x 2 x i8>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vsext.vf2 v9, v8, v0.t
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v9, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i8_unmasked(<vscale x 2 x i8> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i8_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vsext.vf2 v9, v8
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+declare <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i16(<vscale x 2 x i16>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8, v0.t
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i16_unmasked(<vscale x 2 x i16> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e16, mf2, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v9, v8
+; CHECK-NEXT:    vmv1r.v v8, v9
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+declare <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i32_unmasked(<vscale x 2 x i32> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+declare <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8, v0.t
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+define <vscale x 2 x float> @vsitofp_nxv2f32_nxv2i64_unmasked(<vscale x 2 x i64> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f32_nxv2i64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfncvt.f.x.w v10, v8
+; CHECK-NEXT:    vmv.v.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x float> @llvm.vp.sitofp.nxv2f32.nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x float> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i8(<vscale x 2 x i8>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf4 v10, v8, v0.t
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i8_unmasked(<vscale x 2 x i8> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i8_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf4 v10, v8
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i16(<vscale x 2 x i16>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf2 v10, v8, v0.t
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i16_unmasked(<vscale x 2 x i16> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i16_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vsext.vf2 v10, v8
+; CHECK-NEXT:    vfwcvt.f.x.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i32(<vscale x 2 x i32>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v10, v8, v0.t
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i32_unmasked(<vscale x 2 x i32> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i32_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e32, m1, ta, mu
+; CHECK-NEXT:    vfwcvt.f.x.v v10, v8
+; CHECK-NEXT:    vmv2r.v v8, v10
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+declare <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i1>, i32)
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8, v0.t
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> %m, i32 %evl)
+  ret <vscale x 2 x double> %v
+}
+
+define <vscale x 2 x double> @vsitofp_nxv2f64_nxv2i64_unmasked(<vscale x 2 x i64> %va, i32 zeroext %evl) {
+; CHECK-LABEL: vsitofp_nxv2f64_nxv2i64_unmasked:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli zero, a0, e64, m2, ta, mu
+; CHECK-NEXT:    vfcvt.f.x.v v8, v8
+; CHECK-NEXT:    ret
+  %v = call <vscale x 2 x double> @llvm.vp.sitofp.nxv2f64.nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> undef, i1 true, i32 0), <vscale x 2 x i1> undef, <vscale x 2 x i32> zeroinitializer), i32 %evl)
+  ret <vscale x 2 x double> %v
+}