[RISCV] Support vector strict rounding operations.

[fakepaper56]

Summary

The patch basically models custom lowering of base rounding operations to expand
rounding by coverting to ingter and coverting back to FP. The other one thing
the patch does is to covert sNan of the source to qNan.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2572

There shouldn't be an overflow exception. If the value is already an "integer" in the FP register we shouldn't go through the integer conversion. The integer conversion only exists because of an oddity of the RISC-V ISA. That's not part of the definition of these functions.

The one thing we need to do that the non-strict doesn't handle is converting snan to qnan.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2572

And we need to avoid raising an inexact exception for nearbyint.

[fakepaper56]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2572

The one thing we need to do that the non-strict doesn't handle is converting snan to qnan.

Sorry, I don't know why we need to covert snan to qnan? Could you elaborate it ?

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2572

A signaling nan input should set the invalid exception and produce a quiet nan output. I believe the the glibc library implementation handles it by adding the input to itself if the input is nan. The fadd instruction produces qnan for an snan input and raises the invalid exception.

[fakepaper56]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2572

Thank you. I got your idea.

[fakepaper56]

Expanding by converting sNan to qNan and using base opcocdes with the new source.
The change makes the revision be blocked by D148619, a revision supports
vector rint/nearbyint in RISC-V.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

401

STRICT_FNEARBYINT is already the third entry in the list

2550

Src needs a freeze.

2556

This needs to be a masked fadd I think. If Src is a large number instead of a nan, adding a larger number to itself can produce infinity which will set the overflow exception which we don't want.

2583

I'm not sure we can convert to the base opcode. I think we need to keep the chain through the expanded opcodes. We need to make sure rounding mode changes or reads of fflags that should logically happen after the rounding expansion can't move up. The fadd probably protects anything above from sinking down.

[fakepaper56]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2583

Is the strict_nearbyint is the cause that we should not covert to base opcode? If some modification of FRM is after strict_nearbyint and we covert strict_nearbyint to nearbyint, it maybe let compiler move the modification of FRM before the nearbyint.

[fakepaper56]

Just expanding by coverting to integer and coverting back to FP and also
coverting sNan to qNan.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2583

That is one case yes. I think we could also move a write of fflags before the integer conversion. The integer can cause an inexact exception. If the user was trying to clear that inexact exception, it would be a bug to move their fflag write earlier.

[craig.topper]

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

2630

Update the Chain in this if too.

[fakepaper56]

Update chain after coverting back to FP.

[craig.topper]

LGTM

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

setOperationAction({ISD::FREM, ISD::FPOW, ISD::FPOWI,
                         ISD::FEXP2, ISD::FLOG, ISD::FLOG2, ISD::FLOG10},
                        MVT::f16, Promote);
 
     // FIXME: Need to promote f16 STRICT_* to f32 libcalls, but we don't have
     // complete support for all operations in LegalizeDAG.
     setOperationAction({ISD::STRICT_FCEIL, ISD::STRICT_FFLOOR,
                         ISD::STRICT_FNEARBYINT, ISD::STRICT_FRINT,
                         ISD::STRICT_FROUND, ISD::STRICT_FROUNDEVEN,
-                        ISD::STRICT_FTRUNC},
+                        ISD::STRICT_FTRUNC, ISD::STRICT_FNEARBYINT},
                        MVT::f16, Promote);
 
     // We need to custom promote this.
     if (Subtarget.is64Bit())
       setOperationAction(ISD::FPOWI, MVT::i32, Custom);
   }
 
   if (Subtarget.hasStdExtF()) {
const auto SetCommonVFPActions = [&](MVT VT) {
 
       setOperationAction(FloatingPointVPOps, VT, Custom);
 
       setOperationAction({ISD::STRICT_FP_EXTEND, ISD::STRICT_FP_ROUND}, VT,
                          Custom);
       setOperationAction({ISD::STRICT_FADD, ISD::STRICT_FSUB, ISD::STRICT_FMUL,
                           ISD::STRICT_FDIV, ISD::STRICT_FSQRT, ISD::STRICT_FMA},
                          VT, Legal);
-      setOperationAction({ISD::STRICT_FSETCC, ISD::STRICT_FSETCCS}, VT, Custom);
+      setOperationAction({ISD::STRICT_FSETCC, ISD::STRICT_FSETCCS,
+                          ISD::STRICT_FTRUNC, ISD::STRICT_FCEIL,
+                          ISD::STRICT_FFLOOR, ISD::STRICT_FROUND,
+                          ISD::STRICT_FROUNDEVEN, ISD::STRICT_FNEARBYINT},
+                         VT, Custom);
     };
 
     // Sets common extload/truncstore actions on RVV floating-point vector
     // types.
     const auto SetCommonVFPExtLoadTruncStoreActions =
         [&](MVT VT, ArrayRef<MVT::SimpleValueType> SmallerVTs) {
           for (auto SmallVT : SmallerVTs) {
             setTruncStoreAction(VT, SmallVT, Expand);
if (Subtarget.useRVVForFixedLengthVectors()) {
         setOperationAction(ISD::BITCAST, VT, Custom);
 
         setOperationAction(FloatingPointVecReduceOps, VT, Custom);
 
         setOperationAction(FloatingPointVPOps, VT, Custom);
 
         setOperationAction({ISD::STRICT_FP_EXTEND, ISD::STRICT_FP_ROUND}, VT,
                            Custom);
-        setOperationAction({ISD::STRICT_FADD, ISD::STRICT_FSUB,
-                            ISD::STRICT_FMUL, ISD::STRICT_FDIV,
-                            ISD::STRICT_FSQRT, ISD::STRICT_FMA},
-                           VT, Custom);
-        setOperationAction({ISD::STRICT_FSETCC, ISD::STRICT_FSETCCS}, VT,
-                           Custom);
+        setOperationAction(
+            {ISD::STRICT_FADD, ISD::STRICT_FSUB, ISD::STRICT_FMUL,
+             ISD::STRICT_FDIV, ISD::STRICT_FSQRT, ISD::STRICT_FMA,
+             ISD::STRICT_FSETCC, ISD::STRICT_FSETCCS, ISD::STRICT_FTRUNC,
+             ISD::STRICT_FCEIL, ISD::STRICT_FFLOOR, ISD::STRICT_FROUND,
+             ISD::STRICT_FROUNDEVEN},
       }
 
       // Custom-legalize bitcasts from fixed-length vectors to scalar types.
       setOperationAction(ISD::BITCAST, {MVT::i8, MVT::i16, MVT::i32, MVT::i64},
                          Custom);
       if (Subtarget.hasStdExtZfhOrZfhmin())
         setOperationAction(ISD::BITCAST, MVT::f16, Custom);
       if (Subtarget.hasStdExtF())
if (DstVT.isFixedLengthVector())
     Res = convertFromScalableVector(DstVT, Res, DAG, Subtarget);
 
   return Res;
 }
 
 static RISCVFPRndMode::RoundingMode matchRoundingOp(unsigned Opc) {
   switch (Opc) {
   case ISD::FROUNDEVEN:
+  case ISD::STRICT_FROUNDEVEN:
   case ISD::VP_FROUNDEVEN:
     return RISCVFPRndMode::RNE;
   case ISD::FTRUNC:
+  case ISD::STRICT_FTRUNC:
   case ISD::VP_FROUNDTOZERO:
     return RISCVFPRndMode::RTZ;
   case ISD::FFLOOR:
+  case ISD::STRICT_FFLOOR:
   case ISD::VP_FFLOOR:
     return RISCVFPRndMode::RDN;
   case ISD::FCEIL:
+  case ISD::STRICT_FCEIL:
   case ISD::VP_FCEIL:
     return RISCVFPRndMode::RUP;
   case ISD::FROUND:
+  case ISD::STRICT_FROUND:
   case ISD::VP_FROUND:
     return RISCVFPRndMode::RMM;
   case ISD::FRINT:
     return RISCVFPRndMode::DYN;
   }
 
   return RISCVFPRndMode::Invalid;
 }
Truncated = DAG.getNode(RISCVISD::FCOPYSIGN_VL, DL, ContainerVT, Truncated,
                           Src, Src, Mask, VL);
 
   if (!VT.isFixedLengthVector())
     return Truncated;
 
   return convertFromScalableVector(VT, Truncated, DAG, Subtarget);
 }
 
+// Expand vector STRICT_FTRUNC, STRICT_FCEIL, STRICT_FFLOOR, STRICT_FROUND
+// STRICT_FROUNDEVEN and STRICT_FNEARBYINT by converting sNan of the source to
+// qNan and replacing strict opcodes with base opcodes.
+static SDValue
+lowerVectorStrictFTRUNC_FCEIL_FFLOOR_FROUND(SDValue Op, SelectionDAG &DAG,
+                                            const RISCVSubtarget &Subtarget) {
+  SDLoc DL(Op);
+  MVT VT = Op.getSimpleValueType();
+  SDValue Chain = Op.getOperand(0);
+  SDValue Src = Op.getOperand(1);
+
+  MVT MaskVT = VT.changeVectorElementType(MVT::i1);
+  SDValue Unorder =
+      DAG.getNode(ISD::STRICT_FSETCC, DL, DAG.getVTList(MaskVT, MVT::Other),
+                  Chain, Src, Src, DAG.getCondCode(ISD::SETUO));
+  SDValue Fadd = DAG.getNode(ISD::STRICT_FADD, DL,
+                             DAG.getVTList(VT, MVT::Other), Chain, Src, Src);
+  Src = DAG.getSelect(DL, VT, Unorder, Fadd, Src);
+
+  unsigned BaseOpc;
+  switch (Op.getOpcode()) {
+  default:
+    llvm_unreachable("Unexpected opcode!");
+  case ISD::STRICT_FCEIL:
+    BaseOpc = ISD::FCEIL;
+    break;
+  case ISD::STRICT_FFLOOR:
+    BaseOpc = ISD::FFLOOR;
+    break;
+  case ISD::STRICT_FNEARBYINT:
+    BaseOpc = ISD::FNEARBYINT;
+    break;
+  case ISD::STRICT_FROUND:
+    BaseOpc = ISD::FROUND;
+    break;
+  case ISD::STRICT_FROUNDEVEN:
+    BaseOpc = ISD::FROUNDEVEN;
+    break;
+  case ISD::STRICT_FTRUNC:
+    BaseOpc = ISD::FTRUNC;
+    break;
+  }
+  SDValue BaseRes = DAG.getNode(BaseOpc, DL, VT, Src);
+  Chain = DAG.getMergeValues({Unorder.getValue(1), Fadd.getValue(1)}, DL);
+  return DAG.getMergeValues({BaseRes, Chain}, DL);
+}
+
 static SDValue
 lowerFTRUNC_FCEIL_FFLOOR_FROUND(SDValue Op, SelectionDAG &DAG,
                                 const RISCVSubtarget &Subtarget) {
   MVT VT = Op.getSimpleValueType();
   if (VT.isVector())
     return lowerVectorFTRUNC_FCEIL_FFLOOR_FROUND(Op, DAG, Subtarget);
 
   if (DAG.shouldOptForSize())

 static std::optional<uint64_t> getExactInteger(const APFloat &APF,
                                                uint32_t BitWidth) {
   APSInt ValInt(BitWidth, !APF.isNegative());
   // We use an arbitrary rounding mode here. If a floating-point is an exact
   // integer (e.g., 1.0), the rounding mode does not affect the output value. If
   // the rounding mode changes the output value, then it is not an exact
   // integer.
   RoundingMode ArbitraryRM = RoundingMode::TowardZero;
   bool IsExact;
   // If it is out of signed integer range, it will return an invalid operation.
   // If it is not an exact integer, IsExact is false.
   if ((APF.convertToInteger(ValInt, ArbitraryRM, &IsExact) ==
        APFloatBase::opInvalidOp) ||
       !IsExact)
     return std::nullopt;
   return ValInt.extractBitsAsZExtValue(BitWidth, 0);
 }
return lowerToScalableOp(Op, DAG, RISCVISD::STRICT_FDIV_VL,
                              /*HasMergeOp*/ true);
   case ISD::STRICT_FSQRT:
     return lowerToScalableOp(Op, DAG, RISCVISD::STRICT_FSQRT_VL);
   case ISD::STRICT_FMA:
     return lowerToScalableOp(Op, DAG, RISCVISD::STRICT_VFMADD_VL);
   case ISD::STRICT_FSETCC:
   case ISD::STRICT_FSETCCS:
     return lowerVectorStrictFSetcc(Op, DAG);
+  case ISD::STRICT_FCEIL:
+  case ISD::STRICT_FRINT:
+  case ISD::STRICT_FFLOOR:
+  case ISD::STRICT_FTRUNC:
+  case ISD::STRICT_FNEARBYINT:
+  case ISD::STRICT_FROUND:
+  case ISD::STRICT_FROUNDEVEN:
+    return lowerVectorStrictFTRUNC_FCEIL_FFLOOR_FROUND(Op, DAG, Subtarget);
   case ISD::MGATHER:
   case ISD::VP_GATHER:
     return lowerMaskedGather(Op, DAG);
   case ISD::MSCATTER:
   case ISD::VP_SCATTER:
     return lowerMaskedScatter(Op, DAG);
   case ISD::GET_ROUNDING:
     return lowerGET_ROUNDING(Op, DAG);

llvm/test/CodeGen/RISCV/rvv/fceil-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+define <vscale x 1 x half> @ceil_nxv1f16(<vscale x 1 x half> %x) {
+; CHECK-LABEL: ceil_nxv1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x half> @llvm.experimental.constrained.ceil.nxv1f16(<vscale x 1 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x half> %a
+}
+declare <vscale x 1 x half> @llvm.experimental.constrained.ceil.nxv1f16(<vscale x 1 x half>, metadata)
+
+define <vscale x 2 x half> @ceil_nxv2f16(<vscale x 2 x half> %x) {
+; CHECK-LABEL: ceil_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x half> @llvm.experimental.constrained.ceil.nxv2f16(<vscale x 2 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x half> %a
+}
+declare <vscale x 2 x half> @llvm.experimental.constrained.ceil.nxv2f16(<vscale x 2 x half>, metadata)
+
+define <vscale x 4 x half> @ceil_nxv4f16(<vscale x 4 x half> %x) {
+; CHECK-LABEL: ceil_nxv4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x half> @llvm.experimental.constrained.ceil.nxv4f16(<vscale x 4 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x half> %a
+}
+declare <vscale x 4 x half> @llvm.experimental.constrained.ceil.nxv4f16(<vscale x 4 x half>, metadata)
+
+define <vscale x 8 x half> @ceil_nxv8f16(<vscale x 8 x half> %x) {
+; CHECK-LABEL: ceil_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x half> @llvm.experimental.constrained.ceil.nxv8f16(<vscale x 8 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x half> %a
+}
+declare <vscale x 8 x half> @llvm.experimental.constrained.ceil.nxv8f16(<vscale x 8 x half>, metadata)
+
+define <vscale x 16 x half> @ceil_nxv16f16(<vscale x 16 x half> %x) {
+; CHECK-LABEL: ceil_nxv16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x half> @llvm.experimental.constrained.ceil.nxv16f16(<vscale x 16 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x half> %a
+}
+declare <vscale x 16 x half> @llvm.experimental.constrained.ceil.nxv16f16(<vscale x 16 x half>, metadata)
+
+define <vscale x 32 x half> @ceil_nxv32f16(<vscale x 32 x half> %x) {
+; CHECK-LABEL: ceil_nxv32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 32 x half> @llvm.experimental.constrained.ceil.nxv32f16(<vscale x 32 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 32 x half> %a
+}
+declare <vscale x 32 x half> @llvm.experimental.constrained.ceil.nxv32f16(<vscale x 32 x half>, metadata)
+
+define <vscale x 1 x float> @ceil_nxv1f32(<vscale x 1 x float> %x) {
+; CHECK-LABEL: ceil_nxv1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x float> @llvm.experimental.constrained.ceil.nxv1f32(<vscale x 1 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x float> %a
+}
+declare <vscale x 1 x float> @llvm.experimental.constrained.ceil.nxv1f32(<vscale x 1 x float>, metadata)
+
+define <vscale x 2 x float> @ceil_nxv2f32(<vscale x 2 x float> %x) {
+; CHECK-LABEL: ceil_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x float> @llvm.experimental.constrained.ceil.nxv2f32(<vscale x 2 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x float> %a
+}
+declare <vscale x 2 x float> @llvm.experimental.constrained.ceil.nxv2f32(<vscale x 2 x float>, metadata)
+
+define <vscale x 4 x float> @ceil_nxv4f32(<vscale x 4 x float> %x) {
+; CHECK-LABEL: ceil_nxv4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x float> @llvm.experimental.constrained.ceil.nxv4f32(<vscale x 4 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x float> %a
+}
+declare <vscale x 4 x float> @llvm.experimental.constrained.ceil.nxv4f32(<vscale x 4 x float>, metadata)
+
+define <vscale x 8 x float> @ceil_nxv8f32(<vscale x 8 x float> %x) {
+; CHECK-LABEL: ceil_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x float> @llvm.experimental.constrained.ceil.nxv8f32(<vscale x 8 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x float> %a
+}
+declare <vscale x 8 x float> @llvm.experimental.constrained.ceil.nxv8f32(<vscale x 8 x float>, metadata)
+
+define <vscale x 16 x float> @ceil_nxv16f32(<vscale x 16 x float> %x) {
+; CHECK-LABEL: ceil_nxv16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x float> @llvm.experimental.constrained.ceil.nxv16f32(<vscale x 16 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x float> %a
+}
+declare <vscale x 16 x float> @llvm.experimental.constrained.ceil.nxv16f32(<vscale x 16 x float>, metadata)
+
+define <vscale x 1 x double> @ceil_nxv1f64(<vscale x 1 x double> %x) {
+; CHECK-LABEL: ceil_nxv1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x double> @llvm.experimental.constrained.ceil.nxv1f64(<vscale x 1 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x double> %a
+}
+declare <vscale x 1 x double> @llvm.experimental.constrained.ceil.nxv1f64(<vscale x 1 x double>, metadata)
+
+define <vscale x 2 x double> @ceil_nxv2f64(<vscale x 2 x double> %x) {
+; CHECK-LABEL: ceil_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x double> @llvm.experimental.constrained.ceil.nxv2f64(<vscale x 2 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x double> %a
+}
+declare <vscale x 2 x double> @llvm.experimental.constrained.ceil.nxv2f64(<vscale x 2 x double>, metadata)
+
+define <vscale x 4 x double> @ceil_nxv4f64(<vscale x 4 x double> %x) {
+; CHECK-LABEL: ceil_nxv4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x double> @llvm.experimental.constrained.ceil.nxv4f64(<vscale x 4 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x double> %a
+}
+declare <vscale x 4 x double> @llvm.experimental.constrained.ceil.nxv4f64(<vscale x 4 x double>, metadata)
+
+define <vscale x 8 x double> @ceil_nxv8f64(<vscale x 8 x double> %x) {
+; CHECK-LABEL: ceil_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x double> @llvm.experimental.constrained.ceil.nxv8f64(<vscale x 8 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x double> %a
+}
+declare <vscale x 8 x double> @llvm.experimental.constrained.ceil.nxv8f64(<vscale x 8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/ffloor-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+define <vscale x 1 x half> @floor_nxv1f16(<vscale x 1 x half> %x) {
+; CHECK-LABEL: floor_nxv1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x half> @llvm.experimental.constrained.floor.nxv1f16(<vscale x 1 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x half> %a
+}
+declare <vscale x 1 x half> @llvm.experimental.constrained.floor.nxv1f16(<vscale x 1 x half>, metadata)
+
+define <vscale x 2 x half> @floor_nxv2f16(<vscale x 2 x half> %x) {
+; CHECK-LABEL: floor_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x half> @llvm.experimental.constrained.floor.nxv2f16(<vscale x 2 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x half> %a
+}
+declare <vscale x 2 x half> @llvm.experimental.constrained.floor.nxv2f16(<vscale x 2 x half>, metadata)
+
+define <vscale x 4 x half> @floor_nxv4f16(<vscale x 4 x half> %x) {
+; CHECK-LABEL: floor_nxv4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x half> @llvm.experimental.constrained.floor.nxv4f16(<vscale x 4 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x half> %a
+}
+declare <vscale x 4 x half> @llvm.experimental.constrained.floor.nxv4f16(<vscale x 4 x half>, metadata)
+
+define <vscale x 8 x half> @floor_nxv8f16(<vscale x 8 x half> %x) {
+; CHECK-LABEL: floor_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x half> @llvm.experimental.constrained.floor.nxv8f16(<vscale x 8 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x half> %a
+}
+declare <vscale x 8 x half> @llvm.experimental.constrained.floor.nxv8f16(<vscale x 8 x half>, metadata)
+
+define <vscale x 16 x half> @floor_nxv16f16(<vscale x 16 x half> %x) {
+; CHECK-LABEL: floor_nxv16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x half> @llvm.experimental.constrained.floor.nxv16f16(<vscale x 16 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x half> %a
+}
+declare <vscale x 16 x half> @llvm.experimental.constrained.floor.nxv16f16(<vscale x 16 x half>, metadata)
+
+define <vscale x 32 x half> @floor_nxv32f16(<vscale x 32 x half> %x) {
+; CHECK-LABEL: floor_nxv32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 32 x half> @llvm.experimental.constrained.floor.nxv32f16(<vscale x 32 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 32 x half> %a
+}
+declare <vscale x 32 x half> @llvm.experimental.constrained.floor.nxv32f16(<vscale x 32 x half>, metadata)
+
+define <vscale x 1 x float> @floor_nxv1f32(<vscale x 1 x float> %x) {
+; CHECK-LABEL: floor_nxv1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x float> @llvm.experimental.constrained.floor.nxv1f32(<vscale x 1 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x float> %a
+}
+declare <vscale x 1 x float> @llvm.experimental.constrained.floor.nxv1f32(<vscale x 1 x float>, metadata)
+
+define <vscale x 2 x float> @floor_nxv2f32(<vscale x 2 x float> %x) {
+; CHECK-LABEL: floor_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x float> @llvm.experimental.constrained.floor.nxv2f32(<vscale x 2 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x float> %a
+}
+declare <vscale x 2 x float> @llvm.experimental.constrained.floor.nxv2f32(<vscale x 2 x float>, metadata)
+
+define <vscale x 4 x float> @floor_nxv4f32(<vscale x 4 x float> %x) {
+; CHECK-LABEL: floor_nxv4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x float> @llvm.experimental.constrained.floor.nxv4f32(<vscale x 4 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x float> %a
+}
+declare <vscale x 4 x float> @llvm.experimental.constrained.floor.nxv4f32(<vscale x 4 x float>, metadata)
+
+define <vscale x 8 x float> @floor_nxv8f32(<vscale x 8 x float> %x) {
+; CHECK-LABEL: floor_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x float> @llvm.experimental.constrained.floor.nxv8f32(<vscale x 8 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x float> %a
+}
+declare <vscale x 8 x float> @llvm.experimental.constrained.floor.nxv8f32(<vscale x 8 x float>, metadata)
+
+define <vscale x 16 x float> @floor_nxv16f32(<vscale x 16 x float> %x) {
+; CHECK-LABEL: floor_nxv16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x float> @llvm.experimental.constrained.floor.nxv16f32(<vscale x 16 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x float> %a
+}
+declare <vscale x 16 x float> @llvm.experimental.constrained.floor.nxv16f32(<vscale x 16 x float>, metadata)
+
+define <vscale x 1 x double> @floor_nxv1f64(<vscale x 1 x double> %x) {
+; CHECK-LABEL: floor_nxv1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x double> @llvm.experimental.constrained.floor.nxv1f64(<vscale x 1 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x double> %a
+}
+declare <vscale x 1 x double> @llvm.experimental.constrained.floor.nxv1f64(<vscale x 1 x double>, metadata)
+
+define <vscale x 2 x double> @floor_nxv2f64(<vscale x 2 x double> %x) {
+; CHECK-LABEL: floor_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x double> @llvm.experimental.constrained.floor.nxv2f64(<vscale x 2 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x double> %a
+}
+declare <vscale x 2 x double> @llvm.experimental.constrained.floor.nxv2f64(<vscale x 2 x double>, metadata)
+
+define <vscale x 4 x double> @floor_nxv4f64(<vscale x 4 x double> %x) {
+; CHECK-LABEL: floor_nxv4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x double> @llvm.experimental.constrained.floor.nxv4f64(<vscale x 4 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x double> %a
+}
+declare <vscale x 4 x double> @llvm.experimental.constrained.floor.nxv4f64(<vscale x 4 x double>, metadata)
+
+define <vscale x 8 x double> @floor_nxv8f64(<vscale x 8 x double> %x) {
+; CHECK-LABEL: floor_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x double> @llvm.experimental.constrained.floor.nxv8f64(<vscale x 8 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x double> %a
+}
+declare <vscale x 8 x double> @llvm.experimental.constrained.floor.nxv8f64(<vscale x 8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fceil-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+define <1 x half> @ceil_v1f16(<1 x half> %x) {
+; CHECK-LABEL: ceil_v1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x half> @llvm.experimental.constrained.ceil.v1f16(<1 x half> %x, metadata !"fpexcept.strict")
+  ret <1 x half> %a
+}
+declare <1 x half> @llvm.experimental.constrained.ceil.v1f16(<1 x half>, metadata)
+
+define <2 x half> @ceil_v2f16(<2 x half> %x) {
+; CHECK-LABEL: ceil_v2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x half> @llvm.experimental.constrained.ceil.v2f16(<2 x half> %x, metadata !"fpexcept.strict")
+  ret <2 x half> %a
+}
+declare <2 x half> @llvm.experimental.constrained.ceil.v2f16(<2 x half>, metadata)
+
+define <4 x half> @ceil_v4f16(<4 x half> %x) {
+; CHECK-LABEL: ceil_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x half> @llvm.experimental.constrained.ceil.v4f16(<4 x half> %x, metadata !"fpexcept.strict")
+  ret <4 x half> %a
+}
+declare <4 x half> @llvm.experimental.constrained.ceil.v4f16(<4 x half>, metadata)
+
+define <8 x half> @ceil_v8f16(<8 x half> %x) {
+; CHECK-LABEL: ceil_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x half> @llvm.experimental.constrained.ceil.v8f16(<8 x half> %x, metadata !"fpexcept.strict")
+  ret <8 x half> %a
+}
+declare <8 x half> @llvm.experimental.constrained.ceil.v8f16(<8 x half>, metadata)
+
+define <16 x half> @ceil_v16f16(<16 x half> %x) {
+; CHECK-LABEL: ceil_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x half> @llvm.experimental.constrained.ceil.v16f16(<16 x half> %x, metadata !"fpexcept.strict")
+  ret <16 x half> %a
+}
+declare <16 x half> @llvm.experimental.constrained.ceil.v16f16(<16 x half>, metadata)
+
+define <32 x half> @ceil_v32f16(<32 x half> %x) {
+; CHECK-LABEL: ceil_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <32 x half> @llvm.experimental.constrained.ceil.v32f16(<32 x half> %x, metadata !"fpexcept.strict")
+  ret <32 x half> %a
+}
+declare <32 x half> @llvm.experimental.constrained.ceil.v32f16(<32 x half>, metadata)
+
+define <1 x float> @ceil_v1f32(<1 x float> %x) {
+; CHECK-LABEL: ceil_v1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x float> @llvm.experimental.constrained.ceil.v1f32(<1 x float> %x, metadata !"fpexcept.strict")
+  ret <1 x float> %a
+}
+declare <1 x float> @llvm.experimental.constrained.ceil.v1f32(<1 x float>, metadata)
+
+define <2 x float> @ceil_v2f32(<2 x float> %x) {
+; CHECK-LABEL: ceil_v2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x float> @llvm.experimental.constrained.ceil.v2f32(<2 x float> %x, metadata !"fpexcept.strict")
+  ret <2 x float> %a
+}
+declare <2 x float> @llvm.experimental.constrained.ceil.v2f32(<2 x float>, metadata)
+
+define <4 x float> @ceil_v4f32(<4 x float> %x) {
+; CHECK-LABEL: ceil_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x float> @llvm.experimental.constrained.ceil.v4f32(<4 x float> %x, metadata !"fpexcept.strict")
+  ret <4 x float> %a
+}
+declare <4 x float> @llvm.experimental.constrained.ceil.v4f32(<4 x float>, metadata)
+
+define <8 x float> @ceil_v8f32(<8 x float> %x) {
+; CHECK-LABEL: ceil_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x float> @llvm.experimental.constrained.ceil.v8f32(<8 x float> %x, metadata !"fpexcept.strict")
+  ret <8 x float> %a
+}
+declare <8 x float> @llvm.experimental.constrained.ceil.v8f32(<8 x float>, metadata)
+
+define <16 x float> @ceil_v16f32(<16 x float> %x) {
+; CHECK-LABEL: ceil_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x float> @llvm.experimental.constrained.ceil.v16f32(<16 x float> %x, metadata !"fpexcept.strict")
+  ret <16 x float> %a
+}
+declare <16 x float> @llvm.experimental.constrained.ceil.v16f32(<16 x float>, metadata)
+
+define <1 x double> @ceil_v1f64(<1 x double> %x) {
+; CHECK-LABEL: ceil_v1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x double> @llvm.experimental.constrained.ceil.v1f64(<1 x double> %x, metadata !"fpexcept.strict")
+  ret <1 x double> %a
+}
+declare <1 x double> @llvm.experimental.constrained.ceil.v1f64(<1 x double>, metadata)
+
+define <2 x double> @ceil_v2f64(<2 x double> %x) {
+; CHECK-LABEL: ceil_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x double> @llvm.experimental.constrained.ceil.v2f64(<2 x double> %x, metadata !"fpexcept.strict")
+  ret <2 x double> %a
+}
+declare <2 x double> @llvm.experimental.constrained.ceil.v2f64(<2 x double>, metadata)
+
+define <4 x double> @ceil_v4f64(<4 x double> %x) {
+; CHECK-LABEL: ceil_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x double> @llvm.experimental.constrained.ceil.v4f64(<4 x double> %x, metadata !"fpexcept.strict")
+  ret <4 x double> %a
+}
+declare <4 x double> @llvm.experimental.constrained.ceil.v4f64(<4 x double>, metadata)
+
+define <8 x double> @ceil_v8f64(<8 x double> %x) {
+; CHECK-LABEL: ceil_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 3
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x double> @llvm.experimental.constrained.ceil.v8f64(<8 x double> %x, metadata !"fpexcept.strict")
+  ret <8 x double> %a
+}
+declare <8 x double> @llvm.experimental.constrained.ceil.v8f64(<8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-ffloor-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+define <1 x half> @floor_v1f16(<1 x half> %x) {
+; CHECK-LABEL: floor_v1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x half> @llvm.experimental.constrained.floor.v1f16(<1 x half> %x, metadata !"fpexcept.strict")
+  ret <1 x half> %a
+}
+declare <1 x half> @llvm.experimental.constrained.floor.v1f16(<1 x half>, metadata)
+
+define <2 x half> @floor_v2f16(<2 x half> %x) {
+; CHECK-LABEL: floor_v2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x half> @llvm.experimental.constrained.floor.v2f16(<2 x half> %x, metadata !"fpexcept.strict")
+  ret <2 x half> %a
+}
+declare <2 x half> @llvm.experimental.constrained.floor.v2f16(<2 x half>, metadata)
+
+define <4 x half> @floor_v4f16(<4 x half> %x) {
+; CHECK-LABEL: floor_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x half> @llvm.experimental.constrained.floor.v4f16(<4 x half> %x, metadata !"fpexcept.strict")
+  ret <4 x half> %a
+}
+declare <4 x half> @llvm.experimental.constrained.floor.v4f16(<4 x half>, metadata)
+
+define <8 x half> @floor_v8f16(<8 x half> %x) {
+; CHECK-LABEL: floor_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x half> @llvm.experimental.constrained.floor.v8f16(<8 x half> %x, metadata !"fpexcept.strict")
+  ret <8 x half> %a
+}
+declare <8 x half> @llvm.experimental.constrained.floor.v8f16(<8 x half>, metadata)
+
+define <16 x half> @floor_v16f16(<16 x half> %x) {
+; CHECK-LABEL: floor_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x half> @llvm.experimental.constrained.floor.v16f16(<16 x half> %x, metadata !"fpexcept.strict")
+  ret <16 x half> %a
+}
+declare <16 x half> @llvm.experimental.constrained.floor.v16f16(<16 x half>, metadata)
+
+define <32 x half> @floor_v32f16(<32 x half> %x) {
+; CHECK-LABEL: floor_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <32 x half> @llvm.experimental.constrained.floor.v32f16(<32 x half> %x, metadata !"fpexcept.strict")
+  ret <32 x half> %a
+}
+declare <32 x half> @llvm.experimental.constrained.floor.v32f16(<32 x half>, metadata)
+
+define <1 x float> @floor_v1f32(<1 x float> %x) {
+; CHECK-LABEL: floor_v1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x float> @llvm.experimental.constrained.floor.v1f32(<1 x float> %x, metadata !"fpexcept.strict")
+  ret <1 x float> %a
+}
+declare <1 x float> @llvm.experimental.constrained.floor.v1f32(<1 x float>, metadata)
+
+define <2 x float> @floor_v2f32(<2 x float> %x) {
+; CHECK-LABEL: floor_v2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x float> @llvm.experimental.constrained.floor.v2f32(<2 x float> %x, metadata !"fpexcept.strict")
+  ret <2 x float> %a
+}
+declare <2 x float> @llvm.experimental.constrained.floor.v2f32(<2 x float>, metadata)
+
+define <4 x float> @floor_v4f32(<4 x float> %x) {
+; CHECK-LABEL: floor_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x float> @llvm.experimental.constrained.floor.v4f32(<4 x float> %x, metadata !"fpexcept.strict")
+  ret <4 x float> %a
+}
+declare <4 x float> @llvm.experimental.constrained.floor.v4f32(<4 x float>, metadata)
+
+define <8 x float> @floor_v8f32(<8 x float> %x) {
+; CHECK-LABEL: floor_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x float> @llvm.experimental.constrained.floor.v8f32(<8 x float> %x, metadata !"fpexcept.strict")
+  ret <8 x float> %a
+}
+declare <8 x float> @llvm.experimental.constrained.floor.v8f32(<8 x float>, metadata)
+
+define <16 x float> @floor_v16f32(<16 x float> %x) {
+; CHECK-LABEL: floor_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x float> @llvm.experimental.constrained.floor.v16f32(<16 x float> %x, metadata !"fpexcept.strict")
+  ret <16 x float> %a
+}
+declare <16 x float> @llvm.experimental.constrained.floor.v16f32(<16 x float>, metadata)
+
+define <1 x double> @floor_v1f64(<1 x double> %x) {
+; CHECK-LABEL: floor_v1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x double> @llvm.experimental.constrained.floor.v1f64(<1 x double> %x, metadata !"fpexcept.strict")
+  ret <1 x double> %a
+}
+declare <1 x double> @llvm.experimental.constrained.floor.v1f64(<1 x double>, metadata)
+
+define <2 x double> @floor_v2f64(<2 x double> %x) {
+; CHECK-LABEL: floor_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x double> @llvm.experimental.constrained.floor.v2f64(<2 x double> %x, metadata !"fpexcept.strict")
+  ret <2 x double> %a
+}
+declare <2 x double> @llvm.experimental.constrained.floor.v2f64(<2 x double>, metadata)
+
+define <4 x double> @floor_v4f64(<4 x double> %x) {
+; CHECK-LABEL: floor_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x double> @llvm.experimental.constrained.floor.v4f64(<4 x double> %x, metadata !"fpexcept.strict")
+  ret <4 x double> %a
+}
+declare <4 x double> @llvm.experimental.constrained.floor.v4f64(<4 x double>, metadata)
+
+define <8 x double> @floor_v8f64(<8 x double> %x) {
+; CHECK-LABEL: floor_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 2
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x double> @llvm.experimental.constrained.floor.v8f64(<8 x double> %x, metadata !"fpexcept.strict")
+  ret <8 x double> %a
+}
+declare <8 x double> @llvm.experimental.constrained.floor.v8f64(<8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fnearbyint-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+declare <2 x half> @llvm.experimental.constrained.nearbyint.v2f16(<2 x half>, metadata, metadata)
+
+define <2 x half> @nearbyint_v2f16(<2 x half> %v) {
+  %r = call <2 x half> @llvm.experimental.constrained.nearbyint.v2f16(<2 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <2 x half> %r
+}
+
+declare <4 x half> @llvm.experimental.constrained.nearbyint.v4f16(<4 x half>, metadata, metadata)
+
+define <4 x half> @nearbyint_v4f16(<4 x half> %v) {
+  %r = call <4 x half> @llvm.experimental.constrained.nearbyint.v4f16(<4 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <4 x half> %r
+}
+
+declare <8 x half> @llvm.experimental.constrained.nearbyint.v8f16(<8 x half>, metadata, metadata)
+
+define <8 x half> @nearbyint_v8f16(<8 x half> %v) {
+  %r = call <8 x half> @llvm.experimental.constrained.nearbyint.v8f16(<8 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <8 x half> %r
+}
+
+declare <16 x half> @llvm.experimental.constrained.nearbyint.v16f16(<16 x half>, metadata, metadata)
+
+define <16 x half> @nearbyint_v16f16(<16 x half> %v) {
+  %r = call <16 x half> @llvm.experimental.constrained.nearbyint.v16f16(<16 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <16 x half> %r
+}
+
+declare <32 x half> @llvm.experimental.constrained.nearbyint.v32f16(<32 x half>, metadata, metadata)
+
+define <32 x half> @nearbyint_v32f16(<32 x half> %v) {
+  %r = call <32 x half> @llvm.experimental.constrained.nearbyint.v32f16(<32 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <32 x half> %r
+}
+
+declare <2 x float> @llvm.experimental.constrained.nearbyint.v2f32(<2 x float>, metadata, metadata)
+
+define <2 x float> @nearbyint_v2f32(<2 x float> %v) {
+  %r = call <2 x float> @llvm.experimental.constrained.nearbyint.v2f32(<2 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <2 x float> %r
+}
+
+declare <4 x float> @llvm.experimental.constrained.nearbyint.v4f32(<4 x float>, metadata, metadata)
+
+define <4 x float> @nearbyint_v4f32(<4 x float> %v) {
+  %r = call <4 x float> @llvm.experimental.constrained.nearbyint.v4f32(<4 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <4 x float> %r
+}
+
+declare <8 x float> @llvm.experimental.constrained.nearbyint.v8f32(<8 x float>, metadata, metadata)
+
+define <8 x float> @nearbyint_v8f32(<8 x float> %v) {
+  %r = call <8 x float> @llvm.experimental.constrained.nearbyint.v8f32(<8 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <8 x float> %r
+}
+
+declare <16 x float> @llvm.experimental.constrained.nearbyint.v16f32(<16 x float>, metadata, metadata)
+
+define <16 x float> @nearbyint_v16f32(<16 x float> %v) {
+  %r = call <16 x float> @llvm.experimental.constrained.nearbyint.v16f32(<16 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <16 x float> %r
+}
+
+declare <2 x double> @llvm.experimental.constrained.nearbyint.v2f64(<2 x double>, metadata, metadata)
+
+define <2 x double> @nearbyint_v2f64(<2 x double> %v) {
+  %r = call <2 x double> @llvm.experimental.constrained.nearbyint.v2f64(<2 x double> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <2 x double> %r
+}
+
+declare <4 x double> @llvm.experimental.constrained.nearbyint.v4f64(<4 x double>, metadata, metadata)
+
+define <4 x double> @nearbyint_v4f64(<4 x double> %v) {
+  %r = call <4 x double> @llvm.experimental.constrained.nearbyint.v4f64(<4 x double> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <4 x double> %r
+}
+
+declare <8 x double> @llvm.experimental.constrained.nearbyint.v8f64(<8 x double>, metadata, metadata)
+
+define <8 x double> @nearbyint_v8f64(<8 x double> %v) {
+  %r = call <8 x double> @llvm.experimental.constrained.nearbyint.v8f64(<8 x double> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <8 x double> %r
+}

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-fround-costrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+; This file tests the code generation for `llvm.experimental.constrained.round.*` on scalable vector type.
+
+define <1 x half> @round_v1f16(<1 x half> %x) {
+; CHECK-LABEL: round_v1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x half> @llvm.experimental.constrained.round.v1f16(<1 x half> %x, metadata !"fpexcept.strict")
+  ret <1 x half> %a
+}
+declare <1 x half> @llvm.experimental.constrained.round.v1f16(<1 x half>, metadata)
+
+define <2 x half> @round_v2f16(<2 x half> %x) {
+; CHECK-LABEL: round_v2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x half> @llvm.experimental.constrained.round.v2f16(<2 x half> %x, metadata !"fpexcept.strict")
+  ret <2 x half> %a
+}
+declare <2 x half> @llvm.experimental.constrained.round.v2f16(<2 x half>, metadata)
+
+define <4 x half> @round_v4f16(<4 x half> %x) {
+; CHECK-LABEL: round_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x half> @llvm.experimental.constrained.round.v4f16(<4 x half> %x, metadata !"fpexcept.strict")
+  ret <4 x half> %a
+}
+declare <4 x half> @llvm.experimental.constrained.round.v4f16(<4 x half>, metadata)
+
+define <8 x half> @round_v8f16(<8 x half> %x) {
+; CHECK-LABEL: round_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x half> @llvm.experimental.constrained.round.v8f16(<8 x half> %x, metadata !"fpexcept.strict")
+  ret <8 x half> %a
+}
+declare <8 x half> @llvm.experimental.constrained.round.v8f16(<8 x half>, metadata)
+
+define <16 x half> @round_v16f16(<16 x half> %x) {
+; CHECK-LABEL: round_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x half> @llvm.experimental.constrained.round.v16f16(<16 x half> %x, metadata !"fpexcept.strict")
+  ret <16 x half> %a
+}
+declare <16 x half> @llvm.experimental.constrained.round.v16f16(<16 x half>, metadata)
+
+define <32 x half> @round_v32f16(<32 x half> %x) {
+; CHECK-LABEL: round_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <32 x half> @llvm.experimental.constrained.round.v32f16(<32 x half> %x, metadata !"fpexcept.strict")
+  ret <32 x half> %a
+}
+declare <32 x half> @llvm.experimental.constrained.round.v32f16(<32 x half>, metadata)
+
+define <1 x float> @round_v1f32(<1 x float> %x) {
+; CHECK-LABEL: round_v1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x float> @llvm.experimental.constrained.round.v1f32(<1 x float> %x, metadata !"fpexcept.strict")
+  ret <1 x float> %a
+}
+declare <1 x float> @llvm.experimental.constrained.round.v1f32(<1 x float>, metadata)
+
+define <2 x float> @round_v2f32(<2 x float> %x) {
+; CHECK-LABEL: round_v2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x float> @llvm.experimental.constrained.round.v2f32(<2 x float> %x, metadata !"fpexcept.strict")
+  ret <2 x float> %a
+}
+declare <2 x float> @llvm.experimental.constrained.round.v2f32(<2 x float>, metadata)
+
+define <4 x float> @round_v4f32(<4 x float> %x) {
+; CHECK-LABEL: round_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x float> @llvm.experimental.constrained.round.v4f32(<4 x float> %x, metadata !"fpexcept.strict")
+  ret <4 x float> %a
+}
+declare <4 x float> @llvm.experimental.constrained.round.v4f32(<4 x float>, metadata)
+
+define <8 x float> @round_v8f32(<8 x float> %x) {
+; CHECK-LABEL: round_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x float> @llvm.experimental.constrained.round.v8f32(<8 x float> %x, metadata !"fpexcept.strict")
+  ret <8 x float> %a
+}
+declare <8 x float> @llvm.experimental.constrained.round.v8f32(<8 x float>, metadata)
+
+define <16 x float> @round_v16f32(<16 x float> %x) {
+; CHECK-LABEL: round_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x float> @llvm.experimental.constrained.round.v16f32(<16 x float> %x, metadata !"fpexcept.strict")
+  ret <16 x float> %a
+}
+declare <16 x float> @llvm.experimental.constrained.round.v16f32(<16 x float>, metadata)
+
+define <1 x double> @round_v1f64(<1 x double> %x) {
+; CHECK-LABEL: round_v1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x double> @llvm.experimental.constrained.round.v1f64(<1 x double> %x, metadata !"fpexcept.strict")
+  ret <1 x double> %a
+}
+declare <1 x double> @llvm.experimental.constrained.round.v1f64(<1 x double>, metadata)
+
+define <2 x double> @round_v2f64(<2 x double> %x) {
+; CHECK-LABEL: round_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x double> @llvm.experimental.constrained.round.v2f64(<2 x double> %x, metadata !"fpexcept.strict")
+  ret <2 x double> %a
+}
+declare <2 x double> @llvm.experimental.constrained.round.v2f64(<2 x double>, metadata)
+
+define <4 x double> @round_v4f64(<4 x double> %x) {
+; CHECK-LABEL: round_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x double> @llvm.experimental.constrained.round.v4f64(<4 x double> %x, metadata !"fpexcept.strict")
+  ret <4 x double> %a
+}
+declare <4 x double> @llvm.experimental.constrained.round.v4f64(<4 x double>, metadata)
+
+define <8 x double> @round_v8f64(<8 x double> %x) {
+; CHECK-LABEL: round_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x double> @llvm.experimental.constrained.round.v8f64(<8 x double> %x, metadata !"fpexcept.strict")
+  ret <8 x double> %a
+}
+declare <8 x double> @llvm.experimental.constrained.round.v8f64(<8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-froundeven-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+; This file tests the code generation for `llvm.experimental.constrained.roundeven.*` on scalable vector type.
+
+define <1 x half> @roundeven_v1f16(<1 x half> %x) {
+; CHECK-LABEL: roundeven_v1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x half> @llvm.experimental.constrained.roundeven.v1f16(<1 x half> %x, metadata !"fpexcept.strict")
+  ret <1 x half> %a
+}
+declare <1 x half> @llvm.experimental.constrained.roundeven.v1f16(<1 x half>, metadata)
+
+define <2 x half> @roundeven_v2f16(<2 x half> %x) {
+; CHECK-LABEL: roundeven_v2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x half> @llvm.experimental.constrained.roundeven.v2f16(<2 x half> %x, metadata !"fpexcept.strict")
+  ret <2 x half> %a
+}
+declare <2 x half> @llvm.experimental.constrained.roundeven.v2f16(<2 x half>, metadata)
+
+define <4 x half> @roundeven_v4f16(<4 x half> %x) {
+; CHECK-LABEL: roundeven_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x half> @llvm.experimental.constrained.roundeven.v4f16(<4 x half> %x, metadata !"fpexcept.strict")
+  ret <4 x half> %a
+}
+declare <4 x half> @llvm.experimental.constrained.roundeven.v4f16(<4 x half>, metadata)
+
+define <8 x half> @roundeven_v8f16(<8 x half> %x) {
+; CHECK-LABEL: roundeven_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x half> @llvm.experimental.constrained.roundeven.v8f16(<8 x half> %x, metadata !"fpexcept.strict")
+  ret <8 x half> %a
+}
+declare <8 x half> @llvm.experimental.constrained.roundeven.v8f16(<8 x half>, metadata)
+
+define <16 x half> @roundeven_v16f16(<16 x half> %x) {
+; CHECK-LABEL: roundeven_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x half> @llvm.experimental.constrained.roundeven.v16f16(<16 x half> %x, metadata !"fpexcept.strict")
+  ret <16 x half> %a
+}
+declare <16 x half> @llvm.experimental.constrained.roundeven.v16f16(<16 x half>, metadata)
+
+define <32 x half> @roundeven_v32f16(<32 x half> %x) {
+; CHECK-LABEL: roundeven_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <32 x half> @llvm.experimental.constrained.roundeven.v32f16(<32 x half> %x, metadata !"fpexcept.strict")
+  ret <32 x half> %a
+}
+declare <32 x half> @llvm.experimental.constrained.roundeven.v32f16(<32 x half>, metadata)
+
+define <1 x float> @roundeven_v1f32(<1 x float> %x) {
+; CHECK-LABEL: roundeven_v1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x float> @llvm.experimental.constrained.roundeven.v1f32(<1 x float> %x, metadata !"fpexcept.strict")
+  ret <1 x float> %a
+}
+declare <1 x float> @llvm.experimental.constrained.roundeven.v1f32(<1 x float>, metadata)
+
+define <2 x float> @roundeven_v2f32(<2 x float> %x) {
+; CHECK-LABEL: roundeven_v2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x float> @llvm.experimental.constrained.roundeven.v2f32(<2 x float> %x, metadata !"fpexcept.strict")
+  ret <2 x float> %a
+}
+declare <2 x float> @llvm.experimental.constrained.roundeven.v2f32(<2 x float>, metadata)
+
+define <4 x float> @roundeven_v4f32(<4 x float> %x) {
+; CHECK-LABEL: roundeven_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x float> @llvm.experimental.constrained.roundeven.v4f32(<4 x float> %x, metadata !"fpexcept.strict")
+  ret <4 x float> %a
+}
+declare <4 x float> @llvm.experimental.constrained.roundeven.v4f32(<4 x float>, metadata)
+
+define <8 x float> @roundeven_v8f32(<8 x float> %x) {
+; CHECK-LABEL: roundeven_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x float> @llvm.experimental.constrained.roundeven.v8f32(<8 x float> %x, metadata !"fpexcept.strict")
+  ret <8 x float> %a
+}
+declare <8 x float> @llvm.experimental.constrained.roundeven.v8f32(<8 x float>, metadata)
+
+define <16 x float> @roundeven_v16f32(<16 x float> %x) {
+; CHECK-LABEL: roundeven_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x float> @llvm.experimental.constrained.roundeven.v16f32(<16 x float> %x, metadata !"fpexcept.strict")
+  ret <16 x float> %a
+}
+declare <16 x float> @llvm.experimental.constrained.roundeven.v16f32(<16 x float>, metadata)
+
+define <1 x double> @roundeven_v1f64(<1 x double> %x) {
+; CHECK-LABEL: roundeven_v1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x double> @llvm.experimental.constrained.roundeven.v1f64(<1 x double> %x, metadata !"fpexcept.strict")
+  ret <1 x double> %a
+}
+declare <1 x double> @llvm.experimental.constrained.roundeven.v1f64(<1 x double>, metadata)
+
+define <2 x double> @roundeven_v2f64(<2 x double> %x) {
+; CHECK-LABEL: roundeven_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x double> @llvm.experimental.constrained.roundeven.v2f64(<2 x double> %x, metadata !"fpexcept.strict")
+  ret <2 x double> %a
+}
+declare <2 x double> @llvm.experimental.constrained.roundeven.v2f64(<2 x double>, metadata)
+
+define <4 x double> @roundeven_v4f64(<4 x double> %x) {
+; CHECK-LABEL: roundeven_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x double> @llvm.experimental.constrained.roundeven.v4f64(<4 x double> %x, metadata !"fpexcept.strict")
+  ret <4 x double> %a
+}
+declare <4 x double> @llvm.experimental.constrained.roundeven.v4f64(<4 x double>, metadata)
+
+define <8 x double> @roundeven_v8f64(<8 x double> %x) {
+; CHECK-LABEL: roundeven_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x double> @llvm.experimental.constrained.roundeven.v8f64(<8 x double> %x, metadata !"fpexcept.strict")
+  ret <8 x double> %a
+}
+declare <8 x double> @llvm.experimental.constrained.roundeven.v8f64(<8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/fixed-vectors-ftrunc-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+define <1 x half> @trunc_v1f16(<1 x half> %x) {
+; CHECK-LABEL: trunc_v1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x half> @llvm.experimental.constrained.trunc.v1f16(<1 x half> %x, metadata !"fpexcept.strict")
+  ret <1 x half> %a
+}
+declare <1 x half> @llvm.experimental.constrained.trunc.v1f16(<1 x half>, metadata)
+
+define <2 x half> @trunc_v2f16(<2 x half> %x) {
+; CHECK-LABEL: trunc_v2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x half> @llvm.experimental.constrained.trunc.v2f16(<2 x half> %x, metadata !"fpexcept.strict")
+  ret <2 x half> %a
+}
+declare <2 x half> @llvm.experimental.constrained.trunc.v2f16(<2 x half>, metadata)
+
+define <4 x half> @trunc_v4f16(<4 x half> %x) {
+; CHECK-LABEL: trunc_v4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x half> @llvm.experimental.constrained.trunc.v4f16(<4 x half> %x, metadata !"fpexcept.strict")
+  ret <4 x half> %a
+}
+declare <4 x half> @llvm.experimental.constrained.trunc.v4f16(<4 x half>, metadata)
+
+define <8 x half> @trunc_v8f16(<8 x half> %x) {
+; CHECK-LABEL: trunc_v8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x half> @llvm.experimental.constrained.trunc.v8f16(<8 x half> %x, metadata !"fpexcept.strict")
+  ret <8 x half> %a
+}
+declare <8 x half> @llvm.experimental.constrained.trunc.v8f16(<8 x half>, metadata)
+
+define <16 x half> @trunc_v16f16(<16 x half> %x) {
+; CHECK-LABEL: trunc_v16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x half> @llvm.experimental.constrained.trunc.v16f16(<16 x half> %x, metadata !"fpexcept.strict")
+  ret <16 x half> %a
+}
+declare <16 x half> @llvm.experimental.constrained.trunc.v16f16(<16 x half>, metadata)
+
+define <32 x half> @trunc_v32f16(<32 x half> %x) {
+; CHECK-LABEL: trunc_v32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:    vsetvli zero, a0, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <32 x half> @llvm.experimental.constrained.trunc.v32f16(<32 x half> %x, metadata !"fpexcept.strict")
+  ret <32 x half> %a
+}
+declare <32 x half> @llvm.experimental.constrained.trunc.v32f16(<32 x half>, metadata)
+
+define <1 x float> @trunc_v1f32(<1 x float> %x) {
+; CHECK-LABEL: trunc_v1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x float> @llvm.experimental.constrained.trunc.v1f32(<1 x float> %x, metadata !"fpexcept.strict")
+  ret <1 x float> %a
+}
+declare <1 x float> @llvm.experimental.constrained.trunc.v1f32(<1 x float>, metadata)
+
+define <2 x float> @trunc_v2f32(<2 x float> %x) {
+; CHECK-LABEL: trunc_v2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x float> @llvm.experimental.constrained.trunc.v2f32(<2 x float> %x, metadata !"fpexcept.strict")
+  ret <2 x float> %a
+}
+declare <2 x float> @llvm.experimental.constrained.trunc.v2f32(<2 x float>, metadata)
+
+define <4 x float> @trunc_v4f32(<4 x float> %x) {
+; CHECK-LABEL: trunc_v4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x float> @llvm.experimental.constrained.trunc.v4f32(<4 x float> %x, metadata !"fpexcept.strict")
+  ret <4 x float> %a
+}
+declare <4 x float> @llvm.experimental.constrained.trunc.v4f32(<4 x float>, metadata)
+
+define <8 x float> @trunc_v8f32(<8 x float> %x) {
+; CHECK-LABEL: trunc_v8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x float> @llvm.experimental.constrained.trunc.v8f32(<8 x float> %x, metadata !"fpexcept.strict")
+  ret <8 x float> %a
+}
+declare <8 x float> @llvm.experimental.constrained.trunc.v8f32(<8 x float>, metadata)
+
+define <16 x float> @trunc_v16f32(<16 x float> %x) {
+; CHECK-LABEL: trunc_v16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 16, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <16 x float> @llvm.experimental.constrained.trunc.v16f32(<16 x float> %x, metadata !"fpexcept.strict")
+  ret <16 x float> %a
+}
+declare <16 x float> @llvm.experimental.constrained.trunc.v16f32(<16 x float>, metadata)
+
+define <1 x double> @trunc_v1f64(<1 x double> %x) {
+; CHECK-LABEL: trunc_v1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 1, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <1 x double> @llvm.experimental.constrained.trunc.v1f64(<1 x double> %x, metadata !"fpexcept.strict")
+  ret <1 x double> %a
+}
+declare <1 x double> @llvm.experimental.constrained.trunc.v1f64(<1 x double>, metadata)
+
+define <2 x double> @trunc_v2f64(<2 x double> %x) {
+; CHECK-LABEL: trunc_v2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 2, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v9, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v9, v9
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <2 x double> @llvm.experimental.constrained.trunc.v2f64(<2 x double> %x, metadata !"fpexcept.strict")
+  ret <2 x double> %a
+}
+declare <2 x double> @llvm.experimental.constrained.trunc.v2f64(<2 x double>, metadata)
+
+define <4 x double> @trunc_v4f64(<4 x double> %x) {
+; CHECK-LABEL: trunc_v4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v10, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v10, v10
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <4 x double> @llvm.experimental.constrained.trunc.v4f64(<4 x double> %x, metadata !"fpexcept.strict")
+  ret <4 x double> %a
+}
+declare <4 x double> @llvm.experimental.constrained.trunc.v4f64(<4 x double>, metadata)
+
+define <8 x double> @trunc_v8f64(<8 x double> %x) {
+; CHECK-LABEL: trunc_v8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetivli zero, 8, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v12, v8, v8
+; CHECK-NEXT:    vmor.mm v0, v12, v12
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <8 x double> @llvm.experimental.constrained.trunc.v8f64(<8 x double> %x, metadata !"fpexcept.strict")
+  ret <8 x double> %a
+}
+declare <8 x double> @llvm.experimental.constrained.trunc.v8f64(<8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/fnearbyint-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+declare <vscale x 1 x half> @llvm.experimental.constrained.nearbyint.nxv1f16(<vscale x 1 x half>, metadata, metadata)
+
+define <vscale x 1 x half> @nearbyint_nxv1f16(<vscale x 1 x half> %v) {
+; CHECK-LABEL: nearbyint_nxv1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vfcvt.x.f.v v9, v9
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 1 x half> @llvm.experimental.constrained.nearbyint.nxv1f16(<vscale x 1 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 1 x half> %r
+}
+
+declare <vscale x 2 x half> @llvm.experimental.constrained.nearbyint.nxv2f16(<vscale x 2 x half>, metadata, metadata)
+
+define <vscale x 2 x half> @nearbyint_nxv2f16(<vscale x 2 x half> %v) {
+; CHECK-LABEL: nearbyint_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vfcvt.x.f.v v9, v9
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 2 x half> @llvm.experimental.constrained.nearbyint.nxv2f16(<vscale x 2 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 2 x half> %r
+}
+
+declare <vscale x 4 x half> @llvm.experimental.constrained.nearbyint.nxv4f16(<vscale x 4 x half>, metadata, metadata)
+
+define <vscale x 4 x half> @nearbyint_nxv4f16(<vscale x 4 x half> %v) {
+; CHECK-LABEL: nearbyint_nxv4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vfcvt.x.f.v v9, v9
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 4 x half> @llvm.experimental.constrained.nearbyint.nxv4f16(<vscale x 4 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 4 x half> %r
+}
+
+declare <vscale x 8 x half> @llvm.experimental.constrained.nearbyint.nxv8f16(<vscale x 8 x half>, metadata, metadata)
+
+define <vscale x 8 x half> @nearbyint_nxv8f16(<vscale x 8 x half> %v) {
+; CHECK-LABEL: nearbyint_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vfcvt.x.f.v v10, v10
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 8 x half> @llvm.experimental.constrained.nearbyint.nxv8f16(<vscale x 8 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 8 x half> %r
+}
+
+declare <vscale x 16 x half> @llvm.experimental.constrained.nearbyint.nxv16f16(<vscale x 16 x half>, metadata, metadata)
+
+define <vscale x 16 x half> @nearbyint_nxv16f16(<vscale x 16 x half> %v) {
+; CHECK-LABEL: nearbyint_nxv16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vfcvt.x.f.v v12, v12
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 16 x half> @llvm.experimental.constrained.nearbyint.nxv16f16(<vscale x 16 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 16 x half> %r
+}
+
+declare <vscale x 32 x half> @llvm.experimental.constrained.nearbyint.nxv32f16(<vscale x 32 x half>, metadata, metadata)
+
+define <vscale x 32 x half> @nearbyint_nxv32f16(<vscale x 32 x half> %v) {
+; CHECK-LABEL: nearbyint_nxv32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, ma
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vfcvt.x.f.v v16, v16
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 32 x half> @llvm.experimental.constrained.nearbyint.nxv32f16(<vscale x 32 x half> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 32 x half> %r
+}
+
+declare <vscale x 1 x float> @llvm.experimental.constrained.nearbyint.nxv1f32(<vscale x 1 x float>, metadata, metadata)
+
+define <vscale x 1 x float> @nearbyint_nxv1f32(<vscale x 1 x float> %v) {
+; CHECK-LABEL: nearbyint_nxv1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vfcvt.x.f.v v9, v9
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 1 x float> @llvm.experimental.constrained.nearbyint.nxv1f32(<vscale x 1 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 1 x float> %r
+}
+
+declare <vscale x 2 x float> @llvm.experimental.constrained.nearbyint.nxv2f32(<vscale x 2 x float>, metadata, metadata)
+
+define <vscale x 2 x float> @nearbyint_nxv2f32(<vscale x 2 x float> %v) {
+; CHECK-LABEL: nearbyint_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vfcvt.x.f.v v9, v9
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 2 x float> @llvm.experimental.constrained.nearbyint.nxv2f32(<vscale x 2 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 2 x float> %r
+}
+
+declare <vscale x 4 x float> @llvm.experimental.constrained.nearbyint.nxv4f32(<vscale x 4 x float>, metadata, metadata)
+
+define <vscale x 4 x float> @nearbyint_nxv4f32(<vscale x 4 x float> %v) {
+; CHECK-LABEL: nearbyint_nxv4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vfcvt.x.f.v v10, v10
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 4 x float> @llvm.experimental.constrained.nearbyint.nxv4f32(<vscale x 4 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 4 x float> %r
+}
+
+declare <vscale x 8 x float> @llvm.experimental.constrained.nearbyint.nxv8f32(<vscale x 8 x float>, metadata, metadata)
+
+define <vscale x 8 x float> @nearbyint_nxv8f32(<vscale x 8 x float> %v) {
+; CHECK-LABEL: nearbyint_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vfcvt.x.f.v v12, v12
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 8 x float> @llvm.experimental.constrained.nearbyint.nxv8f32(<vscale x 8 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 8 x float> %r
+}
+
+declare <vscale x 16 x float> @llvm.experimental.constrained.nearbyint.nxv16f32(<vscale x 16 x float>, metadata, metadata)
+
+define <vscale x 16 x float> @nearbyint_nxv16f32(<vscale x 16 x float> %v) {
+; CHECK-LABEL: nearbyint_nxv16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, ma
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vfcvt.x.f.v v16, v16
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 16 x float> @llvm.experimental.constrained.nearbyint.nxv16f32(<vscale x 16 x float> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 16 x float> %r
+}
+
+declare <vscale x 1 x double> @llvm.experimental.constrained.nearbyint.nxv1f64(<vscale x 1 x double>, metadata, metadata)
+
+define <vscale x 1 x double> @nearbyint_nxv1f64(<vscale x 1 x double> %v) {
+; CHECK-LABEL: nearbyint_nxv1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vfcvt.x.f.v v9, v9
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 1 x double> @llvm.experimental.constrained.nearbyint.nxv1f64(<vscale x 1 x double> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 1 x double> %r
+}
+
+declare <vscale x 2 x double> @llvm.experimental.constrained.nearbyint.nxv2f64(<vscale x 2 x double>, metadata, metadata)
+
+define <vscale x 2 x double> @nearbyint_nxv2f64(<vscale x 2 x double> %v) {
+; CHECK-LABEL: nearbyint_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vfcvt.x.f.v v10, v10
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 2 x double> @llvm.experimental.constrained.nearbyint.nxv2f64(<vscale x 2 x double> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 2 x double> %r
+}
+
+declare <vscale x 4 x double> @llvm.experimental.constrained.nearbyint.nxv4f64(<vscale x 4 x double>, metadata, metadata)
+
+define <vscale x 4 x double> @nearbyint_nxv4f64(<vscale x 4 x double> %v) {
+; CHECK-LABEL: nearbyint_nxv4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vfcvt.x.f.v v12, v12
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 4 x double> @llvm.experimental.constrained.nearbyint.nxv4f64(<vscale x 4 x double> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 4 x double> %r
+}
+
+declare <vscale x 8 x double> @llvm.experimental.constrained.nearbyint.nxv8f64(<vscale x 8 x double>, metadata, metadata)
+
+define <vscale x 8 x double> @nearbyint_nxv8f64(<vscale x 8 x double> %v) {
+; CHECK-LABEL: nearbyint_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vfcvt.x.f.v v16, v16
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8
+; CHECK-NEXT:    ret
+  %r = call <vscale x 8 x double> @llvm.experimental.constrained.nearbyint.nxv8f64(<vscale x 8 x double> %v, metadata !"round.dynamic", metadata !"fpexcept.strict")
+  ret <vscale x 8 x double> %r
+}

llvm/test/CodeGen/RISCV/rvv/fround-costrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+; This file tests the code generation for `llvm.experimental.constrained.round.*` on scalable vector type.
+
+define <vscale x 1 x half> @round_nxv1f16(<vscale x 1 x half> %x) {
+; CHECK-LABEL: round_nxv1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x half> @llvm.experimental.constrained.round.nxv1f16(<vscale x 1 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x half> %a
+}
+declare <vscale x 1 x half> @llvm.experimental.constrained.round.nxv1f16(<vscale x 1 x half>, metadata)
+
+define <vscale x 2 x half> @round_nxv2f16(<vscale x 2 x half> %x) {
+; CHECK-LABEL: round_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x half> @llvm.experimental.constrained.round.nxv2f16(<vscale x 2 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x half> %a
+}
+declare <vscale x 2 x half> @llvm.experimental.constrained.round.nxv2f16(<vscale x 2 x half>, metadata)
+
+define <vscale x 4 x half> @round_nxv4f16(<vscale x 4 x half> %x) {
+; CHECK-LABEL: round_nxv4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x half> @llvm.experimental.constrained.round.nxv4f16(<vscale x 4 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x half> %a
+}
+declare <vscale x 4 x half> @llvm.experimental.constrained.round.nxv4f16(<vscale x 4 x half>, metadata)
+
+define <vscale x 8 x half> @round_nxv8f16(<vscale x 8 x half> %x) {
+; CHECK-LABEL: round_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x half> @llvm.experimental.constrained.round.nxv8f16(<vscale x 8 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x half> %a
+}
+declare <vscale x 8 x half> @llvm.experimental.constrained.round.nxv8f16(<vscale x 8 x half>, metadata)
+
+define <vscale x 16 x half> @round_nxv16f16(<vscale x 16 x half> %x) {
+; CHECK-LABEL: round_nxv16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x half> @llvm.experimental.constrained.round.nxv16f16(<vscale x 16 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x half> %a
+}
+declare <vscale x 16 x half> @llvm.experimental.constrained.round.nxv16f16(<vscale x 16 x half>, metadata)
+
+define <vscale x 32 x half> @round_nxv32f16(<vscale x 32 x half> %x) {
+; CHECK-LABEL: round_nxv32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 32 x half> @llvm.experimental.constrained.round.nxv32f16(<vscale x 32 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 32 x half> %a
+}
+declare <vscale x 32 x half> @llvm.experimental.constrained.round.nxv32f16(<vscale x 32 x half>, metadata)
+
+define <vscale x 1 x float> @round_nxv1f32(<vscale x 1 x float> %x) {
+; CHECK-LABEL: round_nxv1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x float> @llvm.experimental.constrained.round.nxv1f32(<vscale x 1 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x float> %a
+}
+declare <vscale x 1 x float> @llvm.experimental.constrained.round.nxv1f32(<vscale x 1 x float>, metadata)
+
+define <vscale x 2 x float> @round_nxv2f32(<vscale x 2 x float> %x) {
+; CHECK-LABEL: round_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x float> @llvm.experimental.constrained.round.nxv2f32(<vscale x 2 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x float> %a
+}
+declare <vscale x 2 x float> @llvm.experimental.constrained.round.nxv2f32(<vscale x 2 x float>, metadata)
+
+define <vscale x 4 x float> @round_nxv4f32(<vscale x 4 x float> %x) {
+; CHECK-LABEL: round_nxv4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x float> @llvm.experimental.constrained.round.nxv4f32(<vscale x 4 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x float> %a
+}
+declare <vscale x 4 x float> @llvm.experimental.constrained.round.nxv4f32(<vscale x 4 x float>, metadata)
+
+define <vscale x 8 x float> @round_nxv8f32(<vscale x 8 x float> %x) {
+; CHECK-LABEL: round_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x float> @llvm.experimental.constrained.round.nxv8f32(<vscale x 8 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x float> %a
+}
+declare <vscale x 8 x float> @llvm.experimental.constrained.round.nxv8f32(<vscale x 8 x float>, metadata)
+
+define <vscale x 16 x float> @round_nxv16f32(<vscale x 16 x float> %x) {
+; CHECK-LABEL: round_nxv16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x float> @llvm.experimental.constrained.round.nxv16f32(<vscale x 16 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x float> %a
+}
+declare <vscale x 16 x float> @llvm.experimental.constrained.round.nxv16f32(<vscale x 16 x float>, metadata)
+
+define <vscale x 1 x double> @round_nxv1f64(<vscale x 1 x double> %x) {
+; CHECK-LABEL: round_nxv1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x double> @llvm.experimental.constrained.round.nxv1f64(<vscale x 1 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x double> %a
+}
+declare <vscale x 1 x double> @llvm.experimental.constrained.round.nxv1f64(<vscale x 1 x double>, metadata)
+
+define <vscale x 2 x double> @round_nxv2f64(<vscale x 2 x double> %x) {
+; CHECK-LABEL: round_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x double> @llvm.experimental.constrained.round.nxv2f64(<vscale x 2 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x double> %a
+}
+declare <vscale x 2 x double> @llvm.experimental.constrained.round.nxv2f64(<vscale x 2 x double>, metadata)
+
+define <vscale x 4 x double> @round_nxv4f64(<vscale x 4 x double> %x) {
+; CHECK-LABEL: round_nxv4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x double> @llvm.experimental.constrained.round.nxv4f64(<vscale x 4 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x double> %a
+}
+declare <vscale x 4 x double> @llvm.experimental.constrained.round.nxv4f64(<vscale x 4 x double>, metadata)
+
+define <vscale x 8 x double> @round_nxv8f64(<vscale x 8 x double> %x) {
+; CHECK-LABEL: round_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 4
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x double> @llvm.experimental.constrained.round.nxv8f64(<vscale x 8 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x double> %a
+}
+declare <vscale x 8 x double> @llvm.experimental.constrained.round.nxv8f64(<vscale x 8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/froundeven-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+; This file tests the code generation for `llvm.experimental.constrained.roundeven.*` on scalable vector type.
+
+define <vscale x 1 x half> @roundeven_nxv1f16(<vscale x 1 x half> %x) {
+; CHECK-LABEL: roundeven_nxv1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x half> @llvm.experimental.constrained.roundeven.nxv1f16(<vscale x 1 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x half> %a
+}
+declare <vscale x 1 x half> @llvm.experimental.constrained.roundeven.nxv1f16(<vscale x 1 x half>, metadata)
+
+define <vscale x 2 x half> @roundeven_nxv2f16(<vscale x 2 x half> %x) {
+; CHECK-LABEL: roundeven_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x half> @llvm.experimental.constrained.roundeven.nxv2f16(<vscale x 2 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x half> %a
+}
+declare <vscale x 2 x half> @llvm.experimental.constrained.roundeven.nxv2f16(<vscale x 2 x half>, metadata)
+
+define <vscale x 4 x half> @roundeven_nxv4f16(<vscale x 4 x half> %x) {
+; CHECK-LABEL: roundeven_nxv4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x half> @llvm.experimental.constrained.roundeven.nxv4f16(<vscale x 4 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x half> %a
+}
+declare <vscale x 4 x half> @llvm.experimental.constrained.roundeven.nxv4f16(<vscale x 4 x half>, metadata)
+
+define <vscale x 8 x half> @roundeven_nxv8f16(<vscale x 8 x half> %x) {
+; CHECK-LABEL: roundeven_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x half> @llvm.experimental.constrained.roundeven.nxv8f16(<vscale x 8 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x half> %a
+}
+declare <vscale x 8 x half> @llvm.experimental.constrained.roundeven.nxv8f16(<vscale x 8 x half>, metadata)
+
+define <vscale x 16 x half> @roundeven_nxv16f16(<vscale x 16 x half> %x) {
+; CHECK-LABEL: roundeven_nxv16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x half> @llvm.experimental.constrained.roundeven.nxv16f16(<vscale x 16 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x half> %a
+}
+declare <vscale x 16 x half> @llvm.experimental.constrained.roundeven.nxv16f16(<vscale x 16 x half>, metadata)
+
+define <vscale x 32 x half> @roundeven_nxv32f16(<vscale x 32 x half> %x) {
+; CHECK-LABEL: roundeven_nxv32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 32 x half> @llvm.experimental.constrained.roundeven.nxv32f16(<vscale x 32 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 32 x half> %a
+}
+declare <vscale x 32 x half> @llvm.experimental.constrained.roundeven.nxv32f16(<vscale x 32 x half>, metadata)
+
+define <vscale x 1 x float> @roundeven_nxv1f32(<vscale x 1 x float> %x) {
+; CHECK-LABEL: roundeven_nxv1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x float> @llvm.experimental.constrained.roundeven.nxv1f32(<vscale x 1 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x float> %a
+}
+declare <vscale x 1 x float> @llvm.experimental.constrained.roundeven.nxv1f32(<vscale x 1 x float>, metadata)
+
+define <vscale x 2 x float> @roundeven_nxv2f32(<vscale x 2 x float> %x) {
+; CHECK-LABEL: roundeven_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x float> @llvm.experimental.constrained.roundeven.nxv2f32(<vscale x 2 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x float> %a
+}
+declare <vscale x 2 x float> @llvm.experimental.constrained.roundeven.nxv2f32(<vscale x 2 x float>, metadata)
+
+define <vscale x 4 x float> @roundeven_nxv4f32(<vscale x 4 x float> %x) {
+; CHECK-LABEL: roundeven_nxv4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x float> @llvm.experimental.constrained.roundeven.nxv4f32(<vscale x 4 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x float> %a
+}
+declare <vscale x 4 x float> @llvm.experimental.constrained.roundeven.nxv4f32(<vscale x 4 x float>, metadata)
+
+define <vscale x 8 x float> @roundeven_nxv8f32(<vscale x 8 x float> %x) {
+; CHECK-LABEL: roundeven_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x float> @llvm.experimental.constrained.roundeven.nxv8f32(<vscale x 8 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x float> %a
+}
+declare <vscale x 8 x float> @llvm.experimental.constrained.roundeven.nxv8f32(<vscale x 8 x float>, metadata)
+
+define <vscale x 16 x float> @roundeven_nxv16f32(<vscale x 16 x float> %x) {
+; CHECK-LABEL: roundeven_nxv16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x float> @llvm.experimental.constrained.roundeven.nxv16f32(<vscale x 16 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x float> %a
+}
+declare <vscale x 16 x float> @llvm.experimental.constrained.roundeven.nxv16f32(<vscale x 16 x float>, metadata)
+
+define <vscale x 1 x double> @roundeven_nxv1f64(<vscale x 1 x double> %x) {
+; CHECK-LABEL: roundeven_nxv1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x double> @llvm.experimental.constrained.roundeven.nxv1f64(<vscale x 1 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x double> %a
+}
+declare <vscale x 1 x double> @llvm.experimental.constrained.roundeven.nxv1f64(<vscale x 1 x double>, metadata)
+
+define <vscale x 2 x double> @roundeven_nxv2f64(<vscale x 2 x double> %x) {
+; CHECK-LABEL: roundeven_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x double> @llvm.experimental.constrained.roundeven.nxv2f64(<vscale x 2 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x double> %a
+}
+declare <vscale x 2 x double> @llvm.experimental.constrained.roundeven.nxv2f64(<vscale x 2 x double>, metadata)
+
+define <vscale x 4 x double> @roundeven_nxv4f64(<vscale x 4 x double> %x) {
+; CHECK-LABEL: roundeven_nxv4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x double> @llvm.experimental.constrained.roundeven.nxv4f64(<vscale x 4 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x double> %a
+}
+declare <vscale x 4 x double> @llvm.experimental.constrained.roundeven.nxv4f64(<vscale x 4 x double>, metadata)
+
+define <vscale x 8 x double> @roundeven_nxv8f64(<vscale x 8 x double> %x) {
+; CHECK-LABEL: roundeven_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    fsrmi a0, 0
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vfcvt.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    fsrm a0
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x double> @llvm.experimental.constrained.roundeven.nxv8f64(<vscale x 8 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x double> %a
+}
+declare <vscale x 8 x double> @llvm.experimental.constrained.roundeven.nxv8f64(<vscale x 8 x double>, metadata)

llvm/test/CodeGen/RISCV/rvv/ftrunc-constrained-sdnode.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=ilp32d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zfh,+experimental-zvfh,+v -target-abi=lp64d \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s
+
+define <vscale x 1 x half> @trunc_nxv1f16(<vscale x 1 x half> %x) {
+; CHECK-LABEL: trunc_nxv1f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI0_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI0_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x half> @llvm.experimental.constrained.trunc.nxv1f16(<vscale x 1 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x half> %a
+}
+declare <vscale x 1 x half> @llvm.experimental.constrained.trunc.nxv1f16(<vscale x 1 x half>, metadata)
+
+define <vscale x 2 x half> @trunc_nxv2f16(<vscale x 2 x half> %x) {
+; CHECK-LABEL: trunc_nxv2f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI1_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI1_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x half> @llvm.experimental.constrained.trunc.nxv2f16(<vscale x 2 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x half> %a
+}
+declare <vscale x 2 x half> @llvm.experimental.constrained.trunc.nxv2f16(<vscale x 2 x half>, metadata)
+
+define <vscale x 4 x half> @trunc_nxv4f16(<vscale x 4 x half> %x) {
+; CHECK-LABEL: trunc_nxv4f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI2_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI2_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x half> @llvm.experimental.constrained.trunc.nxv4f16(<vscale x 4 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x half> %a
+}
+declare <vscale x 4 x half> @llvm.experimental.constrained.trunc.nxv4f16(<vscale x 4 x half>, metadata)
+
+define <vscale x 8 x half> @trunc_nxv8f16(<vscale x 8 x half> %x) {
+; CHECK-LABEL: trunc_nxv8f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI3_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI3_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x half> @llvm.experimental.constrained.trunc.nxv8f16(<vscale x 8 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x half> %a
+}
+declare <vscale x 8 x half> @llvm.experimental.constrained.trunc.nxv8f16(<vscale x 8 x half>, metadata)
+
+define <vscale x 16 x half> @trunc_nxv16f16(<vscale x 16 x half> %x) {
+; CHECK-LABEL: trunc_nxv16f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI4_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI4_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x half> @llvm.experimental.constrained.trunc.nxv16f16(<vscale x 16 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x half> %a
+}
+declare <vscale x 16 x half> @llvm.experimental.constrained.trunc.nxv16f16(<vscale x 16 x half>, metadata)
+
+define <vscale x 32 x half> @trunc_nxv32f16(<vscale x 32 x half> %x) {
+; CHECK-LABEL: trunc_nxv32f16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI5_0)
+; CHECK-NEXT:    flh fa5, %lo(.LCPI5_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 32 x half> @llvm.experimental.constrained.trunc.nxv32f16(<vscale x 32 x half> %x, metadata !"fpexcept.strict")
+  ret <vscale x 32 x half> %a
+}
+declare <vscale x 32 x half> @llvm.experimental.constrained.trunc.nxv32f16(<vscale x 32 x half>, metadata)
+
+define <vscale x 1 x float> @trunc_nxv1f32(<vscale x 1 x float> %x) {
+; CHECK-LABEL: trunc_nxv1f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x float> @llvm.experimental.constrained.trunc.nxv1f32(<vscale x 1 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x float> %a
+}
+declare <vscale x 1 x float> @llvm.experimental.constrained.trunc.nxv1f32(<vscale x 1 x float>, metadata)
+
+define <vscale x 2 x float> @trunc_nxv2f32(<vscale x 2 x float> %x) {
+; CHECK-LABEL: trunc_nxv2f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x float> @llvm.experimental.constrained.trunc.nxv2f32(<vscale x 2 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x float> %a
+}
+declare <vscale x 2 x float> @llvm.experimental.constrained.trunc.nxv2f32(<vscale x 2 x float>, metadata)
+
+define <vscale x 4 x float> @trunc_nxv4f32(<vscale x 4 x float> %x) {
+; CHECK-LABEL: trunc_nxv4f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x float> @llvm.experimental.constrained.trunc.nxv4f32(<vscale x 4 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x float> %a
+}
+declare <vscale x 4 x float> @llvm.experimental.constrained.trunc.nxv4f32(<vscale x 4 x float>, metadata)
+
+define <vscale x 8 x float> @trunc_nxv8f32(<vscale x 8 x float> %x) {
+; CHECK-LABEL: trunc_nxv8f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x float> @llvm.experimental.constrained.trunc.nxv8f32(<vscale x 8 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x float> %a
+}
+declare <vscale x 8 x float> @llvm.experimental.constrained.trunc.nxv8f32(<vscale x 8 x float>, metadata)
+
+define <vscale x 16 x float> @trunc_nxv16f32(<vscale x 16 x float> %x) {
+; CHECK-LABEL: trunc_nxv16f32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    lui a0, 307200
+; CHECK-NEXT:    fmv.w.x fa5, a0
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 16 x float> @llvm.experimental.constrained.trunc.nxv16f32(<vscale x 16 x float> %x, metadata !"fpexcept.strict")
+  ret <vscale x 16 x float> %a
+}
+declare <vscale x 16 x float> @llvm.experimental.constrained.trunc.nxv16f32(<vscale x 16 x float>, metadata)
+
+define <vscale x 1 x double> @trunc_nxv1f64(<vscale x 1 x double> %x) {
+; CHECK-LABEL: trunc_nxv1f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI11_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI11_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v9, v8
+; CHECK-NEXT:    vmflt.vf v0, v9, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v9, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v9, v9, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v9, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 1 x double> @llvm.experimental.constrained.trunc.nxv1f64(<vscale x 1 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 1 x double> %a
+}
+declare <vscale x 1 x double> @llvm.experimental.constrained.trunc.nxv1f64(<vscale x 1 x double>, metadata)
+
+define <vscale x 2 x double> @trunc_nxv2f64(<vscale x 2 x double> %x) {
+; CHECK-LABEL: trunc_nxv2f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI12_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI12_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v10, v8
+; CHECK-NEXT:    vmflt.vf v0, v10, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v10, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v10, v10, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v10, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 2 x double> @llvm.experimental.constrained.trunc.nxv2f64(<vscale x 2 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 2 x double> %a
+}
+declare <vscale x 2 x double> @llvm.experimental.constrained.trunc.nxv2f64(<vscale x 2 x double>, metadata)
+
+define <vscale x 4 x double> @trunc_nxv4f64(<vscale x 4 x double> %x) {
+; CHECK-LABEL: trunc_nxv4f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI13_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI13_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    vmflt.vf v0, v12, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v12, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v12, v12, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v12, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 4 x double> @llvm.experimental.constrained.trunc.nxv4f64(<vscale x 4 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 4 x double> %a
+}
+declare <vscale x 4 x double> @llvm.experimental.constrained.trunc.nxv4f64(<vscale x 4 x double>, metadata)
+
+define <vscale x 8 x double> @trunc_nxv8f64(<vscale x 8 x double> %x) {
+; CHECK-LABEL: trunc_nxv8f64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vmfne.vv v0, v8, v8
+; CHECK-NEXT:    lui a0, %hi(.LCPI14_0)
+; CHECK-NEXT:    fld fa5, %lo(.LCPI14_0)(a0)
+; CHECK-NEXT:    vfadd.vv v8, v8, v8, v0.t
+; CHECK-NEXT:    vfabs.v v16, v8
+; CHECK-NEXT:    vmflt.vf v0, v16, fa5
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, ma
+; CHECK-NEXT:    vfcvt.rtz.x.f.v v16, v8, v0.t
+; CHECK-NEXT:    vfcvt.f.x.v v16, v16, v0.t
+; CHECK-NEXT:    vsetvli zero, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vfsgnj.vv v8, v16, v8, v0.t
+; CHECK-NEXT:    ret
+  %a = call <vscale x 8 x double> @llvm.experimental.constrained.trunc.nxv8f64(<vscale x 8 x double> %x, metadata !"fpexcept.strict")
+  ret <vscale x 8 x double> %a
+}
+declare <vscale x 8 x double> @llvm.experimental.constrained.trunc.nxv8f64(<vscale x 8 x double>, metadata)