[RISCV] Add support for fixed vector floating point setcc.

[craig.topper]

Summary

This is annoying because the condition code legalization belongs
to LegalizeDAG, but our custom handler runs in Legalize vector ops
which occurs earlier.

This adds some of the mask binary operations so that we can combine
multiple compares that we need for expansion.

I've also fixed up RISCVISelDAGToDAG.cpp to handle copies of masks.

This patch contains a subset of the integer setcc patch as well.
That patch is dependent on the integer binary ops patch. I'll rebase
based on what order the patches go in.

[craig.topper]

clang-format

[frasercrmck]

Looks good on the whole, just some minor comments. Shame about having to do our own CC legalization though.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2146

I'd probably use Optional<unsigned> for this but maybe that's personal preference.

2201

I think this block could be commented, like why is SETOLT special.

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

99

I was wondering if this warrants a SDTCisVecWithSameNumEltsAndEltVT<opidx, otheropidx, eltvt> to combine these three commonly-used constraints (SDTCisVec<1>, SDTCisSameNumEltsAs<0, 1>, SDTCVecEltisVT<0, X>) but the name alone gives me second thoughts.

[craig.topper]

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

99

I think the tablegen backend maps the name of the class to dispatch to a function that does the type constraining.

I suspect SDTCisVec is redundant if SDTCVecEltisVT is specified.

[craig.topper]

-rebase
-Address review comments.
-Remove SDTCisVec when SDTCVecEltisVT is specified.

[frasercrmck]

LGTM, cheers

llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp

void RISCVDAGToDAGISel::Select(SDNode *Node) {
   case ISD::INSERT_SUBVECTOR: {
     // Bail when not a "cast" like insert_subvector.
     if (Node->getConstantOperandVal(2) != 0)
       break;
     if (!Node->getOperand(0).isUndef())
       break;
 
     // Bail when normal isel should do the job.
-    EVT InVT = Node->getOperand(1).getValueType();
+    MVT InVT = Node->getOperand(1).getSimpleValueType();
     if (VT.isFixedLengthVector() || InVT.isScalableVector())
       break;
 
+    unsigned RegClassID;
+    if (VT.getVectorElementType() == MVT::i1)
+      RegClassID = RISCV::VRRegClassID;
+    else
+      RegClassID = getRegClassIDForLMUL(getLMUL(VT));
+
     SDValue V = Node->getOperand(1);
     SDLoc DL(V);
-    unsigned RegClassID = getRegClassIDForLMUL(getLMUL(VT));
     SDValue RC =
         CurDAG->getTargetConstant(RegClassID, DL, Subtarget->getXLenVT());
     SDNode *NewNode =
         CurDAG->getMachineNode(TargetOpcode::COPY_TO_REGCLASS, DL, VT, V, RC);
     ReplaceNode(Node, NewNode);
     return;
   }
   case ISD::EXTRACT_SUBVECTOR: {
     // Bail when not a "cast" like extract_subvector.
     if (Node->getConstantOperandVal(1) != 0)
       break;
 
     // Bail when normal isel can do the job.
-    EVT InVT = Node->getOperand(0).getValueType();
+    MVT InVT = Node->getOperand(0).getSimpleValueType();
     if (VT.isScalableVector() || InVT.isFixedLengthVector())
       break;
 
+    unsigned RegClassID;
+    if (InVT.getVectorElementType() == MVT::i1)
+      RegClassID = RISCV::VRRegClassID;
+    else
+      RegClassID = getRegClassIDForLMUL(getLMUL(InVT));
+
     SDValue V = Node->getOperand(0);
     SDLoc DL(V);
-    unsigned RegClassID = getRegClassIDForLMUL(getLMUL(InVT.getSimpleVT()));
     SDValue RC =
         CurDAG->getTargetConstant(RegClassID, DL, Subtarget->getXLenVT());
     SDNode *NewNode =
         CurDAG->getMachineNode(TargetOpcode::COPY_TO_REGCLASS, DL, VT, V, RC);
     ReplaceNode(Node, NewNode);
     return;
   }
   }

llvm/lib/Target/RISCV/RISCVISelLowering.h

enum NodeType : unsigned {
   SMAX_VL,
   UMIN_VL,
   UMAX_VL,
 
   // Vector compare producing a mask. Fourth operand is input mask. Fifth
   // operand is VL.
   SETCC_VL,
 
+  // Mask binary operators.
+  VMAND_VL,
+  VMOR_VL,
+  VMXOR_VL,
+
   // Set mask vector to all zeros or ones.
   VMCLR_VL,
   VMSET_VL,
 
   // Matches the semantics of vrgather.vx with an extra operand for VL.
   VRGATHER_VX_VL,
 
   // Memory opcodes start here.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

if (Subtarget.useRVVForFixedLengthVectors()) {
         setOperationAction(ISD::FADD, VT, Custom);
         setOperationAction(ISD::FSUB, VT, Custom);
         setOperationAction(ISD::FMUL, VT, Custom);
         setOperationAction(ISD::FDIV, VT, Custom);
         setOperationAction(ISD::FNEG, VT, Custom);
         setOperationAction(ISD::FABS, VT, Custom);
         setOperationAction(ISD::FSQRT, VT, Custom);
         setOperationAction(ISD::FMA, VT, Custom);
+
+        for (auto CC : VFPCCToExpand)
+          setCondCodeAction(CC, VT, Expand);
       }
     }
   }
 
   // Function alignments.
   const Align FunctionAlignment(Subtarget.hasStdExtC() ? 2 : 4);
   setMinFunctionAlignment(FunctionAlignment);
   setPrefFunctionAlignment(FunctionAlignment);
SDValue Op1 =
       convertToScalableVector(ContainerVT, Op.getOperand(0), DAG, Subtarget);
   SDValue Op2 =
       convertToScalableVector(ContainerVT, Op.getOperand(1), DAG, Subtarget);
 
   SDLoc DL(Op);
   SDValue VL =
       DAG.getConstant(VT.getVectorNumElements(), DL, Subtarget.getXLenVT());
 
+  ISD::CondCode CC = cast<CondCodeSDNode>(Op.getOperand(2))->get();
+
+  bool Invert = false;
+  Optional<unsigned> LogicOpc;
+  if (ContainerVT.isFloatingPoint()) {
+    bool Swap = false;
+    switch (CC) {
+    default:
+      break;
+    case ISD::SETULE:
+    case ISD::SETULT:
+      Swap = true;
+      LLVM_FALLTHROUGH;
+    case ISD::SETUGE:
+    case ISD::SETUGT:
+      CC = getSetCCInverse(CC, ContainerVT);
+      Invert = true;
+      break;
+    case ISD::SETOGE:
+    case ISD::SETOGT:
+    case ISD::SETGE:
+    case ISD::SETGT:
+      Swap = true;
+      break;
+    case ISD::SETUEQ:
+      // Use !((OLT Op1, Op2) || (OLT Op2, Op1))
+      Invert = true;
+      LogicOpc = RISCVISD::VMOR_VL;
+      CC = ISD::SETOLT;
+      break;
+    case ISD::SETONE:
+      // Use ((OLT Op1, Op2) || (OLT Op2, Op1))
+      LogicOpc = RISCVISD::VMOR_VL;
+      CC = ISD::SETOLT;
+      break;
+    case ISD::SETO:
+      // Use (OEQ Op1, Op1) && (OEQ Op2, Op2)
+      LogicOpc = RISCVISD::VMAND_VL;
+      CC = ISD::SETOEQ;
+      break;
+    case ISD::SETUO:
+      // Use (UNE Op1, Op1) || (UNE Op2, Op2)
+      LogicOpc = RISCVISD::VMOR_VL;
+      CC = ISD::SETUNE;
+      break;
+    }
+
+    if (Swap) {
+      CC = getSetCCSwappedOperands(CC);
+      std::swap(Op1, Op2);
+    }
+  }
+
   MVT MaskVT = MVT::getVectorVT(MVT::i1, ContainerVT.getVectorElementCount());
   SDValue Mask = DAG.getNode(RISCVISD::VMSET_VL, DL, MaskVT, VL);
-  SDValue Cmp = DAG.getNode(RISCVISD::SETCC_VL, DL, MaskVT, Op1, Op2,
-                            Op.getOperand(2), Mask, VL);
+
+  // There are 3 cases we need to emit.
+  // 1. For (OEQ Op1, Op1) && (OEQ Op2, Op2) or (UNE Op1, Op1) || (UNE Op2, Op2)
+  //    we need to compare each operand with itself.
+  // 2. For (OLT Op1, Op2) || (OLT Op2, Op1) we need to compare Op1 and Op2 in
+  //    both orders.
+  // 3. For any other case we just need one compare with Op1 and Op2.
+  SDValue Cmp;
+  if (LogicOpc && (CC == ISD::SETOEQ || CC == ISD::SETUNE)) {
+    Cmp = DAG.getNode(RISCVISD::SETCC_VL, DL, MaskVT, Op1, Op1,
+                      DAG.getCondCode(CC), Mask, VL);
+    SDValue Cmp2 = DAG.getNode(RISCVISD::SETCC_VL, DL, MaskVT, Op2, Op2,
+                               DAG.getCondCode(CC), Mask, VL);
+    Cmp = DAG.getNode(*LogicOpc, DL, MaskVT, Cmp, Cmp2, VL);
+  } else {
+    Cmp = DAG.getNode(RISCVISD::SETCC_VL, DL, MaskVT, Op1, Op2,
+                      DAG.getCondCode(CC), Mask, VL);
+    if (LogicOpc) {
+      SDValue Cmp2 = DAG.getNode(RISCVISD::SETCC_VL, DL, MaskVT, Op2, Op1,
+                                 DAG.getCondCode(CC), Mask, VL);
+      Cmp = DAG.getNode(*LogicOpc, DL, MaskVT, Cmp, Cmp2, VL);
+    }
+  }
+
+  if (Invert) {
+    SDValue AllOnes = DAG.getNode(RISCVISD::VMSET_VL, DL, MaskVT, VL);
+    Cmp = DAG.getNode(RISCVISD::VMXOR_VL, DL, MaskVT, Cmp, AllOnes, VL);
+  }
 
   return convertFromScalableVector(VT, Cmp, DAG, Subtarget);
 }
 
 SDValue RISCVTargetLowering::lowerToScalableOp(SDValue Op, SelectionDAG &DAG,
                                                unsigned NewOpc) const {
   MVT VT = Op.getSimpleValueType();
   assert(useRVVForFixedLengthVectorVT(VT) &&
#define NODE_NAME_CASE(NODE)                                                   \
   NODE_NAME_CASE(FABS_VL)
   NODE_NAME_CASE(FSQRT_VL)
   NODE_NAME_CASE(FMA_VL)
   NODE_NAME_CASE(SMIN_VL)
   NODE_NAME_CASE(SMAX_VL)
   NODE_NAME_CASE(UMIN_VL)
   NODE_NAME_CASE(UMAX_VL)
   NODE_NAME_CASE(SETCC_VL)
+  NODE_NAME_CASE(VMAND_VL)
+  NODE_NAME_CASE(VMOR_VL)
+  NODE_NAME_CASE(VMXOR_VL)
   NODE_NAME_CASE(VMCLR_VL)
   NODE_NAME_CASE(VMSET_VL)
   NODE_NAME_CASE(VRGATHER_VX_VL)
   NODE_NAME_CASE(VLE_VL)
   NODE_NAME_CASE(VSE_VL)
   }
   // clang-format on
   return nullptr;

llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td

multiclass VPatIntegerSetCCSDNode_VX_VI<CondCode cc,
                                         string instruction_name,
                                         bit swap = 0> {
   defm : VPatIntegerSetCCSDNode_XI<cc, instruction_name, "VX",
                                    SplatPat, GPR, swap>;
   defm : VPatIntegerSetCCSDNode_XI<cc, instruction_name, "VI",
                                    SplatPat_simm5, simm5, swap>;
 }
 
-multiclass VPatFPSetCCSDNode_VV<CondCode cc, string instruction_name> {
-  foreach fvti = AllFloatVectors in
+multiclass VPatFPSetCCSDNode_VV_VF_FV<CondCode cc,
+                                      string inst_name,
+                                      string swapped_op_inst_name> {
+  foreach fvti = AllFloatVectors in {
     def : Pat<(fvti.Mask (setcc (fvti.Vector fvti.RegClass:$rs1),
                                 (fvti.Vector fvti.RegClass:$rs2),
                                 cc)),
-              (!cast<Instruction>(instruction_name#"_VV_"#fvti.LMul.MX)
+              (!cast<Instruction>(inst_name#"_VV_"#fvti.LMul.MX)
                   fvti.RegClass:$rs1, fvti.RegClass:$rs2, fvti.AVL, fvti.SEW)>;
-}
-
-multiclass VPatFPSetCCSDNode_VF<CondCode cc, string instruction_name> {
-  foreach fvti = AllFloatVectors in
     def : Pat<(fvti.Mask (setcc (fvti.Vector fvti.RegClass:$rs1),
-                                (fvti.Vector (splat_vector fvti.ScalarRegClass:$rs2)),
+                                (splat_vector fvti.ScalarRegClass:$rs2),
                                 cc)),
-              (!cast<Instruction>(instruction_name#"_V"#fvti.ScalarSuffix#"_"#fvti.LMul.MX)
-                  fvti.RegClass:$rs1,
-                  (fvti.Scalar fvti.ScalarRegClass:$rs2),
+              (!cast<Instruction>(inst_name#"_V"#fvti.ScalarSuffix#"_"#fvti.LMul.MX)
+                  fvti.RegClass:$rs1, fvti.ScalarRegClass:$rs2,
                   fvti.AVL, fvti.SEW)>;
-}
-
-multiclass VPatFPSetCCSDNode_FV<CondCode cc, string swapped_op_instruction_name> {
-  foreach fvti = AllFloatVectors in
-    def : Pat<(fvti.Mask (setcc (fvti.Vector (splat_vector fvti.ScalarRegClass:$rs2)),
+    def : Pat<(fvti.Mask (setcc (splat_vector fvti.ScalarRegClass:$rs2),
                                 (fvti.Vector fvti.RegClass:$rs1),
                                 cc)),
-              (!cast<Instruction>(swapped_op_instruction_name#"_V"#fvti.ScalarSuffix#"_"#fvti.LMul.MX)
-                  fvti.RegClass:$rs1,
-                  (fvti.Scalar fvti.ScalarRegClass:$rs2),
+              (!cast<Instruction>(swapped_op_inst_name#"_V"#fvti.ScalarSuffix#"_"#fvti.LMul.MX)
+                  fvti.RegClass:$rs1, fvti.ScalarRegClass:$rs2,
                   fvti.AVL, fvti.SEW)>;
   }
-
-multiclass VPatFPSetCCSDNode_VV_VF_FV<CondCode cc,
-                                      string inst_name,
-                                      string swapped_op_inst_name> {
-  defm : VPatFPSetCCSDNode_VV<cc, inst_name>;
-  defm : VPatFPSetCCSDNode_VF<cc, inst_name>;
-  defm : VPatFPSetCCSDNode_FV<cc, swapped_op_inst_name>;
 }
 
 multiclass VPatExtendSDNode_V<list<SDNode> ops, string inst_name, string suffix,
                               list <VTypeInfoToFraction> fraction_list> {
   foreach vtiTofti = fraction_list in {
     defvar vti = vtiTofti.Vti;
     defvar fti = vtiTofti.Fti;
     foreach op = ops in

llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td

def SDT_RISCVVecFMA_VL : SDTypeProfile<1, 5, [SDTCisSameAs<0, 1>,
                                               SDTCisSameAs<0, 3>,
                                               SDTCisVec<0>, SDTCisFP<0>,
                                               SDTCVecEltisVT<4, i1>,
                                               SDTCisSameNumEltsAs<0, 4>,
                                               SDTCisVT<5, XLenVT>]>;
 def riscv_fma_vl : SDNode<"RISCVISD::FMA_VL", SDT_RISCVVecFMA_VL>;
 
 def riscv_setcc_vl : SDNode<"RISCVISD::SETCC_VL",
-                            SDTypeProfile<1, 5, [SDTCisVec<0>,
-                                                 SDTCVecEltisVT<0, i1>,
+                            SDTypeProfile<1, 5, [SDTCVecEltisVT<0, i1>,
                                                  SDTCisVec<1>,
                                                  SDTCisSameNumEltsAs<0, 1>,
                                                  SDTCisSameAs<1, 2>,
                                                  SDTCisVT<3, OtherVT>,
                                                  SDTCisSameAs<0, 4>,
                                                  SDTCisVT<5, XLenVT>]>>;
 
 def riscv_vrgather_vx_vl : SDNode<"RISCVISD::VRGATHER_VX_VL",
                                   SDTypeProfile<1, 4, [SDTCisVec<0>,
                                                        SDTCisSameAs<0, 1>,
                                                        SDTCisVT<2, XLenVT>,
                                                        SDTCVecEltisVT<3, i1>,
                                                        SDTCisSameNumEltsAs<0, 3>,
                                                        SDTCisVT<4, XLenVT>]>>;
 
-def SDT_RISCVVMSETCLR_VL : SDTypeProfile<1, 1, [SDTCisVec<0>,
+def SDT_RISCVMaskBinOp_VL : SDTypeProfile<1, 3, [SDTCisSameAs<0, 1>,
+                                                 SDTCisSameAs<0, 2>,
                                                  SDTCVecEltisVT<0, i1>,
+                                                 SDTCisVT<3, XLenVT>]>;
+def riscv_vmand_vl : SDNode<"RISCVISD::VMAND_VL", SDT_RISCVMaskBinOp_VL, [SDNPCommutative]>;
+def riscv_vmor_vl  : SDNode<"RISCVISD::VMOR_VL",  SDT_RISCVMaskBinOp_VL, [SDNPCommutative]>;
+def riscv_vmxor_vl : SDNode<"RISCVISD::VMXOR_VL", SDT_RISCVMaskBinOp_VL, [SDNPCommutative]>;
+
+def SDT_RISCVVMSETCLR_VL : SDTypeProfile<1, 1, [SDTCVecEltisVT<0, i1>,
                                                 SDTCisVT<1, XLenVT>]>;
 def riscv_vmclr_vl : SDNode<"RISCVISD::VMCLR_VL", SDT_RISCVVMSETCLR_VL>;
 def riscv_vmset_vl : SDNode<"RISCVISD::VMSET_VL", SDT_RISCVVMSETCLR_VL>;
 
 def true_mask : PatLeaf<(riscv_vmset_vl (XLenVT srcvalue))>;
 
 // Ignore the vl operand.
 def SplatFPOp : PatFrag<(ops node:$op),
def : Pat<(vti.Mask (riscv_setcc_vl (vti.Vector vti.RegClass:$rs1),
             (instruction vti.RegClass:$rs1, XLenVT:$rs2, GPR:$vl, vti.SEW)>;
   def : Pat<(vti.Mask (riscv_setcc_vl (SplatPat_simm5 simm5:$rs2),
                                       (vti.Vector vti.RegClass:$rs1), invcc,
                                       (vti.Mask true_mask),
                                       (XLenVT (VLOp GPR:$vl)))),
             (instruction vti.RegClass:$rs1, simm5:$rs2, GPR:$vl, vti.SEW)>;
 }
 
+multiclass VPatFPSetCCVL_VV_VF_FV<CondCode cc,
+                                  string inst_name,
+                                  string swapped_op_inst_name> {
+  foreach fvti = AllFloatVectors in {
+    def : Pat<(fvti.Mask (riscv_setcc_vl (fvti.Vector fvti.RegClass:$rs1),
+                                         fvti.RegClass:$rs2,
+                                         cc,
+                                         (fvti.Mask true_mask),
+                                         (XLenVT (VLOp GPR:$vl)))),
+              (!cast<Instruction>(inst_name#"_VV_"#fvti.LMul.MX)
+                  fvti.RegClass:$rs1, fvti.RegClass:$rs2, GPR:$vl, fvti.SEW)>;
+    def : Pat<(fvti.Mask (riscv_setcc_vl (fvti.Vector fvti.RegClass:$rs1),
+                                         (SplatFPOp fvti.ScalarRegClass:$rs2),
+                                         cc,
+                                         (fvti.Mask true_mask),
+                                         (XLenVT (VLOp GPR:$vl)))),
+              (!cast<Instruction>(inst_name#"_V"#fvti.ScalarSuffix#"_"#fvti.LMul.MX)
+                  fvti.RegClass:$rs1, fvti.ScalarRegClass:$rs2,
+                  GPR:$vl, fvti.SEW)>;
+    def : Pat<(fvti.Mask (riscv_setcc_vl (SplatFPOp fvti.ScalarRegClass:$rs2),
+                                         (fvti.Vector fvti.RegClass:$rs1),
+                                         cc,
+                                         (fvti.Mask true_mask),
+                                         (XLenVT (VLOp GPR:$vl)))),
+              (!cast<Instruction>(swapped_op_inst_name#"_V"#fvti.ScalarSuffix#"_"#fvti.LMul.MX)
+                  fvti.RegClass:$rs1, fvti.ScalarRegClass:$rs2,
+                  GPR:$vl, fvti.SEW)>;
+  }
+}
+
 //===----------------------------------------------------------------------===//
 // Patterns.
 //===----------------------------------------------------------------------===//
 
 let Predicates = [HasStdExtV] in {
 
 // 7.4. Vector Unit-Stride Instructions
 foreach vti = AllVectors in {
def : Pat<(vti.Vector (riscv_fma_vl (SplatFPOp vti.ScalarRegClass:$rs1),
                                        vti.RegClass:$rd, vti.RegClass:$rs2,
                                        (vti.Mask true_mask),
                                        (XLenVT (VLOp GPR:$vl)))),
             (!cast<Instruction>("PseudoVFMADD_V" # vti.ScalarSuffix # "_" # suffix)
                  vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
                  GPR:$vl, vti.SEW)>;
 }
 
+// 14.11. Vector Floating-Point Compare Instructions
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETEQ,  "PseudoVMFEQ", "PseudoVMFEQ">;
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETOEQ, "PseudoVMFEQ", "PseudoVMFEQ">;
+
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETNE,  "PseudoVMFNE", "PseudoVMFNE">;
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETUNE, "PseudoVMFNE", "PseudoVMFNE">;
+
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETLT,  "PseudoVMFLT", "PseudoVMFGT">;
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETOLT, "PseudoVMFLT", "PseudoVMFGT">;
+
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETLE,  "PseudoVMFLE", "PseudoVMFGE">;
+defm "" : VPatFPSetCCVL_VV_VF_FV<SETOLE, "PseudoVMFLE", "PseudoVMFGE">;
+
+// 14.12. Vector Floating-Point Sign-Injection Instructions
+// Handle fneg with VFSGNJN using the same input for both operands.
 foreach vti = AllFloatVectors in {
   // 14.8. Vector Floating-Point Square-Root Instruction
   def : Pat<(riscv_fsqrt_vl (vti.Vector vti.RegClass:$rs2), (vti.Mask true_mask),
                             (XLenVT (VLOp GPR:$vl))),
             (!cast<Instruction>("PseudoVFSQRT_V_"# vti.LMul.MX)
                  vti.RegClass:$rs2, GPR:$vl, vti.SEW)>;
 
   // 14.12. Vector Floating-Point Sign-Injection Instructions

 // 16.1 Vector Mask-Register Logical Instructions
 let Predicates = [HasStdExtV] in {
 
 foreach mti = AllMasks in {
   def : Pat<(mti.Mask (riscv_vmset_vl (XLenVT (VLOp GPR:$vl)))),
             (!cast<Instruction>("PseudoVMSET_M_" # mti.BX) GPR:$vl, mti.SEW)>;
   def : Pat<(mti.Mask (riscv_vmclr_vl (XLenVT (VLOp GPR:$vl)))),
             (!cast<Instruction>("PseudoVMCLR_M_" # mti.BX) GPR:$vl, mti.SEW)>;
+
+  def : Pat<(mti.Mask (riscv_vmand_vl VR:$rs1, VR:$rs2, (XLenVT (VLOp GPR:$vl)))),
+            (!cast<Instruction>("PseudoVMAND_MM_" # mti.LMul.MX)
+                 VR:$rs1, VR:$rs2, GPR:$vl, mti.SEW)>;
+  def : Pat<(mti.Mask (riscv_vmor_vl VR:$rs1, VR:$rs2, (XLenVT (VLOp GPR:$vl)))),
+            (!cast<Instruction>("PseudoVMOR_MM_" # mti.LMul.MX)
+                 VR:$rs1, VR:$rs2, GPR:$vl, mti.SEW)>;
+  def : Pat<(mti.Mask (riscv_vmxor_vl VR:$rs1, VR:$rs2, (XLenVT (VLOp GPR:$vl)))),
+            (!cast<Instruction>("PseudoVMXOR_MM_" # mti.LMul.MX)
+                 VR:$rs1, VR:$rs2, GPR:$vl, mti.SEW)>;
+
+  // FIXME: Add remaining mask instructions.
+  def : Pat<(mti.Mask (riscv_vmxor_vl (riscv_vmor_vl VR:$rs1, VR:$rs2,
+                                                     (XLenVT (VLOp GPR:$vl))),
+                                      true_mask, (XLenVT (VLOp GPR:$vl)))),
+            (!cast<Instruction>("PseudoVMNOR_MM_" # mti.LMul.MX)
+                 VR:$rs1, VR:$rs2, GPR:$vl, mti.SEW)>;
+
+  // Match the not idiom to the vnot.mm pseudo.
+  def : Pat<(mti.Mask (riscv_vmxor_vl VR:$rs, true_mask, (XLenVT (VLOp GPR:$vl)))),
+            (!cast<Instruction>("PseudoVMNAND_MM_" # mti.LMul.MX)
+                 VR:$rs, VR:$rs, GPR:$vl, mti.SEW)>;
 }
 
 } // Predicates = [HasStdExtV]
 
 // 17.4. Vector Register GAther Instruction
 let Predicates = [HasStdExtV] in {
 
 foreach vti = AllIntegerVectors in {

 
 } // Predicates = [HasStdExtV, HasStdExtF]
 
 //===----------------------------------------------------------------------===//
 // Miscellaneous RISCVISD SDNodes
 //===----------------------------------------------------------------------===//
 
 def riscv_vid_vl : SDNode<"RISCVISD::VID_VL", SDTypeProfile<1, 2,
-                          [SDTCisVec<0>, SDTCisVec<1>, SDTCVecEltisVT<1, i1>,
+                          [SDTCisVec<0>, SDTCVecEltisVT<1, i1>,
                            SDTCisSameNumEltsAs<0, 1>, SDTCisVT<2, XLenVT>]>, []>;
 
 def SDTRVVSlide : SDTypeProfile<1, 5, [
   SDTCisVec<0>, SDTCisSameAs<1, 0>, SDTCisSameAs<2, 0>, SDTCisVT<3, XLenVT>,
   SDTCVecEltisVT<4, i1>, SDTCisSameNumEltsAs<0, 4>, SDTCisVT<5, XLenVT>
 ]>;
 
 def riscv_slideup_vl   : SDNode<"RISCVISD::VSLIDEUP_VL", SDTRVVSlide, []>;

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fp-setcc.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -target-abi=ilp32d -mattr=+experimental-v,+experimental-zfh,+f,+d -verify-machineinstrs -riscv-v-vector-bits-min=128 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK
+; RUN: llc -mtriple=riscv64 -target-abi=lp64d -mattr=+experimental-v,+experimental-zfh,+f,+d -verify-machineinstrs -riscv-v-vector-bits-min=128 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK
+
+define void @fcmp_oeq_vv_v8f16(<8 x half>* %x, <8 x half>* %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oeq_vv_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 8
+; CHECK-NEXT:    vsetvli a4, a3, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vmfeq.vv v27, v25, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v27, (a2)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = load <8 x half>, <8 x half>* %y
+  %c = fcmp oeq <8 x half> %a, %b
+  store <8 x i1> %c, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oeq_vv_v8f16_nonans(<8 x half>* %x, <8 x half>* %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oeq_vv_v8f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 8
+; CHECK-NEXT:    vsetvli a4, a3, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vle16.v v26, (a1)
+; CHECK-NEXT:    vmfeq.vv v27, v25, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v27, (a2)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = load <8 x half>, <8 x half>* %y
+  %c = fcmp nnan oeq <8 x half> %a, %b
+  store <8 x i1> %c, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_une_vv_v4f32(<4 x float>* %x, <4 x float>* %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_une_vv_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 4
+; CHECK-NEXT:    vsetvli a4, a3, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vmfne.vv v27, v25, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v27, (a2)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = load <4 x float>, <4 x float>* %y
+  %c = fcmp une <4 x float> %a, %b
+  store <4 x i1> %c, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_une_vv_v4f32_nonans(<4 x float>* %x, <4 x float>* %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_une_vv_v4f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 4
+; CHECK-NEXT:    vsetvli a4, a3, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vle32.v v26, (a1)
+; CHECK-NEXT:    vmfne.vv v27, v25, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v27, (a2)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = load <4 x float>, <4 x float>* %y
+  %c = fcmp nnan une <4 x float> %a, %b
+  store <4 x i1> %c, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ogt_vv_v2f64(<2 x double>* %x, <2 x double>* %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_ogt_vv_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 2
+; CHECK-NEXT:    vsetvli a4, a3, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vmflt.vv v27, v26, v25
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v27, (a2)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = load <2 x double>, <2 x double>* %y
+  %c = fcmp ogt <2 x double> %a, %b
+  store <2 x i1> %c, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ogt_vv_v2f64_nonans(<2 x double>* %x, <2 x double>* %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_ogt_vv_v2f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 2
+; CHECK-NEXT:    vsetvli a4, a3, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vle64.v v26, (a1)
+; CHECK-NEXT:    vmflt.vv v27, v26, v25
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v27, (a2)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = load <2 x double>, <2 x double>* %y
+  %c = fcmp nnan ogt <2 x double> %a, %b
+  store <2 x i1> %c, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_olt_vv_v16f16(<16 x half>* %x, <16 x half>* %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_olt_vv_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 16
+; CHECK-NEXT:    vsetvli a4, a3, e16,m2,ta,mu
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v26, v28
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = load <16 x half>, <16 x half>* %y
+  %c = fcmp olt <16 x half> %a, %b
+  store <16 x i1> %c, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_olt_vv_v16f16_nonans(<16 x half>* %x, <16 x half>* %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_olt_vv_v16f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 16
+; CHECK-NEXT:    vsetvli a4, a3, e16,m2,ta,mu
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v26, v28
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = load <16 x half>, <16 x half>* %y
+  %c = fcmp nnan olt <16 x half> %a, %b
+  store <16 x i1> %c, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oge_vv_v8f32(<8 x float>* %x, <8 x float>* %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oge_vv_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 8
+; CHECK-NEXT:    vsetvli a4, a3, e32,m2,ta,mu
+; CHECK-NEXT:    vle32.v v26, (a0)
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v28, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = load <8 x float>, <8 x float>* %y
+  %c = fcmp oge <8 x float> %a, %b
+  store <8 x i1> %c, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oge_vv_v8f32_nonans(<8 x float>* %x, <8 x float>* %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oge_vv_v8f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 8
+; CHECK-NEXT:    vsetvli a4, a3, e32,m2,ta,mu
+; CHECK-NEXT:    vle32.v v26, (a0)
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v28, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = load <8 x float>, <8 x float>* %y
+  %c = fcmp nnan oge <8 x float> %a, %b
+  store <8 x i1> %c, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ole_vv_v4f64(<4 x double>* %x, <4 x double>* %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ole_vv_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 4
+; CHECK-NEXT:    vsetvli a4, a3, e64,m2,ta,mu
+; CHECK-NEXT:    vle64.v v26, (a0)
+; CHECK-NEXT:    vle64.v v28, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v26, v28
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = load <4 x double>, <4 x double>* %y
+  %c = fcmp ole <4 x double> %a, %b
+  store <4 x i1> %c, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ole_vv_v4f64_nonans(<4 x double>* %x, <4 x double>* %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ole_vv_v4f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 4
+; CHECK-NEXT:    vsetvli a4, a3, e64,m2,ta,mu
+; CHECK-NEXT:    vle64.v v26, (a0)
+; CHECK-NEXT:    vle64.v v28, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v26, v28
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = load <4 x double>, <4 x double>* %y
+  %c = fcmp nnan ole <4 x double> %a, %b
+  store <4 x i1> %c, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ule_vv_v32f16(<32 x half>* %x, <32 x half>* %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ule_vv_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 32
+; CHECK-NEXT:    vsetvli a4, a3, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a0)
+; CHECK-NEXT:    vle16.v v8, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v8, v28
+; CHECK-NEXT:    vsetvli a0, a3, e8,m2,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <32 x half>, <32 x half>* %x
+  %b = load <32 x half>, <32 x half>* %y
+  %c = fcmp ule <32 x half> %a, %b
+  store <32 x i1> %c, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ule_vv_v32f16_nonans(<32 x half>* %x, <32 x half>* %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ule_vv_v32f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 32
+; CHECK-NEXT:    vsetvli a4, a3, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a0)
+; CHECK-NEXT:    vle16.v v8, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v28, v8
+; CHECK-NEXT:    vsetvli a0, a3, e8,m2,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <32 x half>, <32 x half>* %x
+  %b = load <32 x half>, <32 x half>* %y
+  %c = fcmp nnan ule <32 x half> %a, %b
+  store <32 x i1> %c, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uge_vv_v16f32(<16 x float>* %x, <16 x float>* %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_uge_vv_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 16
+; CHECK-NEXT:    vsetvli a4, a3, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a0)
+; CHECK-NEXT:    vle32.v v8, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v28, v8
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <16 x float>, <16 x float>* %x
+  %b = load <16 x float>, <16 x float>* %y
+  %c = fcmp uge <16 x float> %a, %b
+  store <16 x i1> %c, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uge_vv_v16f32_nonans(<16 x float>* %x, <16 x float>* %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_uge_vv_v16f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 16
+; CHECK-NEXT:    vsetvli a4, a3, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a0)
+; CHECK-NEXT:    vle32.v v8, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v8, v28
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <16 x float>, <16 x float>* %x
+  %b = load <16 x float>, <16 x float>* %y
+  %c = fcmp nnan uge <16 x float> %a, %b
+  store <16 x i1> %c, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ult_vv_v8f64(<8 x double>* %x, <8 x double>* %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_ult_vv_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 8
+; CHECK-NEXT:    vsetvli a4, a3, e64,m4,ta,mu
+; CHECK-NEXT:    vle64.v v28, (a0)
+; CHECK-NEXT:    vle64.v v8, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v8, v28
+; CHECK-NEXT:    vsetvli a0, a3, e8,mf2,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <8 x double>, <8 x double>* %x
+  %b = load <8 x double>, <8 x double>* %y
+  %c = fcmp ult <8 x double> %a, %b
+  store <8 x i1> %c, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ult_vv_v8f64_nonans(<8 x double>* %x, <8 x double>* %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_ult_vv_v8f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 8
+; CHECK-NEXT:    vsetvli a4, a3, e64,m4,ta,mu
+; CHECK-NEXT:    vle64.v v28, (a0)
+; CHECK-NEXT:    vle64.v v8, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v28, v8
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <8 x double>, <8 x double>* %x
+  %b = load <8 x double>, <8 x double>* %y
+  %c = fcmp nnan ult <8 x double> %a, %b
+  store <8 x i1> %c, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ugt_vv_v64f16(<64 x half>* %x, <64 x half>* %y, <64 x i1>* %z) {
+; CHECK-LABEL: fcmp_ugt_vv_v64f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 64
+; CHECK-NEXT:    vsetvli a4, a3, e16,m8,ta,mu
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vle16.v v16, (a1)
+; CHECK-NEXT:    vmfle.vv v25, v8, v16
+; CHECK-NEXT:    vsetvli a0, a3, e8,m4,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <64 x half>, <64 x half>* %x
+  %b = load <64 x half>, <64 x half>* %y
+  %c = fcmp ugt <64 x half> %a, %b
+  store <64 x i1> %c, <64 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ugt_vv_v64f16_nonans(<64 x half>* %x, <64 x half>* %y, <64 x i1>* %z) {
+; CHECK-LABEL: fcmp_ugt_vv_v64f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 64
+; CHECK-NEXT:    vsetvli a4, a3, e16,m8,ta,mu
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vle16.v v16, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v16, v8
+; CHECK-NEXT:    vsetvli a0, a3, e8,m4,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <64 x half>, <64 x half>* %x
+  %b = load <64 x half>, <64 x half>* %y
+  %c = fcmp nnan ugt <64 x half> %a, %b
+  store <64 x i1> %c, <64 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ueq_vv_v32f32(<32 x float>* %x, <32 x float>* %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ueq_vv_v32f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 32
+; CHECK-NEXT:    vsetvli a4, a3, e32,m8,ta,mu
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vle32.v v16, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v16, v8
+; CHECK-NEXT:    vmflt.vv v26, v8, v16
+; CHECK-NEXT:    vsetvli a0, a3, e8,m2,ta,mu
+; CHECK-NEXT:    vmnor.mm v25, v26, v25
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <32 x float>, <32 x float>* %x
+  %b = load <32 x float>, <32 x float>* %y
+  %c = fcmp ueq <32 x float> %a, %b
+  store <32 x i1> %c, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ueq_vv_v32f32_nonans(<32 x float>* %x, <32 x float>* %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ueq_vv_v32f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 32
+; CHECK-NEXT:    vsetvli a4, a3, e32,m8,ta,mu
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vle32.v v16, (a1)
+; CHECK-NEXT:    vmfeq.vv v25, v8, v16
+; CHECK-NEXT:    vsetvli a0, a3, e8,m2,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <32 x float>, <32 x float>* %x
+  %b = load <32 x float>, <32 x float>* %y
+  %c = fcmp nnan ueq <32 x float> %a, %b
+  store <32 x i1> %c, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_one_vv_v8f64(<16 x double>* %x, <16 x double>* %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_one_vv_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 16
+; CHECK-NEXT:    vsetvli a4, a3, e64,m8,ta,mu
+; CHECK-NEXT:    vle64.v v8, (a0)
+; CHECK-NEXT:    vle64.v v16, (a1)
+; CHECK-NEXT:    vmflt.vv v25, v16, v8
+; CHECK-NEXT:    vmflt.vv v26, v8, v16
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vmor.mm v25, v26, v25
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <16 x double>, <16 x double>* %x
+  %b = load <16 x double>, <16 x double>* %y
+  %c = fcmp one <16 x double> %a, %b
+  store <16 x i1> %c, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_one_vv_v8f64_nonans(<16 x double>* %x, <16 x double>* %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_one_vv_v8f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 16
+; CHECK-NEXT:    vsetvli a4, a3, e64,m8,ta,mu
+; CHECK-NEXT:    vle64.v v8, (a0)
+; CHECK-NEXT:    vle64.v v16, (a1)
+; CHECK-NEXT:    vmfne.vv v25, v8, v16
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <16 x double>, <16 x double>* %x
+  %b = load <16 x double>, <16 x double>* %y
+  %c = fcmp nnan one <16 x double> %a, %b
+  store <16 x i1> %c, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ord_vv_v4f16(<4 x half>* %x, <4 x half>* %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ord_vv_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 4
+; CHECK-NEXT:    vsetvli a4, a3, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a1)
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vmfeq.vv v27, v25, v25
+; CHECK-NEXT:    vmfeq.vv v25, v26, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,mf2,ta,mu
+; CHECK-NEXT:    vmand.mm v25, v25, v27
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <4 x half>, <4 x half>* %x
+  %b = load <4 x half>, <4 x half>* %y
+  %c = fcmp ord <4 x half> %a, %b
+  store <4 x i1> %c, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uno_vv_v4f16(<2 x half>* %x, <2 x half>* %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_uno_vv_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a3, zero, 2
+; CHECK-NEXT:    vsetvli a4, a3, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a1)
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vmfne.vv v27, v25, v25
+; CHECK-NEXT:    vmfne.vv v25, v26, v26
+; CHECK-NEXT:    vsetvli a0, a3, e8,mf2,ta,mu
+; CHECK-NEXT:    vmor.mm v25, v25, v27
+; CHECK-NEXT:    vsetvli a0, a3, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a2)
+; CHECK-NEXT:    ret
+  %a = load <2 x half>, <2 x half>* %x
+  %b = load <2 x half>, <2 x half>* %y
+  %c = fcmp uno <2 x half> %a, %b
+  store <2 x i1> %c, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oeq_vf_v8f16(<8 x half>* %x, half %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oeq_vf_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vmfeq.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = insertelement <8 x half> undef, half %y, i32 0
+  %c = shufflevector <8 x half> %b, <8 x half> undef, <8 x i32> zeroinitializer
+  %d = fcmp oeq <8 x half> %a, %c
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oeq_vf_v8f16_nonans(<8 x half>* %x, half %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oeq_vf_v8f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vmfeq.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = insertelement <8 x half> undef, half %y, i32 0
+  %c = shufflevector <8 x half> %b, <8 x half> undef, <8 x i32> zeroinitializer
+  %d = fcmp nnan oeq <8 x half> %a, %c
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_une_vf_v4f32(<4 x float>* %x, float %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_une_vf_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vmfne.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = insertelement <4 x float> undef, float %y, i32 0
+  %c = shufflevector <4 x float> %b, <4 x float> undef, <4 x i32> zeroinitializer
+  %d = fcmp une <4 x float> %a, %c
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_une_vf_v4f32_nonans(<4 x float>* %x, float %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_une_vf_v4f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vmfne.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = insertelement <4 x float> undef, float %y, i32 0
+  %c = shufflevector <4 x float> %b, <4 x float> undef, <4 x i32> zeroinitializer
+  %d = fcmp nnan une <4 x float> %a, %c
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ogt_vf_v2f64(<2 x double>* %x, double %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_ogt_vf_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a3, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vmfgt.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = insertelement <2 x double> undef, double %y, i32 0
+  %c = shufflevector <2 x double> %b, <2 x double> undef, <2 x i32> zeroinitializer
+  %d = fcmp ogt <2 x double> %a, %c
+  store <2 x i1> %d, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ogt_vf_v2f64_nonans(<2 x double>* %x, double %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_ogt_vf_v2f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a3, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vmfgt.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = insertelement <2 x double> undef, double %y, i32 0
+  %c = shufflevector <2 x double> %b, <2 x double> undef, <2 x i32> zeroinitializer
+  %d = fcmp nnan ogt <2 x double> %a, %c
+  store <2 x i1> %d, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_olt_vf_v16f16(<16 x half>* %x, half %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_olt_vf_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e16,m2,ta,mu
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = insertelement <16 x half> undef, half %y, i32 0
+  %c = shufflevector <16 x half> %b, <16 x half> undef, <16 x i32> zeroinitializer
+  %d = fcmp olt <16 x half> %a, %c
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_olt_vf_v16f16_nonans(<16 x half>* %x, half %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_olt_vf_v16f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e16,m2,ta,mu
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = insertelement <16 x half> undef, half %y, i32 0
+  %c = shufflevector <16 x half> %b, <16 x half> undef, <16 x i32> zeroinitializer
+  %d = fcmp nnan olt <16 x half> %a, %c
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oge_vf_v8f32(<8 x float>* %x, float %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oge_vf_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e32,m2,ta,mu
+; CHECK-NEXT:    vle32.v v26, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = insertelement <8 x float> undef, float %y, i32 0
+  %c = shufflevector <8 x float> %b, <8 x float> undef, <8 x i32> zeroinitializer
+  %d = fcmp oge <8 x float> %a, %c
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oge_vf_v8f32_nonans(<8 x float>* %x, float %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oge_vf_v8f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e32,m2,ta,mu
+; CHECK-NEXT:    vle32.v v26, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = insertelement <8 x float> undef, float %y, i32 0
+  %c = shufflevector <8 x float> %b, <8 x float> undef, <8 x i32> zeroinitializer
+  %d = fcmp nnan oge <8 x float> %a, %c
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ole_vf_v4f64(<4 x double>* %x, double %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ole_vf_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e64,m2,ta,mu
+; CHECK-NEXT:    vle64.v v26, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = insertelement <4 x double> undef, double %y, i32 0
+  %c = shufflevector <4 x double> %b, <4 x double> undef, <4 x i32> zeroinitializer
+  %d = fcmp ole <4 x double> %a, %c
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ole_vf_v4f64_nonans(<4 x double>* %x, double %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ole_vf_v4f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e64,m2,ta,mu
+; CHECK-NEXT:    vle64.v v26, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = insertelement <4 x double> undef, double %y, i32 0
+  %c = shufflevector <4 x double> %b, <4 x double> undef, <4 x i32> zeroinitializer
+  %d = fcmp nnan ole <4 x double> %a, %c
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ule_vf_v32f16(<32 x half>* %x, half %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ule_vf_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x half>, <32 x half>* %x
+  %b = insertelement <32 x half> undef, half %y, i32 0
+  %c = shufflevector <32 x half> %b, <32 x half> undef, <32 x i32> zeroinitializer
+  %d = fcmp ule <32 x half> %a, %c
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ule_vf_v32f16_nonans(<32 x half>* %x, half %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ule_vf_v32f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x half>, <32 x half>* %x
+  %b = insertelement <32 x half> undef, half %y, i32 0
+  %c = shufflevector <32 x half> %b, <32 x half> undef, <32 x i32> zeroinitializer
+  %d = fcmp nnan ule <32 x half> %a, %c
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uge_vf_v16f32(<16 x float>* %x, float %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_uge_vf_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x float>, <16 x float>* %x
+  %b = insertelement <16 x float> undef, float %y, i32 0
+  %c = shufflevector <16 x float> %b, <16 x float> undef, <16 x i32> zeroinitializer
+  %d = fcmp uge <16 x float> %a, %c
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uge_vf_v16f32_nonans(<16 x float>* %x, float %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_uge_vf_v16f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x float>, <16 x float>* %x
+  %b = insertelement <16 x float> undef, float %y, i32 0
+  %c = shufflevector <16 x float> %b, <16 x float> undef, <16 x i32> zeroinitializer
+  %d = fcmp nnan uge <16 x float> %a, %c
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ult_vf_v8f64(<8 x double>* %x, double %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_ult_vf_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e64,m4,ta,mu
+; CHECK-NEXT:    vle64.v v28, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,mf2,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x double>, <8 x double>* %x
+  %b = insertelement <8 x double> undef, double %y, i32 0
+  %c = shufflevector <8 x double> %b, <8 x double> undef, <8 x i32> zeroinitializer
+  %d = fcmp ult <8 x double> %a, %c
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ult_vf_v8f64_nonans(<8 x double>* %x, double %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_ult_vf_v8f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e64,m4,ta,mu
+; CHECK-NEXT:    vle64.v v28, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x double>, <8 x double>* %x
+  %b = insertelement <8 x double> undef, double %y, i32 0
+  %c = shufflevector <8 x double> %b, <8 x double> undef, <8 x i32> zeroinitializer
+  %d = fcmp nnan ult <8 x double> %a, %c
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ugt_vf_v64f16(<64 x half>* %x, half %y, <64 x i1>* %z) {
+; CHECK-LABEL: fcmp_ugt_vf_v64f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 64
+; CHECK-NEXT:    vsetvli a3, a2, e16,m8,ta,mu
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m4,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <64 x half>, <64 x half>* %x
+  %b = insertelement <64 x half> undef, half %y, i32 0
+  %c = shufflevector <64 x half> %b, <64 x half> undef, <64 x i32> zeroinitializer
+  %d = fcmp ugt <64 x half> %a, %c
+  store <64 x i1> %d, <64 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ugt_vf_v64f16_nonans(<64 x half>* %x, half %y, <64 x i1>* %z) {
+; CHECK-LABEL: fcmp_ugt_vf_v64f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 64
+; CHECK-NEXT:    vsetvli a3, a2, e16,m8,ta,mu
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m4,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <64 x half>, <64 x half>* %x
+  %b = insertelement <64 x half> undef, half %y, i32 0
+  %c = shufflevector <64 x half> %b, <64 x half> undef, <64 x i32> zeroinitializer
+  %d = fcmp nnan ugt <64 x half> %a, %c
+  store <64 x i1> %d, <64 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ueq_vf_v32f32(<32 x float>* %x, float %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ueq_vf_v32f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e32,m8,ta,mu
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v8, fa0
+; CHECK-NEXT:    vmflt.vf v26, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vmnor.mm v25, v26, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x float>, <32 x float>* %x
+  %b = insertelement <32 x float> undef, float %y, i32 0
+  %c = shufflevector <32 x float> %b, <32 x float> undef, <32 x i32> zeroinitializer
+  %d = fcmp ueq <32 x float> %a, %c
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ueq_vf_v32f32_nonans(<32 x float>* %x, float %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ueq_vf_v32f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e32,m8,ta,mu
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vmfeq.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x float>, <32 x float>* %x
+  %b = insertelement <32 x float> undef, float %y, i32 0
+  %c = shufflevector <32 x float> %b, <32 x float> undef, <32 x i32> zeroinitializer
+  %d = fcmp nnan ueq <32 x float> %a, %c
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_one_vf_v8f64(<16 x double>* %x, double %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_one_vf_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e64,m8,ta,mu
+; CHECK-NEXT:    vle64.v v8, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v8, fa0
+; CHECK-NEXT:    vmflt.vf v26, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vmor.mm v25, v26, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x double>, <16 x double>* %x
+  %b = insertelement <16 x double> undef, double %y, i32 0
+  %c = shufflevector <16 x double> %b, <16 x double> undef, <16 x i32> zeroinitializer
+  %d = fcmp one <16 x double> %a, %c
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_one_vf_v8f64_nonans(<16 x double>* %x, double %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_one_vf_v8f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e64,m8,ta,mu
+; CHECK-NEXT:    vle64.v v8, (a0)
+; CHECK-NEXT:    vmfne.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x double>, <16 x double>* %x
+  %b = insertelement <16 x double> undef, double %y, i32 0
+  %c = shufflevector <16 x double> %b, <16 x double> undef, <16 x i32> zeroinitializer
+  %d = fcmp nnan one <16 x double> %a, %c
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ord_vf_v4f16(<4 x half>* %x, half %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ord_vf_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vfmv.v.f v26, fa0
+; CHECK-NEXT:    vmfeq.vf v27, v26, fa0
+; CHECK-NEXT:    vmfeq.vv v26, v25, v25
+; CHECK-NEXT:    vsetvli a0, a2, e8,mf2,ta,mu
+; CHECK-NEXT:    vmand.mm v25, v26, v27
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x half>, <4 x half>* %x
+  %b = insertelement <4 x half> undef, half %y, i32 0
+  %c = shufflevector <4 x half> %b, <4 x half> undef, <4 x i32> zeroinitializer
+  %d = fcmp ord <4 x half> %a, %c
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uno_vf_v4f16(<2 x half>* %x, half %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_uno_vf_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vfmv.v.f v26, fa0
+; CHECK-NEXT:    vmfne.vf v27, v26, fa0
+; CHECK-NEXT:    vmfne.vv v26, v25, v25
+; CHECK-NEXT:    vsetvli a0, a2, e8,mf2,ta,mu
+; CHECK-NEXT:    vmor.mm v25, v26, v27
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <2 x half>, <2 x half>* %x
+  %b = insertelement <2 x half> undef, half %y, i32 0
+  %c = shufflevector <2 x half> %b, <2 x half> undef, <2 x i32> zeroinitializer
+  %d = fcmp uno <2 x half> %a, %c
+  store <2 x i1> %d, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oeq_fv_v8f16(<8 x half>* %x, half %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oeq_fv_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vmfeq.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = insertelement <8 x half> undef, half %y, i32 0
+  %c = shufflevector <8 x half> %b, <8 x half> undef, <8 x i32> zeroinitializer
+  %d = fcmp oeq <8 x half> %c, %a
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oeq_fv_v8f16_nonans(<8 x half>* %x, half %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oeq_fv_v8f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vmfeq.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x half>, <8 x half>* %x
+  %b = insertelement <8 x half> undef, half %y, i32 0
+  %c = shufflevector <8 x half> %b, <8 x half> undef, <8 x i32> zeroinitializer
+  %d = fcmp nnan oeq <8 x half> %c, %a
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_une_fv_v4f32(<4 x float>* %x, float %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_une_fv_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vmfne.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = insertelement <4 x float> undef, float %y, i32 0
+  %c = shufflevector <4 x float> %b, <4 x float> undef, <4 x i32> zeroinitializer
+  %d = fcmp une <4 x float> %c, %a
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_une_fv_v4f32_nonans(<4 x float>* %x, float %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_une_fv_v4f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e32,m1,ta,mu
+; CHECK-NEXT:    vle32.v v25, (a0)
+; CHECK-NEXT:    vmfne.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x float>, <4 x float>* %x
+  %b = insertelement <4 x float> undef, float %y, i32 0
+  %c = shufflevector <4 x float> %b, <4 x float> undef, <4 x i32> zeroinitializer
+  %d = fcmp nnan une <4 x float> %c, %a
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ogt_fv_v2f64(<2 x double>* %x, double %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_ogt_fv_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a3, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vmflt.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = insertelement <2 x double> undef, double %y, i32 0
+  %c = shufflevector <2 x double> %b, <2 x double> undef, <2 x i32> zeroinitializer
+  %d = fcmp ogt <2 x double> %c, %a
+  store <2 x i1> %d, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ogt_fv_v2f64_nonans(<2 x double>* %x, double %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_ogt_fv_v2f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a3, a2, e64,m1,ta,mu
+; CHECK-NEXT:    vle64.v v25, (a0)
+; CHECK-NEXT:    vmflt.vf v26, v25, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v26, (a1)
+; CHECK-NEXT:    ret
+  %a = load <2 x double>, <2 x double>* %x
+  %b = insertelement <2 x double> undef, double %y, i32 0
+  %c = shufflevector <2 x double> %b, <2 x double> undef, <2 x i32> zeroinitializer
+  %d = fcmp nnan ogt <2 x double> %c, %a
+  store <2 x i1> %d, <2 x i1>* %z
+  ret void
+}
+
+define void @fcmp_olt_fv_v16f16(<16 x half>* %x, half %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_olt_fv_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e16,m2,ta,mu
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = insertelement <16 x half> undef, half %y, i32 0
+  %c = shufflevector <16 x half> %b, <16 x half> undef, <16 x i32> zeroinitializer
+  %d = fcmp olt <16 x half> %c, %a
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_olt_fv_v16f16_nonans(<16 x half>* %x, half %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_olt_fv_v16f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e16,m2,ta,mu
+; CHECK-NEXT:    vle16.v v26, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x half>, <16 x half>* %x
+  %b = insertelement <16 x half> undef, half %y, i32 0
+  %c = shufflevector <16 x half> %b, <16 x half> undef, <16 x i32> zeroinitializer
+  %d = fcmp nnan olt <16 x half> %c, %a
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oge_fv_v8f32(<8 x float>* %x, float %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oge_fv_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e32,m2,ta,mu
+; CHECK-NEXT:    vle32.v v26, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = insertelement <8 x float> undef, float %y, i32 0
+  %c = shufflevector <8 x float> %b, <8 x float> undef, <8 x i32> zeroinitializer
+  %d = fcmp oge <8 x float> %c, %a
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_oge_fv_v8f32_nonans(<8 x float>* %x, float %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_oge_fv_v8f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e32,m2,ta,mu
+; CHECK-NEXT:    vle32.v v26, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x float>, <8 x float>* %x
+  %b = insertelement <8 x float> undef, float %y, i32 0
+  %c = shufflevector <8 x float> %b, <8 x float> undef, <8 x i32> zeroinitializer
+  %d = fcmp nnan oge <8 x float> %c, %a
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ole_fv_v4f64(<4 x double>* %x, double %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ole_fv_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e64,m2,ta,mu
+; CHECK-NEXT:    vle64.v v26, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = insertelement <4 x double> undef, double %y, i32 0
+  %c = shufflevector <4 x double> %b, <4 x double> undef, <4 x i32> zeroinitializer
+  %d = fcmp ole <4 x double> %c, %a
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ole_fv_v4f64_nonans(<4 x double>* %x, double %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ole_fv_v4f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e64,m2,ta,mu
+; CHECK-NEXT:    vle64.v v26, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v26, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x double>, <4 x double>* %x
+  %b = insertelement <4 x double> undef, double %y, i32 0
+  %c = shufflevector <4 x double> %b, <4 x double> undef, <4 x i32> zeroinitializer
+  %d = fcmp nnan ole <4 x double> %c, %a
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ule_fv_v32f16(<32 x half>* %x, half %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ule_fv_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x half>, <32 x half>* %x
+  %b = insertelement <32 x half> undef, half %y, i32 0
+  %c = shufflevector <32 x half> %b, <32 x half> undef, <32 x i32> zeroinitializer
+  %d = fcmp ule <32 x half> %c, %a
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ule_fv_v32f16_nonans(<32 x half>* %x, half %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ule_fv_v32f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x half>, <32 x half>* %x
+  %b = insertelement <32 x half> undef, half %y, i32 0
+  %c = shufflevector <32 x half> %b, <32 x half> undef, <32 x i32> zeroinitializer
+  %d = fcmp nnan ule <32 x half> %c, %a
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uge_fv_v16f32(<16 x float>* %x, float %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_uge_fv_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x float>, <16 x float>* %x
+  %b = insertelement <16 x float> undef, float %y, i32 0
+  %c = shufflevector <16 x float> %b, <16 x float> undef, <16 x i32> zeroinitializer
+  %d = fcmp uge <16 x float> %c, %a
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uge_fv_v16f32_nonans(<16 x float>* %x, float %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_uge_fv_v16f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x float>, <16 x float>* %x
+  %b = insertelement <16 x float> undef, float %y, i32 0
+  %c = shufflevector <16 x float> %b, <16 x float> undef, <16 x i32> zeroinitializer
+  %d = fcmp nnan uge <16 x float> %c, %a
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ult_fv_v8f64(<8 x double>* %x, double %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_ult_fv_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e64,m4,ta,mu
+; CHECK-NEXT:    vle64.v v28, (a0)
+; CHECK-NEXT:    vmfle.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,mf2,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x double>, <8 x double>* %x
+  %b = insertelement <8 x double> undef, double %y, i32 0
+  %c = shufflevector <8 x double> %b, <8 x double> undef, <8 x i32> zeroinitializer
+  %d = fcmp ult <8 x double> %c, %a
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ult_fv_v8f64_nonans(<8 x double>* %x, double %y, <8 x i1>* %z) {
+; CHECK-LABEL: fcmp_ult_fv_v8f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 8
+; CHECK-NEXT:    vsetvli a3, a2, e64,m4,ta,mu
+; CHECK-NEXT:    vle64.v v28, (a0)
+; CHECK-NEXT:    vmfgt.vf v25, v28, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <8 x double>, <8 x double>* %x
+  %b = insertelement <8 x double> undef, double %y, i32 0
+  %c = shufflevector <8 x double> %b, <8 x double> undef, <8 x i32> zeroinitializer
+  %d = fcmp nnan ult <8 x double> %c, %a
+  store <8 x i1> %d, <8 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ugt_fv_v64f16(<64 x half>* %x, half %y, <64 x i1>* %z) {
+; CHECK-LABEL: fcmp_ugt_fv_v64f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 64
+; CHECK-NEXT:    vsetvli a3, a2, e16,m8,ta,mu
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vmfge.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m4,ta,mu
+; CHECK-NEXT:    vmnand.mm v25, v25, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <64 x half>, <64 x half>* %x
+  %b = insertelement <64 x half> undef, half %y, i32 0
+  %c = shufflevector <64 x half> %b, <64 x half> undef, <64 x i32> zeroinitializer
+  %d = fcmp ugt <64 x half> %c, %a
+  store <64 x i1> %d, <64 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ugt_fv_v64f16_nonans(<64 x half>* %x, half %y, <64 x i1>* %z) {
+; CHECK-LABEL: fcmp_ugt_fv_v64f16_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 64
+; CHECK-NEXT:    vsetvli a3, a2, e16,m8,ta,mu
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m4,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <64 x half>, <64 x half>* %x
+  %b = insertelement <64 x half> undef, half %y, i32 0
+  %c = shufflevector <64 x half> %b, <64 x half> undef, <64 x i32> zeroinitializer
+  %d = fcmp nnan ugt <64 x half> %c, %a
+  store <64 x i1> %d, <64 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ueq_fv_v32f32(<32 x float>* %x, float %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ueq_fv_v32f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e32,m8,ta,mu
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v8, fa0
+; CHECK-NEXT:    vmfgt.vf v26, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vmnor.mm v25, v26, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x float>, <32 x float>* %x
+  %b = insertelement <32 x float> undef, float %y, i32 0
+  %c = shufflevector <32 x float> %b, <32 x float> undef, <32 x i32> zeroinitializer
+  %d = fcmp ueq <32 x float> %c, %a
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ueq_fv_v32f32_nonans(<32 x float>* %x, float %y, <32 x i1>* %z) {
+; CHECK-LABEL: fcmp_ueq_fv_v32f32_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 32
+; CHECK-NEXT:    vsetvli a3, a2, e32,m8,ta,mu
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vmfeq.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m2,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <32 x float>, <32 x float>* %x
+  %b = insertelement <32 x float> undef, float %y, i32 0
+  %c = shufflevector <32 x float> %b, <32 x float> undef, <32 x i32> zeroinitializer
+  %d = fcmp nnan ueq <32 x float> %c, %a
+  store <32 x i1> %d, <32 x i1>* %z
+  ret void
+}
+
+define void @fcmp_one_fv_v8f64(<16 x double>* %x, double %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_one_fv_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e64,m8,ta,mu
+; CHECK-NEXT:    vle64.v v8, (a0)
+; CHECK-NEXT:    vmflt.vf v25, v8, fa0
+; CHECK-NEXT:    vmfgt.vf v26, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vmor.mm v25, v26, v25
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x double>, <16 x double>* %x
+  %b = insertelement <16 x double> undef, double %y, i32 0
+  %c = shufflevector <16 x double> %b, <16 x double> undef, <16 x i32> zeroinitializer
+  %d = fcmp one <16 x double> %c, %a
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_one_fv_v8f64_nonans(<16 x double>* %x, double %y, <16 x i1>* %z) {
+; CHECK-LABEL: fcmp_one_fv_v8f64_nonans:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 16
+; CHECK-NEXT:    vsetvli a3, a2, e64,m8,ta,mu
+; CHECK-NEXT:    vle64.v v8, (a0)
+; CHECK-NEXT:    vmfne.vf v25, v8, fa0
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <16 x double>, <16 x double>* %x
+  %b = insertelement <16 x double> undef, double %y, i32 0
+  %c = shufflevector <16 x double> %b, <16 x double> undef, <16 x i32> zeroinitializer
+  %d = fcmp nnan one <16 x double> %c, %a
+  store <16 x i1> %d, <16 x i1>* %z
+  ret void
+}
+
+define void @fcmp_ord_fv_v4f16(<4 x half>* %x, half %y, <4 x i1>* %z) {
+; CHECK-LABEL: fcmp_ord_fv_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 4
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vfmv.v.f v26, fa0
+; CHECK-NEXT:    vmfeq.vf v27, v26, fa0
+; CHECK-NEXT:    vmfeq.vv v26, v25, v25
+; CHECK-NEXT:    vsetvli a0, a2, e8,mf2,ta,mu
+; CHECK-NEXT:    vmand.mm v25, v27, v26
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <4 x half>, <4 x half>* %x
+  %b = insertelement <4 x half> undef, half %y, i32 0
+  %c = shufflevector <4 x half> %b, <4 x half> undef, <4 x i32> zeroinitializer
+  %d = fcmp ord <4 x half> %c, %a
+  store <4 x i1> %d, <4 x i1>* %z
+  ret void
+}
+
+define void @fcmp_uno_fv_v4f16(<2 x half>* %x, half %y, <2 x i1>* %z) {
+; CHECK-LABEL: fcmp_uno_fv_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    addi a2, zero, 2
+; CHECK-NEXT:    vsetvli a3, a2, e16,m1,ta,mu
+; CHECK-NEXT:    vle16.v v25, (a0)
+; CHECK-NEXT:    vfmv.v.f v26, fa0
+; CHECK-NEXT:    vmfne.vf v27, v26, fa0
+; CHECK-NEXT:    vmfne.vv v26, v25, v25
+; CHECK-NEXT:    vsetvli a0, a2, e8,mf2,ta,mu
+; CHECK-NEXT:    vmor.mm v25, v27, v26
+; CHECK-NEXT:    vsetvli a0, a2, e8,m1,ta,mu
+; CHECK-NEXT:    vse1.v v25, (a1)
+; CHECK-NEXT:    ret
+  %a = load <2 x half>, <2 x half>* %x
+  %b = insertelement <2 x half> undef, half %y, i32 0
+  %c = shufflevector <2 x half> %b, <2 x half> undef, <2 x i32> zeroinitializer
+  %d = fcmp uno <2 x half> %c, %a
+  store <2 x i1> %d, <2 x i1>* %z
+  ret void
+}