[RISCV] Define different pseudo instructions for different FPR.

[HsiangKai]

Summary

When spilling, the spill size will depend on the size of register class.
For .vf vector instructions, it may spill the floating point scalar
argument. In order to use the correct load/store instructions for
spilling, we need to provide the correct floating point register class
for the .vf vector pseudo instructions.

In this commit, we define the .vf pseudo instructions as three
different kinds of pseudo instructions for half/float/double. For
example, PseudoVFADD_M1 will become as PseudoVFADD_F16_M1,
PseudoVFADD_F32_M1, and PseudoVFADD_F64_M1.

[HsiangKai]

• Avoid defining redundant pseudo instructions for widening instructions.

[craig.topper]

L:GTM with that one comment.

llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td

42

Please mark this with a FIXME to cleanup. Fraser or I can take care of it as a follow up.

[frasercrmck]

llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td

4195

How come you're not using GetScalarSuffix in these cases too?

llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td

42

Yeah, no problem, I can take a look.

[craig.topper]

Is this patch really dependent on frame lowering and spilling or can we land this before the branch tomorrow?

It's looking like we may have to cherry pick frame lowering and spilling after the branch. So if this patch isn't dependent we should go ahead so we don't have to cherry pick it too.

[rogfer01]

We're multiplying by 3 the _VF pseudos but to be honest I don't see any better alternative.

[craig.topper]

llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td

42

@frasercrmck I'm looking at this and might be proposing changes to this patch as well.

[HsiangKai]

In D95234#2520781, @craig.topper wrote:

Is this patch really dependent on frame lowering and spilling or can we land this before the branch tomorrow?

It's looking like we may have to cherry pick frame lowering and spilling after the branch. So if this patch isn't dependent we should go ahead so we don't have to cherry pick it too.

I will land this patch first.

llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td

4195

getScalarSuffix() may need to be defined in more general way.

llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td

 def V_MF4 : LMULInfo<0b110, VR, VR,          VR,/*NoVReg*/VR,/*NoVReg*/VR, "MF4">;
 def V_MF2 : LMULInfo<0b111, VR, VR,          VR,          VR,/*NoVReg*/VR, "MF2">;
 
 // Used to iterate over all possible LMULs.
 def MxList {
   list<LMULInfo> m = [V_MF8, V_MF4, V_MF2, V_M1, V_M2, V_M4, V_M8];
 }
 
+class FPR_Info<RegisterClass regclass, string fx> {
+  RegisterClass fprclass = regclass;
+  string FX = fx;
+}
+
+def SCALAR_F16 : FPR_Info<FPR16, "F16">;
+def SCALAR_F32 : FPR_Info<FPR32, "F32">;
+def SCALAR_F64 : FPR_Info<FPR64, "F64">;
+
+def FPList {
+  list<FPR_Info> fpinfo = [SCALAR_F16, SCALAR_F32, SCALAR_F64];
+}
+
+class getScalarSuffix<ValueType type> {
+  string suffix = !cond(!eq(type, XLenVT): "VX",
+                        !eq(type, f16): "VF_F16",
+                        !eq(type, f32): "VF_F32",
+                        !eq(type, f64): "VF_F64");
+}
+
+class getWScalarSuffix<ValueType type> {
+  string suffix = !cond(!eq(type, XLenVT): "WX",
+                        !eq(type, f16): "WF_F16",
+                        !eq(type, f32): "WF_F32",
+                        !eq(type, f64): "WF_F64");
+}
+
 class MxSet<int eew> {
   list<LMULInfo> m = !cond(!eq(eew, 8) : [V_MF8, V_MF4, V_MF2, V_M1, V_M2, V_M4, V_M8],
                            !eq(eew, 16) : [V_MF4, V_MF2, V_M1, V_M2, V_M4, V_M8],
                            !eq(eew, 32) : [V_MF2, V_M1, V_M2, V_M4, V_M8],
                            !eq(eew, 64) : [V_M1, V_M2, V_M4, V_M8]);
 }
 
 class NFSet<LMULInfo m> {

 }
 
 // Output pattern for X0 used to represent VLMAX in the pseudo instructions.
 def VLMax : OutPatFrag<(ops), (XLenVT X0)>;
 
 // List of EEW.
 defvar EEWList = [8, 16, 32, 64];
 
-// We only model FPR32 for V instructions in RISCVInstrInfoV.td.
-// FP16/FP32/FP64 registers are alias each other. Convert FPR16 and FPR64
-// to FPR32 for V instructions is enough.
-class ToFPR32<ValueType type, DAGOperand operand, string name> {
-  dag ret = !cond(!eq(!cast<string>(operand), !cast<string>(FPR64)):
-                      (EXTRACT_SUBREG !dag(type, [FPR64], [name]), sub_32),
-                  !eq(!cast<string>(operand), !cast<string>(FPR16)):
-                      (SUBREG_TO_REG (i16 -1), !dag(type, [FPR16], [name]), sub_16),
-                  !eq(1, 1):
-                      !dag(type, [operand], [name]));
-}
-
 class SegRegClass<LMULInfo m, int nf> {
   VReg RC = !cast<VReg>("VRN" # nf # !cond(!eq(m.value, V_MF8.value): V_M1.MX,
                                            !eq(m.value, V_MF4.value): V_M1.MX,
                                            !eq(m.value, V_MF2.value): V_M1.MX,
                                            true: m.MX));
 }
 
 //===----------------------------------------------------------------------===//
string VInst = !subst("_M8", "",
                  !subst("_B1", "",
                  !subst("_B2", "",
                  !subst("_B4", "",
                  !subst("_B8", "",
                  !subst("_B16", "",
                  !subst("_B32", "",
                  !subst("_B64", "",
                  !subst("_MASK", "",
-                 !subst("Pseudo", "", PseudoInst))))))))))))))));
+                 !subst("_F16", "",
+                 !subst("_F32", "",
+                 !subst("_F64", "",
+                 !subst("Pseudo", "", PseudoInst)))))))))))))))))));
 }
 
 class ToLowerCase<string Upper> {
   string L = !subst("VLSEG", "vlseg",
              !subst("VLSSEG", "vlsseg",
              !subst("VSSEG", "vsseg",
              !subst("VSSSEG", "vssseg",
              !subst("VLOXSEG", "vloxseg",
foreach sew = EEWList in {
         defvar emulMX = octuple_to_str<octuple_emul>.ret;
         defvar emul = !cast<LMULInfo>("V_" # emulMX);
         defm _VV : VPseudoBinaryEmul<m.vrclass, m.vrclass, emul.vrclass, m, emul, Constraint>;
       }
     }
   }
 }
 
-multiclass VPseudoBinaryV_VX<bit IsFloat, string Constraint = ""> {
+multiclass VPseudoBinaryV_VX<string Constraint = ""> {
+  foreach m = MxList.m in
+    defm "_VX" : VPseudoBinary<m.vrclass, m.vrclass, GPR, m, Constraint>;
+}
+
+multiclass VPseudoBinaryV_VF<string Constraint = ""> {
   foreach m = MxList.m in
-    defm !if(IsFloat, "_VF", "_VX") : VPseudoBinary<m.vrclass, m.vrclass,
-                             !if(IsFloat, FPR32, GPR), m, Constraint>;
+    foreach f = FPList.fpinfo in
+      defm "_VF_" # f.FX : VPseudoBinary<m.vrclass, m.vrclass,
+                                         f.fprclass, m, Constraint>;
 }
 
 multiclass VPseudoBinaryV_VI<Operand ImmType = simm5, string Constraint = ""> {
   foreach m = MxList.m in
     defm _VI : VPseudoBinary<m.vrclass, m.vrclass, ImmType, m, Constraint>;
 }
 
 multiclass VPseudoBinaryM_MM {

 //   at least 1, and the overlap is in the highest-numbered part of the
 //   destination register group is legal. Otherwise, it is illegal.
 multiclass VPseudoBinaryW_VV {
   foreach m = MxList.m[0-5] in
     defm _VV : VPseudoBinary<m.wvrclass, m.vrclass, m.vrclass, m,
                              "@earlyclobber $rd">;
 }
 
-multiclass VPseudoBinaryW_VX<bit IsFloat> {
+multiclass VPseudoBinaryW_VX {
   foreach m = MxList.m[0-5] in
-    defm !if(IsFloat, "_VF", "_VX") : VPseudoBinary<m.wvrclass, m.vrclass,
-                             !if(IsFloat, FPR32, GPR), m,
+    defm "_VX" : VPseudoBinary<m.wvrclass, m.vrclass, GPR, m,
+                               "@earlyclobber $rd">;
+}
+
+multiclass VPseudoBinaryW_VF {
+  foreach m = MxList.m[0-5] in
+    foreach f = FPList.fpinfo[0-1] in
+      defm "_VF_" # f.FX : VPseudoBinary<m.wvrclass, m.vrclass,
+                                         f.fprclass, m,
                                          "@earlyclobber $rd">;
 }
 
 multiclass VPseudoBinaryW_WV {
   foreach m = MxList.m[0-5] in
     defm _WV : VPseudoBinary<m.wvrclass, m.wvrclass, m.vrclass, m,
                              "@earlyclobber $rd">;
 }
 
-multiclass VPseudoBinaryW_WX<bit IsFloat> {
+multiclass VPseudoBinaryW_WX {
+  foreach m = MxList.m[0-5] in
+    defm "_WX" : VPseudoBinary<m.wvrclass, m.wvrclass, GPR, m,
+                               "@earlyclobber $rd">;
+}
+
+multiclass VPseudoBinaryW_WF {
   foreach m = MxList.m[0-5] in
-    defm !if(IsFloat, "_WF", "_WX") : VPseudoBinary<m.wvrclass, m.wvrclass,
-                             !if(IsFloat, FPR32, GPR), m,
+    foreach f = FPList.fpinfo[0-1] in
+      defm "_WF_" # f.FX : VPseudoBinary<m.wvrclass, m.wvrclass,
+                                         f.fprclass, m,
                                          "@earlyclobber $rd">;
 }
 
 multiclass VPseudoBinaryV_WV {
   foreach m = MxList.m[0-5] in
     defm _WV : VPseudoBinary<m.vrclass, m.wvrclass, m.vrclass, m,
                              "@earlyclobber $rd">;
 }
 
foreach m = MxList.m in
     def "_VV" # !if(CarryIn, "M", "") # "_" # m.MX :
       VPseudoBinaryCarryIn<!if(CarryOut, VR,
                            !if(!and(CarryIn, !not(CarryOut)),
                                GetVRegNoV0<m.vrclass>.R, m.vrclass)),
                            m.vrclass, m.vrclass, m, CarryIn, Constraint>;
 }
 
 multiclass VPseudoBinaryV_XM<bit CarryOut = 0, bit CarryIn = 1,
-                             string Constraint = "", bit IsFloat = 0> {
+                             string Constraint = ""> {
   foreach m = MxList.m in
-    def !if(IsFloat, "_VF", "_VX") # !if(CarryIn, "M", "") # "_" # m.MX :
+    def "_VX" # !if(CarryIn, "M", "") # "_" # m.MX :
       VPseudoBinaryCarryIn<!if(CarryOut, VR,
                            !if(!and(CarryIn, !not(CarryOut)),
                                GetVRegNoV0<m.vrclass>.R, m.vrclass)),
-                           m.vrclass, !if(IsFloat, FPR32, GPR),
-                           m, CarryIn, Constraint>;
+                           m.vrclass, GPR, m, CarryIn, Constraint>;
+}
+
+multiclass VPseudoBinaryV_FM {
+  foreach m = MxList.m in
+    foreach f = FPList.fpinfo in
+      def "_VFM_" # f.FX # "_" # m.MX :
+        VPseudoBinaryCarryIn<GetVRegNoV0<m.vrclass>.R,
+                             m.vrclass, f.fprclass, m, /*CarryIn=*/1, "">;
 }
 
 multiclass VPseudoBinaryV_IM<bit CarryOut = 0, bit CarryIn = 1,
                              string Constraint = ""> {
   foreach m = MxList.m in
     def "_VI" # !if(CarryIn, "M", "") # "_" # m.MX :
       VPseudoBinaryCarryIn<!if(CarryOut, VR,
                            !if(!and(CarryIn, !not(CarryOut)),
                                GetVRegNoV0<m.vrclass>.R, m.vrclass)),
                            m.vrclass, simm5, m, CarryIn, Constraint>;
 }
 
 multiclass VPseudoUnaryV_V_X_I_NoDummyMask {
   foreach m = MxList.m in {
     let VLMul = m.value in {
       def "_V_" # m.MX : VPseudoUnaryNoDummyMask<m.vrclass, m.vrclass>;
       def "_X_" # m.MX : VPseudoUnaryNoDummyMask<m.vrclass, GPR>;
       def "_I_" # m.MX : VPseudoUnaryNoDummyMask<m.vrclass, simm5>;
     }
   }
 }
 
 multiclass VPseudoUnaryV_F_NoDummyMask {
   foreach m = MxList.m in {
+    foreach f = FPList.fpinfo in {
       let VLMul = m.value in {
-      def "_F_" # m.MX : VPseudoUnaryNoDummyMask<m.vrclass, FPR32>;
+        def "_F_" # f.FX # "_" # m.MX : VPseudoUnaryNoDummyMask<m.vrclass, f.fprclass>;
+      }
     }
   }
 }
 
 multiclass VPseudoUnaryV_V {
   foreach m = MxList.m in {
     let VLMul = m.value in {
       def "_V_" # m.MX : VPseudoUnaryNoMask<m.vrclass, m.vrclass>;

 // The source EEW is 8, 16, 32, or 64.
 // When the destination EEW is different from source EEW, we need to use
 // @earlyclobber to avoid the overlap between destination and source registers.
 multiclass VPseudoBinaryM_VV {
   foreach m = MxList.m in
     defm _VV : VPseudoBinary<VR, m.vrclass, m.vrclass, m, "@earlyclobber $rd">;
 }
 
-multiclass VPseudoBinaryM_VX<bit IsFloat> {
+multiclass VPseudoBinaryM_VX {
   foreach m = MxList.m in
-    defm !if(IsFloat, "_VF", "_VX") :
-      VPseudoBinary<VR, m.vrclass, !if(IsFloat, FPR32, GPR), m,
-                    "@earlyclobber $rd">;
+    defm "_VX" :
+      VPseudoBinary<VR, m.vrclass, GPR, m, "@earlyclobber $rd">;
+}
+
+multiclass VPseudoBinaryM_VF {
+  foreach m = MxList.m in
+    foreach f = FPList.fpinfo in
+      defm "_VF_" # f.FX :
+        VPseudoBinary<VR, m.vrclass, f.fprclass, m, "@earlyclobber $rd">;
 }
 
 multiclass VPseudoBinaryM_VI {
   foreach m = MxList.m in
     defm _VI : VPseudoBinary<VR, m.vrclass, simm5, m, "@earlyclobber $rd">;
 }
 
 multiclass VPseudoBinaryV_VV_VX_VI<Operand ImmType = simm5, string Constraint = ""> {
   defm "" : VPseudoBinaryV_VV<Constraint>;
-  defm "" : VPseudoBinaryV_VX</*IsFloat=*/false, Constraint>;
+  defm "" : VPseudoBinaryV_VX<Constraint>;
   defm "" : VPseudoBinaryV_VI<ImmType, Constraint>;
 }
 
-multiclass VPseudoBinaryV_VV_VX<bit IsFloat = 0> {
+multiclass VPseudoBinaryV_VV_VX {
   defm "" : VPseudoBinaryV_VV;
-  defm "" : VPseudoBinaryV_VX<IsFloat>;
+  defm "" : VPseudoBinaryV_VX;
+}
+
+multiclass VPseudoBinaryV_VV_VF {
+  defm "" : VPseudoBinaryV_VV;
+  defm "" : VPseudoBinaryV_VF;
 }
 
 multiclass VPseudoBinaryV_VX_VI<Operand ImmType = simm5> {
-  defm "" : VPseudoBinaryV_VX</*IsFloat=*/0>;
+  defm "" : VPseudoBinaryV_VX;
   defm "" : VPseudoBinaryV_VI<ImmType>;
 }
 
-multiclass VPseudoBinaryW_VV_VX<bit IsFloat = 0> {
+multiclass VPseudoBinaryW_VV_VX {
+  defm "" : VPseudoBinaryW_VV;
+  defm "" : VPseudoBinaryW_VX;
+}
+
+multiclass VPseudoBinaryW_VV_VF {
   defm "" : VPseudoBinaryW_VV;
-  defm "" : VPseudoBinaryW_VX<IsFloat>;
+  defm "" : VPseudoBinaryW_VF;
 }
 
-multiclass VPseudoBinaryW_WV_WX<bit IsFloat = 0> {
+multiclass VPseudoBinaryW_WV_WX {
   defm "" : VPseudoBinaryW_WV;
-  defm "" : VPseudoBinaryW_WX<IsFloat>;
+  defm "" : VPseudoBinaryW_WX;
+}
+
+multiclass VPseudoBinaryW_WV_WF {
+  defm "" : VPseudoBinaryW_WV;
+  defm "" : VPseudoBinaryW_WF;
 }
 
 multiclass VPseudoBinaryV_VM_XM_IM {
   defm "" : VPseudoBinaryV_VM;
   defm "" : VPseudoBinaryV_XM;
   defm "" : VPseudoBinaryV_IM;
 }
 
foreach m = MxList.m in
     defm _VV : VPseudoTernary<m.vrclass, m.vrclass, m.vrclass, m, Constraint>;
 }
 
 multiclass VPseudoTernaryV_VX<string Constraint = ""> {
   foreach m = MxList.m in
     defm _VX : VPseudoTernary<m.vrclass, m.vrclass, GPR, m, Constraint>;
 }
 
-multiclass VPseudoTernaryV_VX_AAXA<bit IsFloat, string Constraint = ""> {
+multiclass VPseudoTernaryV_VX_AAXA<string Constraint = ""> {
+  foreach m = MxList.m in
+    defm "_VX" : VPseudoTernary<m.vrclass, GPR, m.vrclass, m, Constraint>;
+}
+
+multiclass VPseudoTernaryV_VF_AAXA<string Constraint = ""> {
   foreach m = MxList.m in
-    defm !if(IsFloat, "_VF", "_VX") : VPseudoTernary<m.vrclass,
-                             !if(IsFloat, FPR32, GPR), m.vrclass, m, Constraint>;
+    foreach f = FPList.fpinfo in
+      defm "_VF_" # f.FX : VPseudoTernary<m.vrclass, f.fprclass, m.vrclass,
+                                          m, Constraint>;
 }
 
 multiclass VPseudoTernaryW_VV {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxList.m[0-5] in
     defm _VV : VPseudoTernary<m.wvrclass, m.vrclass, m.vrclass, m, constraint>;
 }
 
-multiclass VPseudoTernaryW_VX<bit IsFloat> {
+multiclass VPseudoTernaryW_VX {
+  defvar constraint = "@earlyclobber $rd";
+  foreach m = MxList.m[0-5] in
+    defm "_VX" : VPseudoTernary<m.wvrclass, GPR, m.vrclass, m, constraint>;
+}
+
+multiclass VPseudoTernaryW_VF {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxList.m[0-5] in
-    defm !if(IsFloat, "_VF", "_VX")  : VPseudoTernary<m.wvrclass,
-                             !if(IsFloat, FPR32, GPR), m.vrclass, m, constraint>;
+    foreach f = FPList.fpinfo[0-1] in
+      defm "_VF_" # f.FX : VPseudoTernary<m.wvrclass, f.fprclass, m.vrclass, m,
+                                          constraint>;
 }
 
 multiclass VPseudoTernaryV_VI<Operand ImmType = simm5, string Constraint = ""> {
   foreach m = MxList.m in
     defm _VI : VPseudoTernary<m.vrclass, m.vrclass, ImmType, m, Constraint>;
 }
 
-multiclass VPseudoTernaryV_VV_VX_AAXA<bit IsFloat = 0, string Constraint = ""> {
+multiclass VPseudoTernaryV_VV_VX_AAXA<string Constraint = ""> {
+  defm "" : VPseudoTernaryV_VV<Constraint>;
+  defm "" : VPseudoTernaryV_VX_AAXA<Constraint>;
+}
+
+multiclass VPseudoTernaryV_VV_VF_AAXA<string Constraint = ""> {
   defm "" : VPseudoTernaryV_VV<Constraint>;
-  defm "" : VPseudoTernaryV_VX_AAXA<IsFloat, Constraint>;
+  defm "" : VPseudoTernaryV_VF_AAXA<Constraint>;
 }
 
 multiclass VPseudoTernaryV_VX_VI<Operand ImmType = simm5, string Constraint = ""> {
   defm "" : VPseudoTernaryV_VX<Constraint>;
   defm "" : VPseudoTernaryV_VI<ImmType, Constraint>;
 }
 
-multiclass VPseudoTernaryW_VV_VX<bit IsFloat = false> {
+multiclass VPseudoTernaryW_VV_VX {
+  defm "" : VPseudoTernaryW_VV;
+  defm "" : VPseudoTernaryW_VX;
+}
+
+multiclass VPseudoTernaryW_VV_VF {
   defm "" : VPseudoTernaryW_VV;
-  defm "" : VPseudoTernaryW_VX<IsFloat>;
+  defm "" : VPseudoTernaryW_VF;
 }
 
 multiclass VPseudoBinaryM_VV_VX_VI {
   defm "" : VPseudoBinaryM_VV;
-  defm "" : VPseudoBinaryM_VX</*IsFloat=*/0>;
+  defm "" : VPseudoBinaryM_VX;
   defm "" : VPseudoBinaryM_VI;
 }
 
-multiclass VPseudoBinaryM_VV_VX<bit IsFloat = 0> {
+multiclass VPseudoBinaryM_VV_VX {
   defm "" : VPseudoBinaryM_VV;
-  defm "" : VPseudoBinaryM_VX<IsFloat>;
+  defm "" : VPseudoBinaryM_VX;
+}
+
+multiclass VPseudoBinaryM_VV_VF {
+  defm "" : VPseudoBinaryM_VV;
+  defm "" : VPseudoBinaryM_VF;
 }
 
 multiclass VPseudoBinaryM_VX_VI {
-  defm "" : VPseudoBinaryM_VX</*IsFloat=*/0>;
+  defm "" : VPseudoBinaryM_VX;
   defm "" : VPseudoBinaryM_VI;
 }
 
 multiclass VPseudoReductionV_VS {
   foreach m = MxList.m in {
     let WritesElement0 = 1 in
     defm _VS : VPseudoTernary<V_M1.vrclass, m.vrclass, V_M1.vrclass, m>;
   }
class VPatBinaryNoMask<string intrinsic_name,
                        VReg op1_reg_class,
                        DAGOperand op2_kind> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic_name)
                    (op1_type op1_reg_class:$rs1),
                    (op2_type op2_kind:$rs2),
                    (XLenVT GPR:$vl))),
                    (!cast<Instruction>(inst)
                    (op1_type op1_reg_class:$rs1),
-                   ToFPR32<op2_type, op2_kind, "rs2">.ret,
+                   (op2_type op2_kind:$rs2),
                    (NoX0 GPR:$vl), sew)>;
 
 class VPatBinaryMask<string intrinsic_name,
                      string inst,
                      ValueType result_type,
                      ValueType op1_type,
                      ValueType op2_type,
                      ValueType mask_type,
                      int sew,
                      VReg result_reg_class,
                      VReg op1_reg_class,
                      DAGOperand op2_kind> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic_name#"_mask")
                    (result_type result_reg_class:$merge),
                    (op1_type op1_reg_class:$rs1),
                    (op2_type op2_kind:$rs2),
                    (mask_type V0),
                    (XLenVT GPR:$vl))),
                    (!cast<Instruction>(inst#"_MASK")
                    (result_type result_reg_class:$merge),
                    (op1_type op1_reg_class:$rs1),
-                   ToFPR32<op2_type, op2_kind, "rs2">.ret,
+                   (op2_type op2_kind:$rs2),
                    (mask_type V0), (NoX0 GPR:$vl), sew)>;
 
 class VPatTernaryNoMask<string intrinsic,
                         string inst,
                         string kind,
                         ValueType result_type,
                         ValueType op1_type,
                         ValueType op2_type,
                         ValueType mask_type,
                         int sew,
                         LMULInfo vlmul,
                         VReg result_reg_class,
                         RegisterClass op1_reg_class,
                         DAGOperand op2_kind> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic)
                     (result_type result_reg_class:$rs3),
                     (op1_type op1_reg_class:$rs1),
                     (op2_type op2_kind:$rs2),
                     (XLenVT GPR:$vl))),
-                   (!cast<Instruction>(inst#_#kind#"_"# vlmul.MX)
+                   (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX)
                     result_reg_class:$rs3,
-                    ToFPR32<op1_type, op1_reg_class, "rs1">.ret,
+                    (op1_type op1_reg_class:$rs1),
                     op2_kind:$rs2,
                     (NoX0 GPR:$vl), sew)>;
 
 class VPatTernaryMask<string intrinsic,
                       string inst,
                       string kind,
                       ValueType result_type,
                       ValueType op1_type,
                       ValueType op2_type,
                       ValueType mask_type,
                       int sew,
                       LMULInfo vlmul,
                       VReg result_reg_class,
                       RegisterClass op1_reg_class,
                       DAGOperand op2_kind> :
   Pat<(result_type (!cast<Intrinsic>(intrinsic#"_mask")
                     (result_type result_reg_class:$rs3),
                     (op1_type op1_reg_class:$rs1),
                     (op2_type op2_kind:$rs2),
                     (mask_type V0),
                     (XLenVT GPR:$vl))),
-                   (!cast<Instruction>(inst#_#kind#"_"# vlmul.MX # "_MASK")
+                   (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX # "_MASK")
                     result_reg_class:$rs3,
-                    ToFPR32<op1_type, op1_reg_class, "rs1">.ret,
+                    (op1_type op1_reg_class:$rs1),
                     op2_kind:$rs2,
                     (mask_type V0),
                     (NoX0 GPR:$vl), sew)>;
 
 class VPatAMOWDNoMask<string intrinsic_name,
                     string inst,
                     ValueType result_type,
                     ValueType op1_type,

 {
   def : Pat<(result_type (!cast<Intrinsic>(intrinsic)
                          (op1_type op1_reg_class:$rs1),
                          (op2_type op2_kind:$rs2),
                          (mask_type V0),
                          (XLenVT GPR:$vl))),
                          (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX)
                          (op1_type op1_reg_class:$rs1),
-                         ToFPR32<op2_type, op2_kind, "rs2">.ret,
+                         (op2_type op2_kind:$rs2),
                          (mask_type V0), (NoX0 GPR:$vl), sew)>;
 }
 
 multiclass VPatBinaryMaskOut<string intrinsic,
                              string inst,
                              string kind,
                              ValueType result_type,
                              ValueType op1_type,
                              ValueType op2_type,
                              int sew,
                              LMULInfo vlmul,
                              VReg op1_reg_class,
                              DAGOperand op2_kind>
 {
   def : Pat<(result_type (!cast<Intrinsic>(intrinsic)
                          (op1_type op1_reg_class:$rs1),
                          (op2_type op2_kind:$rs2),
                          (XLenVT GPR:$vl))),
                          (!cast<Instruction>(inst#"_"#kind#"_"#vlmul.MX)
                          (op1_type op1_reg_class:$rs1),
-                         ToFPR32<op2_type, op2_kind, "rs2">.ret,
+                         (op2_type op2_kind:$rs2),
                          (NoX0 GPR:$vl), sew)>;
 }
 
 multiclass VPatConversion<string intrinsic,
                           string inst,
                           string kind,
                           ValueType result_type,
                           ValueType op1_type,
if !and(!ge(octuple_emul, 1), !le(octuple_emul, 64)) then {
                         vti.RegClass, ivti.RegClass>;
     }
   }
 }
 
 multiclass VPatBinaryV_VX<string intrinsic, string instruction,
                           list<VTypeInfo> vtilist> {
   foreach vti = vtilist in {
-    defvar kind = !if(!eq(vti.Scalar, XLenVT), "_VX_", "_VF_");
-    defm : VPatBinary<intrinsic, instruction # kind # vti.LMul.MX,
+    defvar kind = getScalarSuffix<vti.Scalar>.suffix;
+    defm : VPatBinary<intrinsic, instruction#"_"#kind#"_"#vti.LMul.MX,
                       vti.Vector, vti.Vector, vti.Scalar, vti.Mask,
                       vti.SEW, vti.RegClass,
                       vti.RegClass, vti.ScalarRegClass>;
   }
 }
 
 multiclass VPatBinaryV_VX_INT<string intrinsic, string instruction,
                           list<VTypeInfo> vtilist> {
multiclass VPatBinaryW_VV<string intrinsic, string instruction,
   }
 }
 
 multiclass VPatBinaryW_VX<string intrinsic, string instruction,
                           list<VTypeInfoToWide> vtilist> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
-    defvar kind = !if(!eq(Vti.Scalar, XLenVT), "_VX_", "_VF_");
-    defm : VPatBinary<intrinsic, instruction # kind # Vti.LMul.MX,
+    defvar kind = getScalarSuffix<Vti.Scalar>.suffix;
+    defm : VPatBinary<intrinsic, instruction#"_"#kind#"_"#Vti.LMul.MX,
                       Wti.Vector, Vti.Vector, Vti.Scalar, Vti.Mask,
                       Vti.SEW, Wti.RegClass,
                       Vti.RegClass, Vti.ScalarRegClass>;
   }
 }
 
 multiclass VPatBinaryW_WV<string intrinsic, string instruction,
                           list<VTypeInfoToWide> vtilist> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
     defm : VPatBinary<intrinsic, instruction # "_WV_" # Vti.LMul.MX,
                       Wti.Vector, Wti.Vector, Vti.Vector, Vti.Mask,
                       Vti.SEW, Wti.RegClass,
                       Wti.RegClass, Vti.RegClass>;
   }
 }
 
 multiclass VPatBinaryW_WX<string intrinsic, string instruction,
                           list<VTypeInfoToWide> vtilist> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
-    defvar kind = !if(!eq(Vti.Scalar, XLenVT), "_WX_", "_WF_");
-    defm : VPatBinary<intrinsic, instruction # kind  # Vti.LMul.MX,
+    defvar kind = getWScalarSuffix<Vti.Scalar>.suffix;
+    defm : VPatBinary<intrinsic, instruction#"_"#kind#"_"#Vti.LMul.MX,
                       Wti.Vector, Wti.Vector, Vti.Scalar, Vti.Mask,
                       Vti.SEW, Wti.RegClass,
                       Wti.RegClass, Vti.ScalarRegClass>;
   }
 }
 
 multiclass VPatBinaryV_WV<string intrinsic, string instruction,
                           list<VTypeInfoToWide> vtilist> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
     defm : VPatBinary<intrinsic, instruction # "_WV_" # Vti.LMul.MX,
                       Vti.Vector, Wti.Vector, Vti.Vector, Vti.Mask,
                       Vti.SEW, Vti.RegClass,
                       Wti.RegClass, Vti.RegClass>;
   }
 }
 
 multiclass VPatBinaryV_WX<string intrinsic, string instruction,
                           list<VTypeInfoToWide> vtilist> {
   foreach VtiToWti = vtilist in {
     defvar Vti = VtiToWti.Vti;
     defvar Wti = VtiToWti.Wti;
-    defvar kind = !if(!eq(Vti.Scalar, XLenVT), "_WX_", "_WF_");
-    defm : VPatBinary<intrinsic, instruction # kind # Vti.LMul.MX,
+    defvar kind = getWScalarSuffix<Vti.Scalar>.suffix;
+    defm : VPatBinary<intrinsic, instruction#"_"#kind#"_"#Vti.LMul.MX,
                       Vti.Vector, Wti.Vector, Vti.Scalar, Vti.Mask,
                       Vti.SEW, Vti.RegClass,
                       Wti.RegClass, Vti.ScalarRegClass>;
   }
 }
 
 multiclass VPatBinaryV_WI<string intrinsic, string instruction,
                           list<VTypeInfoToWide> vtilist> {
defm : VPatBinaryCarryIn<intrinsic, instruction, "VVM",
                              vti.RegClass, vti.RegClass>;
 }
 
 multiclass VPatBinaryV_XM<string intrinsic, string instruction,
                           bit CarryOut = 0,
                           list<VTypeInfo> vtilist = AllIntegerVectors> {
   foreach vti = vtilist in
     defm : VPatBinaryCarryIn<intrinsic, instruction,
-                             !if(!eq(vti.Scalar, XLenVT), "VXM", "VFM"),
+                             !cond(!eq(vti.Scalar, XLenVT): "VXM",
+                                   !eq(vti.Scalar, f16): "VFM_F16",
+                                   !eq(vti.Scalar, f32): "VFM_F32",
+                                   !eq(vti.Scalar, f64): "VFM_F64"),
                              !if(CarryOut, vti.Mask, vti.Vector),
                              vti.Vector, vti.Scalar, vti.Mask,
                              vti.SEW, vti.LMul,
                              vti.RegClass, vti.ScalarRegClass>;
 }
 
 multiclass VPatBinaryV_IM<string intrinsic, string instruction,
                           bit CarryOut = 0> {
defm : VPatBinary<intrinsic, instruction # "_VV_" # vti.LMul.MX,
                       vti.Mask, vti.Vector, vti.Vector, vti.Mask,
                       vti.SEW, VR,
                       vti.RegClass, vti.RegClass>;
 }
 
 multiclass VPatBinaryM_VX<string intrinsic, string instruction,
                           list<VTypeInfo> vtilist> {
   foreach vti = vtilist in {
-    defvar kind = !if(!eq(vti.Scalar, XLenVT), "_VX_", "_VF_");
-    defm : VPatBinary<intrinsic, instruction # kind # vti.LMul.MX,
+    defvar kind = getScalarSuffix<vti.Scalar>.suffix;
+    defm : VPatBinary<intrinsic, instruction#"_"#kind#"_"#vti.LMul.MX,
                       vti.Mask, vti.Vector, vti.Scalar, vti.Mask,
                       vti.SEW, VR,
                       vti.RegClass, vti.ScalarRegClass>;
   }
 }
 
 multiclass VPatBinaryM_VI<string intrinsic, string instruction,
                           list<VTypeInfo> vtilist> {
defm : VPatTernary<intrinsic, instruction, "VX",
                        vti.SEW, vti.LMul, vti.RegClass,
                        vti.RegClass, GPR>;
 }
 
 multiclass VPatTernaryV_VX_AAXA<string intrinsic, string instruction,
                            list<VTypeInfo> vtilist> {
   foreach vti = vtilist in
     defm : VPatTernary<intrinsic, instruction,
-                       !if(!eq(vti.Scalar, XLenVT), "VX", "VF"),
+                       getScalarSuffix<vti.Scalar>.suffix,
                        vti.Vector, vti.Scalar, vti.Vector, vti.Mask,
                        vti.SEW, vti.LMul, vti.RegClass,
                        vti.ScalarRegClass, vti.RegClass>;
 }
 
 multiclass VPatTernaryV_VI<string intrinsic, string instruction,
                            list<VTypeInfo> vtilist, Operand Imm_type> {
   foreach vti = vtilist in

 }
 
 multiclass VPatTernaryW_VX<string intrinsic, string instruction,
                            list<VTypeInfoToWide> vtilist> {
   foreach vtiToWti = vtilist in {
     defvar vti = vtiToWti.Vti;
     defvar wti = vtiToWti.Wti;
     defm : VPatTernary<intrinsic, instruction,
-                       !if(!eq(vti.Scalar, XLenVT), "VX", "VF"),
+                       getScalarSuffix<vti.Scalar>.suffix,
                        wti.Vector, vti.Scalar, vti.Vector,
                        vti.Mask, vti.SEW, vti.LMul,
                        wti.RegClass, vti.ScalarRegClass, vti.RegClass>;
   }
 }
 
 multiclass VPatTernaryV_VV_VX_AAXA<string intrinsic, string instruction,
                               list<VTypeInfo> vtilist> {

 defm PseudoVNMSUB      : VPseudoTernaryV_VV_VX_AAXA;
 
 //===----------------------------------------------------------------------===//
 // 12.14. Vector Widening Integer Multiply-Add Instructions
 //===----------------------------------------------------------------------===//
 defm PseudoVWMACCU     : VPseudoTernaryW_VV_VX;
 defm PseudoVWMACC      : VPseudoTernaryW_VV_VX;
 defm PseudoVWMACCSU    : VPseudoTernaryW_VV_VX;
-defm PseudoVWMACCUS    : VPseudoTernaryW_VX</*IsFloat*/false>;
+defm PseudoVWMACCUS    : VPseudoTernaryW_VX;
 
 //===----------------------------------------------------------------------===//
 // 12.16. Vector Integer Merge Instructions
 //===----------------------------------------------------------------------===//
 defm PseudoVMERGE      : VPseudoBinaryV_VM_XM_IM;
 
 //===----------------------------------------------------------------------===//
 // 12.17. Vector Integer Move Instructions

 }
 
 } // Predicates = [HasStdExtV]
 
 let Predicates = [HasStdExtV, HasStdExtF] in {
 //===----------------------------------------------------------------------===//
 // 14.2. Vector Single-Width Floating-Point Add/Subtract Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVFADD       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
-defm PseudoVFSUB       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
-defm PseudoVFRSUB      : VPseudoBinaryV_VX</*IsFloat=*/1>;
+defm PseudoVFADD       : VPseudoBinaryV_VV_VF;
+defm PseudoVFSUB       : VPseudoBinaryV_VV_VF;
+defm PseudoVFRSUB      : VPseudoBinaryV_VF;
 
 //===----------------------------------------------------------------------===//
 // 14.3. Vector Widening Floating-Point Add/Subtract Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVFWADD     : VPseudoBinaryW_VV_VX</*IsFloat=*/1>;
-defm PseudoVFWSUB     : VPseudoBinaryW_VV_VX</*IsFloat=*/1>;
-defm PseudoVFWADD     : VPseudoBinaryW_WV_WX</*IsFloat=*/1>;
-defm PseudoVFWSUB     : VPseudoBinaryW_WV_WX</*IsFloat=*/1>;
+defm PseudoVFWADD     : VPseudoBinaryW_VV_VF;
+defm PseudoVFWSUB     : VPseudoBinaryW_VV_VF;
+defm PseudoVFWADD     : VPseudoBinaryW_WV_WF;
+defm PseudoVFWSUB     : VPseudoBinaryW_WV_WF;
 
 //===----------------------------------------------------------------------===//
 // 14.4. Vector Single-Width Floating-Point Multiply/Divide Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVFMUL       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
-defm PseudoVFDIV       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
-defm PseudoVFRDIV      : VPseudoBinaryV_VX</*IsFloat=*/1>;
+defm PseudoVFMUL       : VPseudoBinaryV_VV_VF;
+defm PseudoVFDIV       : VPseudoBinaryV_VV_VF;
+defm PseudoVFRDIV      : VPseudoBinaryV_VF;
 
 //===----------------------------------------------------------------------===//
 // 14.5. Vector Widening Floating-Point Multiply
 //===----------------------------------------------------------------------===//
-defm PseudoVFWMUL      : VPseudoBinaryW_VV_VX</*IsFloat=*/1>;
+defm PseudoVFWMUL      : VPseudoBinaryW_VV_VF;
 
 //===----------------------------------------------------------------------===//
 // 14.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVFMACC      : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
-defm PseudoVFNMACC     : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
-defm PseudoVFMSAC      : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
-defm PseudoVFNMSAC     : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
-defm PseudoVFMADD      : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
-defm PseudoVFNMADD     : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
-defm PseudoVFMSUB      : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
-defm PseudoVFNMSUB     : VPseudoTernaryV_VV_VX_AAXA</*IsFloat*/true>;
+defm PseudoVFMACC      : VPseudoTernaryV_VV_VF_AAXA;
+defm PseudoVFNMACC     : VPseudoTernaryV_VV_VF_AAXA;
+defm PseudoVFMSAC      : VPseudoTernaryV_VV_VF_AAXA;
+defm PseudoVFNMSAC     : VPseudoTernaryV_VV_VF_AAXA;
+defm PseudoVFMADD      : VPseudoTernaryV_VV_VF_AAXA;
+defm PseudoVFNMADD     : VPseudoTernaryV_VV_VF_AAXA;
+defm PseudoVFMSUB      : VPseudoTernaryV_VV_VF_AAXA;
+defm PseudoVFNMSUB     : VPseudoTernaryV_VV_VF_AAXA;
 
 //===----------------------------------------------------------------------===//
 // 14.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVFWMACC     : VPseudoTernaryW_VV_VX</*IsFloat*/true>;
-defm PseudoVFWNMACC    : VPseudoTernaryW_VV_VX</*IsFloat*/true>;
-defm PseudoVFWMSAC     : VPseudoTernaryW_VV_VX</*IsFloat*/true>;
-defm PseudoVFWNMSAC    : VPseudoTernaryW_VV_VX</*IsFloat*/true>;
+defm PseudoVFWMACC     : VPseudoTernaryW_VV_VF;
+defm PseudoVFWNMACC    : VPseudoTernaryW_VV_VF;
+defm PseudoVFWMSAC     : VPseudoTernaryW_VV_VF;
+defm PseudoVFWNMSAC    : VPseudoTernaryW_VV_VF;
 
 //===----------------------------------------------------------------------===//
 // 14.8. Vector Floating-Point Square-Root Instruction
 //===----------------------------------------------------------------------===//
 defm PseudoVFSQRT      : VPseudoUnaryV_V;
 
 //===----------------------------------------------------------------------===//
 // 14.9. Vector Floating-Point Reciprocal Square-Root Estimate Instruction
 //===----------------------------------------------------------------------===//
 defm PseudoVFRSQRTE7   : VPseudoUnaryV_V;
 
 //===----------------------------------------------------------------------===//
 // 14.10. Vector Floating-Point Reciprocal Estimate Instruction
 //===----------------------------------------------------------------------===//
 defm PseudoVFRECE7     : VPseudoUnaryV_V;
 
 //===----------------------------------------------------------------------===//
 // 14.11. Vector Floating-Point Min/Max Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVFMIN       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
-defm PseudoVFMAX       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
+defm PseudoVFMIN       : VPseudoBinaryV_VV_VF;
+defm PseudoVFMAX       : VPseudoBinaryV_VV_VF;
 
 //===----------------------------------------------------------------------===//
 // 14.12. Vector Floating-Point Sign-Injection Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVFSGNJ      : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
-defm PseudoVFSGNJN     : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
-defm PseudoVFSGNJX     : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
+defm PseudoVFSGNJ      : VPseudoBinaryV_VV_VF;
+defm PseudoVFSGNJN     : VPseudoBinaryV_VV_VF;
+defm PseudoVFSGNJX     : VPseudoBinaryV_VV_VF;
 
 //===----------------------------------------------------------------------===//
 // 14.13. Vector Floating-Point Compare Instructions
 //===----------------------------------------------------------------------===//
-defm PseudoVMFEQ       : VPseudoBinaryM_VV_VX</*IsFloat=*/1>;
-defm PseudoVMFNE       : VPseudoBinaryM_VV_VX</*IsFloat=*/1>;
-defm PseudoVMFLT       : VPseudoBinaryM_VV_VX</*IsFloat=*/1>;
-defm PseudoVMFLE       : VPseudoBinaryM_VV_VX</*IsFloat=*/1>;
-defm PseudoVMFGT       : VPseudoBinaryM_VX</*IsFloat=*/1>;
-defm PseudoVMFGE       : VPseudoBinaryM_VX</*IsFloat=*/1>;
+defm PseudoVMFEQ       : VPseudoBinaryM_VV_VF;
+defm PseudoVMFNE       : VPseudoBinaryM_VV_VF;
+defm PseudoVMFLT       : VPseudoBinaryM_VV_VF;
+defm PseudoVMFLE       : VPseudoBinaryM_VV_VF;
+defm PseudoVMFGT       : VPseudoBinaryM_VF;
+defm PseudoVMFGE       : VPseudoBinaryM_VF;
 
 //===----------------------------------------------------------------------===//
 // 14.14. Vector Floating-Point Classify Instruction
 //===----------------------------------------------------------------------===//
 defm PseudoVFCLASS     : VPseudoUnaryV_V;
 
 //===----------------------------------------------------------------------===//
 // 14.15. Vector Floating-Point Merge Instruction
 //===----------------------------------------------------------------------===//
-defm PseudoVFMERGE     : VPseudoBinaryV_XM</*CarryOut =*/0,/*CarryIn =*/true,
-                                           /*Constraint =*/"", /*IsFloat=*/true>;
+defm PseudoVFMERGE     : VPseudoBinaryV_FM;
 
 //===----------------------------------------------------------------------===//
 // 14.16. Vector Floating-Point Move Instruction
 //===----------------------------------------------------------------------===//
 defm PseudoVFMV_V      : VPseudoUnaryV_F_NoDummyMask;
 
 //===----------------------------------------------------------------------===//
 // 14.17. Single-Width Floating-Point/Integer Type-Convert Instructions

 //===----------------------------------------------------------------------===//
 // 17.2. Floating-Point Scalar Move Instructions
 //===----------------------------------------------------------------------===//
 
 let Predicates = [HasStdExtV, HasStdExtF] in {
 let mayLoad = 0, mayStore = 0, hasSideEffects = 0, usesCustomInserter = 1,
     Uses = [VL, VTYPE] in {
   foreach m = MxList.m in {
+    foreach f = FPList.fpinfo in {
       let VLMul = m.value in {
         let HasSEWOp = 1, BaseInstr = VFMV_F_S in
-      def PseudoVFMV_F_S # "_" # m.MX : Pseudo<(outs FPR32:$rd),
+        def PseudoVFMV_F_S # "_" # f.FX # "_" # m.MX :
+                                          Pseudo<(outs f.fprclass:$rd),
                                                  (ins m.vrclass:$rs2,
                                                       ixlenimm:$sew),
                                                  []>, RISCVVPseudo;
         let HasVLOp = 1, HasSEWOp = 1, BaseInstr = VFMV_S_F, WritesElement0 = 1,
             Constraints = "$rd = $rs1" in
-      def PseudoVFMV_S_F # "_" # m.MX : Pseudo<(outs m.vrclass:$rd),
-                                               (ins m.vrclass:$rs1, FPR32:$rs2,
+        def PseudoVFMV_S_F # "_" # f.FX #"_" # m.MX :
+                                          Pseudo<(outs m.vrclass:$rd),
+                                                 (ins m.vrclass:$rs1, f.fprclass:$rs2,
                                                       GPR:$vl, ixlenimm:$sew),
                                                  []>, RISCVVPseudo;
       }
     }
   }
+}
 } // Predicates = [HasStdExtV, HasStdExtF]
 
 //===----------------------------------------------------------------------===//
 // 17.3. Vector Slide Instructions
 //===----------------------------------------------------------------------===//
 let Predicates = [HasStdExtV] in {
   defm PseudoVSLIDEUP    : VPseudoTernaryV_VX_VI<uimm5, "@earlyclobber $rd">;
   defm PseudoVSLIDEDOWN  : VPseudoTernaryV_VX_VI<uimm5>;
-  defm PseudoVSLIDE1UP   : VPseudoBinaryV_VX</*IsFloat*/false, "@earlyclobber $rd">;
-  defm PseudoVSLIDE1DOWN : VPseudoBinaryV_VX</*IsFloat*/false>;
+  defm PseudoVSLIDE1UP   : VPseudoBinaryV_VX<"@earlyclobber $rd">;
+  defm PseudoVSLIDE1DOWN : VPseudoBinaryV_VX;
 } // Predicates = [HasStdExtV]
 
 let Predicates = [HasStdExtV, HasStdExtF] in {
-  defm PseudoVFSLIDE1UP  : VPseudoBinaryV_VX</*IsFloat*/true, "@earlyclobber $rd">;
-  defm PseudoVFSLIDE1DOWN : VPseudoBinaryV_VX</*IsFloat*/true>;
+  defm PseudoVFSLIDE1UP  : VPseudoBinaryV_VF<"@earlyclobber $rd">;
+  defm PseudoVFSLIDE1DOWN : VPseudoBinaryV_VF;
 } // Predicates = [HasStdExtV, HasStdExtF]
 
 //===----------------------------------------------------------------------===//
 // 17.4. Vector Register Gather Instructions
 //===----------------------------------------------------------------------===//
 defm PseudoVRGATHER    : VPseudoBinaryV_VV_VX_VI<uimm5, "@earlyclobber $rd">;
 defm PseudoVRGATHEREI16 : VPseudoBinaryV_VV_EEW</* eew */ 16, "@earlyclobber $rd">;
 
foreach fvti = AllFloatVectors in {
   // If we're splatting fpimm0, use vmv.v.x vd, x0.
   def : Pat<(fvti.Vector (int_riscv_vfmv_v_f
                          (fvti.Scalar (fpimm0)), GPR:$vl)),
             (!cast<Instruction>("PseudoVMV_V_I_"#fvti.LMul.MX)
              0, (NoX0 GPR:$vl), fvti.SEW)>;
 
   def : Pat<(fvti.Vector (int_riscv_vfmv_v_f
                          (fvti.Scalar fvti.ScalarRegClass:$rs2), GPR:$vl)),
-            (!cast<Instruction>("PseudoVFMV_V_F_"#fvti.LMul.MX)
-             ToFPR32<fvti.Scalar, fvti.ScalarRegClass, "rs2">.ret,
+            (!cast<Instruction>("PseudoVFMV_V_F_" #
+                                !cond(!eq(fvti.Scalar, f16): "F16_",
+                                      !eq(fvti.Scalar, f32): "F32_",
+                                      !eq(fvti.Scalar, f64): "F64_") #
+                                fvti.LMul.MX)
+             (fvti.Scalar fvti.ScalarRegClass:$rs2),
              (NoX0 GPR:$vl), fvti.SEW)>;
 }
 
 //===----------------------------------------------------------------------===//
 // 14.17. Single-Width Floating-Point/Integer Type-Convert Instructions
 //===----------------------------------------------------------------------===//
 defm "" : VPatConversionVI_VF<"int_riscv_vfcvt_xu_f_v", "PseudoVFCVT_XU_F">;
 defm "" : VPatConversionVI_VF<"int_riscv_vfcvt_rtz_xu_f_v", "PseudoVFCVT_RTZ_XU_F">;

 } // Predicates = [HasStdExtV]
 
 //===----------------------------------------------------------------------===//
 // 17.2. Floating-Point Scalar Move Instructions
 //===----------------------------------------------------------------------===//
 
 let Predicates = [HasStdExtV, HasStdExtF] in {
 foreach fvti = AllFloatVectors in {
-  defvar instr = !cast<Instruction>("PseudoVFMV_F_S_" # fvti.LMul.MX);
+  defvar instr = !cast<Instruction>("PseudoVFMV_F_S_" #
+                                    !cond(!eq(fvti.Scalar, f16): "F16_",
+                                          !eq(fvti.Scalar, f32): "F32_",
+                                          !eq(fvti.Scalar, f64): "F64_") #
+                                    fvti.LMul.MX);
   def : Pat<(fvti.Scalar (int_riscv_vfmv_f_s (fvti.Vector fvti.RegClass:$rs2))),
-             // Floating point instructions with a scalar result will always
-             // generate the result in a register of class FPR32. When dealing
-             // with the f64 variant of a pattern we need to promote the FPR32
-             // subregister generated by the instruction to the FPR64 base
-             // register expected by the type in the pattern
-             !cond(!eq(!cast<string>(fvti.ScalarRegClass),
-                       !cast<string>(FPR64)):
-                      (SUBREG_TO_REG (i32 -1),
-                                     (instr $rs2, fvti.SEW), sub_32),
-                   !eq(!cast<string>(fvti.ScalarRegClass),
-                       !cast<string>(FPR16)):
-                      (EXTRACT_SUBREG (instr $rs2, fvti.SEW), sub_16),
-                   !eq(1, 1):
-                      (instr $rs2, fvti.SEW))>;
+                         (instr $rs2, fvti.SEW)>;
 
   def : Pat<(fvti.Vector (int_riscv_vfmv_s_f (fvti.Vector fvti.RegClass:$rs1),
                          (fvti.Scalar fvti.ScalarRegClass:$rs2), GPR:$vl)),
-            (!cast<Instruction>("PseudoVFMV_S_F_" # fvti.LMul.MX)
-             (fvti.Vector $rs1), ToFPR32<fvti.Scalar, fvti.ScalarRegClass, "rs2">.ret,
+            (!cast<Instruction>("PseudoVFMV_S_F_" #
+                                !cond(!eq(fvti.Scalar, f16): "F16_",
+                                      !eq(fvti.Scalar, f32): "F32_",
+                                      !eq(fvti.Scalar, f64): "F64_") #
+                                fvti.LMul.MX)
+             (fvti.Vector $rs1),
+             (fvti.Scalar fvti.ScalarRegClass:$rs2),
              (NoX0 GPR:$vl), fvti.SEW)>;
 }
 } // Predicates = [HasStdExtV, HasStdExtF]
 
 //===----------------------------------------------------------------------===//
 // 17.3. Vector Slide Instructions
 //===----------------------------------------------------------------------===//
 let Predicates = [HasStdExtV] in {

llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td

 
 // Penalize the generic form with Complexity=1 to give the simm5/uimm5 variants
 // precedence
 def SplatPat       : ComplexPattern<vAny, 1, "selectVSplat", [], [], 1>;
 
 def SplatPat_simm5 : ComplexPattern<vAny, 1, "selectVSplatSimm5", []>;
 def SplatPat_uimm5 : ComplexPattern<vAny, 1, "selectVSplatUimm5", []>;
 
+// We only model FPR32 for V instructions in RISCVInstrInfoV.td.
+// FP16/FP32/FP64 registers are alias each other. Convert FPR16 and FPR64
+// to FPR32 for V instructions is enough.
+class ToFPR32<ValueType type, DAGOperand operand, string name> {
+  dag ret = !cond(!eq(!cast<string>(operand), !cast<string>(FPR64)):
+                      (EXTRACT_SUBREG !dag(type, [FPR64], [name]), sub_32),
+                  !eq(!cast<string>(operand), !cast<string>(FPR16)):
+                      (SUBREG_TO_REG (i16 -1), !dag(type, [FPR16], [name]), sub_16),
+                  !eq(1, 1):
+                      !dag(type, [operand], [name]));
+}
+
 class SwapHelper<dag Prefix, dag A, dag B, dag Suffix, bit swap> {
    dag Value = !con(Prefix, !if(swap, B, A), !if(swap, A, B), Suffix);
 }
 
 multiclass VPatUSLoadStoreSDNode<LLVMType type,
                                  int sew,
                                  LMULInfo vlmul,
                                  OutPatFrag avl,
class VPatBinarySDNode_VF<SDNode vop,
                           int sew,
                           LMULInfo vlmul,
                           OutPatFrag avl,
                           VReg RetClass,
                           VReg vop_reg_class,
                           DAGOperand xop_kind> :
     Pat<(result_type (vop (vop_type vop_reg_class:$rs1),
                           (vop_type (splat_vector xop_kind:$rs2)))),
-        (!cast<Instruction>(instruction_name#"_VF_"#vlmul.MX)
+        (!cast<Instruction>(instruction_name#"_VF_F32_"#vlmul.MX)
                      vop_reg_class:$rs1,
                      ToFPR32<xop_type, xop_kind, "rs2">.ret,
                      avl, sew)>;
 
 multiclass VPatBinaryFPSDNode_VV_VF<SDNode vop, string instruction_name> {
   foreach vti = AllFloatVectors in {
     def : VPatBinarySDNode_VV<vop, instruction_name,
                               vti.Vector, vti.Vector, vti.Mask, vti.SEW,
                               vti.LMul, vti.AVL, vti.RegClass, vti.RegClass>;
     def : VPatBinarySDNode_VF<vop, instruction_name,
                               vti.Vector, vti.Vector, vti.Scalar, vti.Mask,
                               vti.SEW, vti.LMul, vti.AVL, vti.RegClass, vti.RegClass,
                               vti.ScalarRegClass>;
   }
 }
 
 multiclass VPatBinaryFPSDNode_R_VF<SDNode vop, string instruction_name> {
   foreach fvti = AllFloatVectors in
     def : Pat<(fvti.Vector (vop (fvti.Vector (splat_vector fvti.Scalar:$rs2)),
                                 (fvti.Vector fvti.RegClass:$rs1))),
-              (!cast<Instruction>(instruction_name#"_VF_"#fvti.LMul.MX)
+              (!cast<Instruction>(instruction_name#"_VF_F32_"#fvti.LMul.MX)
                            fvti.RegClass:$rs1,
                            ToFPR32<fvti.Scalar, fvti.ScalarRegClass, "rs2">.ret,
                            fvti.AVL, fvti.SEW)>;
 }
 
 multiclass VPatIntegerSetCCSDNode_VV<CondCode cc,
                                      string instruction_name,
                                      bit swap = 0> {
def : Pat<(fvti.Mask (setcc (fvti.Vector fvti.RegClass:$rs1),
                   fvti.RegClass:$rs1, fvti.RegClass:$rs2, fvti.AVL, fvti.SEW)>;
 }
 
 multiclass VPatFPSetCCSDNode_VF<CondCode cc, string instruction_name> {
   foreach fvti = AllFloatVectors in
     def : Pat<(fvti.Mask (setcc (fvti.Vector fvti.RegClass:$rs1),
                                 (fvti.Vector (splat_vector fvti.ScalarRegClass:$rs2)),
                                 cc)),
-              (!cast<Instruction>(instruction_name#"_VF_"#fvti.LMul.MX)
+              (!cast<Instruction>(instruction_name#"_VF_F32_"#fvti.LMul.MX)
                   fvti.RegClass:$rs1,
                   ToFPR32<fvti.Scalar, fvti.ScalarRegClass, "rs2">.ret,
                   fvti.AVL, fvti.SEW)>;
 }
 
 multiclass VPatFPSetCCSDNode_FV<CondCode cc, string swapped_op_instruction_name> {
   foreach fvti = AllFloatVectors in
     def : Pat<(fvti.Mask (setcc (fvti.Vector (splat_vector fvti.ScalarRegClass:$rs2)),
                                 (fvti.Vector fvti.RegClass:$rs1),
                                 cc)),
-              (!cast<Instruction>(swapped_op_instruction_name#"_VF_"#fvti.LMul.MX)
+              (!cast<Instruction>(swapped_op_instruction_name#"_VF_F32_"#fvti.LMul.MX)
                   fvti.RegClass:$rs1,
                   ToFPR32<fvti.Scalar, fvti.ScalarRegClass, "rs2">.ret,
                   fvti.AVL, fvti.SEW)>;
 }
 
 multiclass VPatFPSetCCSDNode_VV_VF_FV<CondCode cc,
                                       string inst_name,
                                       string swapped_op_inst_name> {
def : Pat<(fvti.Vector (vselect (fvti.Mask VMV0:$vm), fvti.RegClass:$rs1,
                                                         fvti.RegClass:$rs2)),
             (!cast<Instruction>("PseudoVMERGE_VVM_"#fvti.LMul.MX)
                  fvti.RegClass:$rs2, fvti.RegClass:$rs1, VMV0:$vm,
                  fvti.AVL, fvti.SEW)>;
 
   def : Pat<(fvti.Vector (vselect (fvti.Mask VMV0:$vm),
                                   (splat_vector fvti.ScalarRegClass:$rs1),
                                   fvti.RegClass:$rs2)),
-            (!cast<Instruction>("PseudoVFMERGE_VFM_"#fvti.LMul.MX)
+            (!cast<Instruction>("PseudoVFMERGE_VFM_F32_"#fvti.LMul.MX)
                  fvti.RegClass:$rs2,
                  ToFPR32<fvti.Scalar, fvti.ScalarRegClass, "rs1">.ret,
                  VMV0:$vm, fvti.AVL, fvti.SEW)>;
 
   def : Pat<(fvti.Vector (vselect (fvti.Mask VMV0:$vm),
                                   (splat_vector (fvti.Scalar fpimm0)),
                                   fvti.RegClass:$rs2)),
             (!cast<Instruction>("PseudoVMERGE_VIM_"#fvti.LMul.MX)
def : Pat<(vti.Vector (rv32_splat_i64 simm5:$rs1)),
                 simm5:$rs1, vti.AVL, vti.SEW)>;
   }
 }
 } // Predicates = [HasStdExtV, IsRV32]
 
 let Predicates = [HasStdExtV, HasStdExtF] in {
 foreach fvti = AllFloatVectors in {
   def : Pat<(fvti.Vector (splat_vector fvti.ScalarRegClass:$rs1)),
-            (!cast<Instruction>("PseudoVFMV_V_F_"#fvti.LMul.MX)
+            (!cast<Instruction>("PseudoVFMV_V_F_F32_"#fvti.LMul.MX)
               ToFPR32<fvti.Scalar, fvti.ScalarRegClass, "rs1">.ret,
               fvti.AVL, fvti.SEW)>;
 
   def : Pat<(fvti.Vector (splat_vector (fvti.Scalar fpimm0))),
             (!cast<Instruction>("PseudoVMV_V_I_"#fvti.LMul.MX)
               0, fvti.AVL, fvti.SEW)>;
 }
 } // Predicates = [HasStdExtV, HasStdExtF]

llvm/lib/Target/RISCV/RISCVMCInstLower.cpp

for (const MachineOperand &MO : MI->explicit_operands()) {
     case MachineOperand::MO_Register: {
       unsigned Reg = MO.getReg();
 
       if (RISCV::VRM2RegClass.contains(Reg) ||
           RISCV::VRM4RegClass.contains(Reg) ||
           RISCV::VRM8RegClass.contains(Reg)) {
         Reg = TRI->getSubReg(Reg, RISCV::sub_vrm1_0);
         assert(Reg && "Subregister does not exist");
+      } else if (RISCV::FPR16RegClass.contains(Reg)) {
+        Reg = TRI->getMatchingSuperReg(Reg, RISCV::sub_16, &RISCV::FPR32RegClass);
+        assert(Reg && "Subregister does not exist");
+      } else if (RISCV::FPR64RegClass.contains(Reg)) {
+        Reg = TRI->getSubReg(Reg, RISCV::sub_32);
+        assert(Reg && "Superregister does not exist");
       }
 
       MCOp = MCOperand::createReg(Reg);
       break;
     }
     case MachineOperand::MO_Immediate:
       MCOp = MCOperand::createImm(MO.getImm());
       break;

llvm/test/CodeGen/RISCV/rvv/spill-fpr-scalar.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+d,+experimental-zfh \
+; RUN:   -verify-machineinstrs --riscv-no-aliases < %s \
+; RUN:   | FileCheck %s
+
+declare half @llvm.riscv.vfmv.f.s.nxv1f16(<vscale x 1 x half>)
+declare float @llvm.riscv.vfmv.f.s.nxv1f32(<vscale x 1 x float>)
+declare double @llvm.riscv.vfmv.f.s.nxv1f64(<vscale x 1 x double>)
+
+declare <vscale x 1 x half> @llvm.riscv.vfmv.v.f.nxv1f16(half, i64);
+declare <vscale x 1 x float> @llvm.riscv.vfmv.v.f.nxv1f32(float, i64);
+declare <vscale x 1 x double> @llvm.riscv.vfmv.v.f.nxv1f64(double, i64);
+
+define <vscale x 1 x half> @intrinsic_vfmv.f.s_s_nxv1f16(<vscale x 1 x half> %0, i64 %1) nounwind {
+; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    vsetvli zero, zero, e16,mf4,ta,mu
+; CHECK-NEXT:    vfmv.f.s ft0, v8
+; CHECK-NEXT:    fsh ft0, 14(sp) # 2-byte Folded Spill
+; CHECK-NEXT:    #APP
+; CHECK-NEXT:    #NO_APP
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    flh ft0, 14(sp) # 2-byte Folded Reload
+; CHECK-NEXT:    vfmv.v.f v8, ft0
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call half @llvm.riscv.vfmv.f.s.nxv1f16(<vscale x 1 x half> %0)
+  tail call void asm sideeffect "", "~{f0_d},~{f1_d},~{f2_d},~{f3_d},~{f4_d},~{f5_d},~{f6_d},~{f7_d},~{f8_d},~{f9_d},~{f10_d},~{f11_d},~{f12_d},~{f13_d},~{f14_d},~{f15_d},~{f16_d},~{f17_d},~{f18_d},~{f19_d},~{f20_d},~{f21_d},~{f22_d},~{f23_d},~{f24_d},~{f25_d},~{f26_d},~{f27_d},~{f28_d},~{f29_d},~{f30_d},~{f31_d}"()
+  %b = call <vscale x 1 x half> @llvm.riscv.vfmv.v.f.nxv1f16(half %a, i64 %1)
+  ret <vscale x 1 x half> %b
+}
+
+define <vscale x 1 x float> @intrinsic_vfmv.f.s_s_nxv1f32(<vscale x 1 x float> %0, i64 %1) nounwind {
+; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    vsetvli zero, zero, e32,mf2,ta,mu
+; CHECK-NEXT:    vfmv.f.s ft0, v8
+; CHECK-NEXT:    fsw ft0, 12(sp) # 4-byte Folded Spill
+; CHECK-NEXT:    #APP
+; CHECK-NEXT:    #NO_APP
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    flw ft0, 12(sp) # 4-byte Folded Reload
+; CHECK-NEXT:    vfmv.v.f v8, ft0
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call float @llvm.riscv.vfmv.f.s.nxv1f32(<vscale x 1 x float> %0)
+  tail call void asm sideeffect "", "~{f0_d},~{f1_d},~{f2_d},~{f3_d},~{f4_d},~{f5_d},~{f6_d},~{f7_d},~{f8_d},~{f9_d},~{f10_d},~{f11_d},~{f12_d},~{f13_d},~{f14_d},~{f15_d},~{f16_d},~{f17_d},~{f18_d},~{f19_d},~{f20_d},~{f21_d},~{f22_d},~{f23_d},~{f24_d},~{f25_d},~{f26_d},~{f27_d},~{f28_d},~{f29_d},~{f30_d},~{f31_d}"()
+  %b = call <vscale x 1 x float> @llvm.riscv.vfmv.v.f.nxv1f32(float %a, i64 %1)
+  ret <vscale x 1 x float> %b
+}
+
+define <vscale x 1 x double> @intrinsic_vfmv.f.s_s_nxv1f64(<vscale x 1 x double> %0, i64 %1) nounwind {
+; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv1f64:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    addi sp, sp, -16
+; CHECK-NEXT:    vsetvli zero, zero, e64,m1,ta,mu
+; CHECK-NEXT:    vfmv.f.s ft0, v8
+; CHECK-NEXT:    fsd ft0, 8(sp) # 8-byte Folded Spill
+; CHECK-NEXT:    #APP
+; CHECK-NEXT:    #NO_APP
+; CHECK-NEXT:    vsetvli a0, a0, e64,m1,ta,mu
+; CHECK-NEXT:    fld ft0, 8(sp) # 8-byte Folded Reload
+; CHECK-NEXT:    vfmv.v.f v8, ft0
+; CHECK-NEXT:    addi sp, sp, 16
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call double @llvm.riscv.vfmv.f.s.nxv1f64(<vscale x 1 x double> %0)
+  tail call void asm sideeffect "", "~{f0_d},~{f1_d},~{f2_d},~{f3_d},~{f4_d},~{f5_d},~{f6_d},~{f7_d},~{f8_d},~{f9_d},~{f10_d},~{f11_d},~{f12_d},~{f13_d},~{f14_d},~{f15_d},~{f16_d},~{f17_d},~{f18_d},~{f19_d},~{f20_d},~{f21_d},~{f22_d},~{f23_d},~{f24_d},~{f25_d},~{f26_d},~{f27_d},~{f28_d},~{f29_d},~{f30_d},~{f31_d}"()
+  %b = call <vscale x 1 x double> @llvm.riscv.vfmv.v.f.nxv1f64(double %a, i64 %1)
+  ret <vscale x 1 x double> %b
+}

llvm/test/CodeGen/RISCV/rvv/vfmv.f.s.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv64 -mattr=+d,+experimental-v,+experimental-zfh -target-abi lp64d -verify-machineinstrs < %s | FileCheck %s
 ; RUN: llc -mtriple=riscv32 -mattr=+d,+experimental-v,+experimental-zfh -target-abi ilp32d -verify-machineinstrs < %s | FileCheck %s
 
 declare half @llvm.riscv.vfmv.f.s.nxv1f16(<vscale x 1 x half>)
 
 define half @intrinsic_vfmv.f.s_s_nxv1f16(<vscale x 1 x half> %0) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv1f16:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vsetvli zero, zero, e16,mf4,ta,mu
 ; CHECK-NEXT:    vfmv.f.s fa0, v8
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h killed $f10_f
 ; CHECK-NEXT:    ret
 entry:
   %a = call half @llvm.riscv.vfmv.f.s.nxv1f16(<vscale x 1 x half> %0)
   ret half %a
 }
 
 declare half @llvm.riscv.vfmv.f.s.nxv2f16(<vscale x 2 x half>)
 
 define half @intrinsic_vfmv.f.s_s_nxv2f16(<vscale x 2 x half> %0) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv2f16:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vsetvli zero, zero, e16,mf2,ta,mu
 ; CHECK-NEXT:    vfmv.f.s fa0, v8
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h killed $f10_f
 ; CHECK-NEXT:    ret
 entry:
   %a = call half @llvm.riscv.vfmv.f.s.nxv2f16(<vscale x 2 x half> %0)
   ret half %a
 }
 
 declare half @llvm.riscv.vfmv.f.s.nxv4f16(<vscale x 4 x half>)
 
 define half @intrinsic_vfmv.f.s_s_nxv4f16(<vscale x 4 x half> %0) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv4f16:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vsetvli zero, zero, e16,m1,ta,mu
 ; CHECK-NEXT:    vfmv.f.s fa0, v8
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h killed $f10_f
 ; CHECK-NEXT:    ret
 entry:
   %a = call half @llvm.riscv.vfmv.f.s.nxv4f16(<vscale x 4 x half> %0)
   ret half %a
 }
 
 declare half @llvm.riscv.vfmv.f.s.nxv8f16(<vscale x 8 x half>)
 
 define half @intrinsic_vfmv.f.s_s_nxv8f16(<vscale x 8 x half> %0) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv8f16:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vsetvli zero, zero, e16,m2,ta,mu
 ; CHECK-NEXT:    vfmv.f.s fa0, v8
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h killed $f10_f
 ; CHECK-NEXT:    ret
 entry:
   %a = call half @llvm.riscv.vfmv.f.s.nxv8f16(<vscale x 8 x half> %0)
   ret half %a
 }
 
 declare half @llvm.riscv.vfmv.f.s.nxv16f16(<vscale x 16 x half>)
 
 define half @intrinsic_vfmv.f.s_s_nxv16f16(<vscale x 16 x half> %0) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv16f16:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vsetvli zero, zero, e16,m4,ta,mu
 ; CHECK-NEXT:    vfmv.f.s fa0, v8
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h killed $f10_f
 ; CHECK-NEXT:    ret
 entry:
   %a = call half @llvm.riscv.vfmv.f.s.nxv16f16(<vscale x 16 x half> %0)
   ret half %a
 }
 
 declare half @llvm.riscv.vfmv.f.s.nxv32f16(<vscale x 32 x half>)
 
 define half @intrinsic_vfmv.f.s_s_nxv32f16(<vscale x 32 x half> %0) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.f.s_s_nxv32f16:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    vsetvli zero, zero, e16,m8,ta,mu
 ; CHECK-NEXT:    vfmv.f.s fa0, v8
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h killed $f10_f
 ; CHECK-NEXT:    ret
 entry:
   %a = call half @llvm.riscv.vfmv.f.s.nxv32f16(<vscale x 32 x half> %0)
   ret half %a
 }
 
 declare float @llvm.riscv.vfmv.f.s.nxv1f32(<vscale x 1 x float>)
 

llvm/test/CodeGen/RISCV/rvv/vfmv.s.f-rv32.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+d,+experimental-v,+experimental-zfh -target-abi ilp32d -verify-machineinstrs < %s | FileCheck %s
 
 declare <vscale x 1 x half> @llvm.riscv.vfmv.s.f.nxv1f16(<vscale x 1 x half>, half, i32)
 
 define <vscale x 1 x half> @intrinsic_vfmv.s.f_f_nxv1f16(<vscale x 1 x half> %0, half %1, i32 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv1f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 1 x half> @llvm.riscv.vfmv.s.f.nxv1f16(<vscale x 1 x half> %0, half %1, i32 %2)
   ret <vscale x 1 x half> %a
 }
 
 declare <vscale x 2 x half> @llvm.riscv.vfmv.s.f.nxv2f16(<vscale x 2 x half>, half, i32)
 
 define <vscale x 2 x half> @intrinsic_vfmv.s.f_f_nxv2f16(<vscale x 2 x half> %0, half %1, i32 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv2f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 2 x half> @llvm.riscv.vfmv.s.f.nxv2f16(<vscale x 2 x half> %0, half %1, i32 %2)
   ret <vscale x 2 x half> %a
 }
 
 declare <vscale x 4 x half> @llvm.riscv.vfmv.s.f.nxv4f16(<vscale x 4 x half>, half, i32)
 
 define <vscale x 4 x half> @intrinsic_vfmv.s.f_f_nxv4f16(<vscale x 4 x half> %0, half %1, i32 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv4f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 4 x half> @llvm.riscv.vfmv.s.f.nxv4f16(<vscale x 4 x half> %0, half %1, i32 %2)
   ret <vscale x 4 x half> %a
 }
 
 declare <vscale x 8 x half> @llvm.riscv.vfmv.s.f.nxv8f16(<vscale x 8 x half>, half, i32)
 
 define <vscale x 8 x half> @intrinsic_vfmv.s.f_f_nxv8f16(<vscale x 8 x half> %0, half %1, i32 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv8f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 8 x half> @llvm.riscv.vfmv.s.f.nxv8f16(<vscale x 8 x half> %0, half %1, i32 %2)
   ret <vscale x 8 x half> %a
 }
 
 declare <vscale x 16 x half> @llvm.riscv.vfmv.s.f.nxv16f16(<vscale x 16 x half>, half, i32)
 
 define <vscale x 16 x half> @intrinsic_vfmv.s.f_f_nxv16f16(<vscale x 16 x half> %0, half %1, i32 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv16f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m4,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 16 x half> @llvm.riscv.vfmv.s.f.nxv16f16(<vscale x 16 x half> %0, half %1, i32 %2)
   ret <vscale x 16 x half> %a
 }
 
 declare <vscale x 32 x half> @llvm.riscv.vfmv.s.f.nxv32f16(<vscale x 32 x half>, half, i32)
 
 define <vscale x 32 x half> @intrinsic_vfmv.s.f_f_nxv32f16(<vscale x 32 x half> %0, half %1, i32 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv32f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m8,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 32 x half> @llvm.riscv.vfmv.s.f.nxv32f16(<vscale x 32 x half> %0, half %1, i32 %2)
   ret <vscale x 32 x half> %a
 }
 

llvm/test/CodeGen/RISCV/rvv/vfmv.s.f-rv64.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv64 -mattr=+d,+experimental-v,+experimental-zfh -target-abi lp64d -verify-machineinstrs < %s | FileCheck %s
 
 declare <vscale x 1 x half> @llvm.riscv.vfmv.s.f.nxv1f16(<vscale x 1 x half>, half, i64)
 
 define <vscale x 1 x half> @intrinsic_vfmv.s.f_f_nxv1f16(<vscale x 1 x half> %0, half %1, i64 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv1f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 1 x half> @llvm.riscv.vfmv.s.f.nxv1f16(<vscale x 1 x half> %0, half %1, i64 %2)
   ret <vscale x 1 x half> %a
 }
 
 declare <vscale x 2 x half> @llvm.riscv.vfmv.s.f.nxv2f16(<vscale x 2 x half>, half, i64)
 
 define <vscale x 2 x half> @intrinsic_vfmv.s.f_f_nxv2f16(<vscale x 2 x half> %0, half %1, i64 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv2f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 2 x half> @llvm.riscv.vfmv.s.f.nxv2f16(<vscale x 2 x half> %0, half %1, i64 %2)
   ret <vscale x 2 x half> %a
 }
 
 declare <vscale x 4 x half> @llvm.riscv.vfmv.s.f.nxv4f16(<vscale x 4 x half>, half, i64)
 
 define <vscale x 4 x half> @intrinsic_vfmv.s.f_f_nxv4f16(<vscale x 4 x half> %0, half %1, i64 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv4f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 4 x half> @llvm.riscv.vfmv.s.f.nxv4f16(<vscale x 4 x half> %0, half %1, i64 %2)
   ret <vscale x 4 x half> %a
 }
 
 declare <vscale x 8 x half> @llvm.riscv.vfmv.s.f.nxv8f16(<vscale x 8 x half>, half, i64)
 
 define <vscale x 8 x half> @intrinsic_vfmv.s.f_f_nxv8f16(<vscale x 8 x half> %0, half %1, i64 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv8f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 8 x half> @llvm.riscv.vfmv.s.f.nxv8f16(<vscale x 8 x half> %0, half %1, i64 %2)
   ret <vscale x 8 x half> %a
 }
 
 declare <vscale x 16 x half> @llvm.riscv.vfmv.s.f.nxv16f16(<vscale x 16 x half>, half, i64)
 
 define <vscale x 16 x half> @intrinsic_vfmv.s.f_f_nxv16f16(<vscale x 16 x half> %0, half %1, i64 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv16f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m4,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 16 x half> @llvm.riscv.vfmv.s.f.nxv16f16(<vscale x 16 x half> %0, half %1, i64 %2)
   ret <vscale x 16 x half> %a
 }
 
 declare <vscale x 32 x half> @llvm.riscv.vfmv.s.f.nxv32f16(<vscale x 32 x half>, half, i64)
 
 define <vscale x 32 x half> @intrinsic_vfmv.s.f_f_nxv32f16(<vscale x 32 x half> %0, half %1, i64 %2) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.s.f_f_nxv32f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m8,ta,mu
 ; CHECK-NEXT:    vfmv.s.f v8, fa0
 ; CHECK-NEXT:    ret
 entry:
   %a = call <vscale x 32 x half> @llvm.riscv.vfmv.s.f.nxv32f16(<vscale x 32 x half> %0, half %1, i64 %2)
   ret <vscale x 32 x half> %a
 }
 

llvm/test/CodeGen/RISCV/rvv/vfmv.v.f-rv32.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv32 -mattr=+experimental-v,+d,+experimental-zfh -target-abi ilp32d -verify-machineinstrs \
 ; RUN:   --riscv-no-aliases < %s | FileCheck %s
 declare <vscale x 1 x half> @llvm.riscv.vfmv.v.f.nxv1f16(
   half,
   i32);
 
 define <vscale x 1 x half> @intrinsic_vfmv.v.f_f_nxv1f16(half %0, i32 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv1f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 1 x half> @llvm.riscv.vfmv.v.f.nxv1f16(
     half %0,
     i32 %1)
 
   ret <vscale x 1 x half> %a
 }
 
 declare <vscale x 2 x half> @llvm.riscv.vfmv.v.f.nxv2f16(
   half,
   i32);
 
 define <vscale x 2 x half> @intrinsic_vfmv.v.f_f_nxv2f16(half %0, i32 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv2f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 2 x half> @llvm.riscv.vfmv.v.f.nxv2f16(
     half %0,
     i32 %1)
 
   ret <vscale x 2 x half> %a
 }
 
 declare <vscale x 4 x half> @llvm.riscv.vfmv.v.f.nxv4f16(
   half,
   i32);
 
 define <vscale x 4 x half> @intrinsic_vfmv.v.f_f_nxv4f16(half %0, i32 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv4f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 4 x half> @llvm.riscv.vfmv.v.f.nxv4f16(
     half %0,
     i32 %1)
 
   ret <vscale x 4 x half> %a
 }
 
 declare <vscale x 8 x half> @llvm.riscv.vfmv.v.f.nxv8f16(
   half,
   i32);
 
 define <vscale x 8 x half> @intrinsic_vfmv.v.f_f_nxv8f16(half %0, i32 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv8f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 8 x half> @llvm.riscv.vfmv.v.f.nxv8f16(
     half %0,
     i32 %1)
 
   ret <vscale x 8 x half> %a
 }
 
 declare <vscale x 16 x half> @llvm.riscv.vfmv.v.f.nxv16f16(
   half,
   i32);
 
 define <vscale x 16 x half> @intrinsic_vfmv.v.f_f_nxv16f16(half %0, i32 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv16f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m4,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 16 x half> @llvm.riscv.vfmv.v.f.nxv16f16(
     half %0,
     i32 %1)
 
   ret <vscale x 16 x half> %a
 }
 
 declare <vscale x 32 x half> @llvm.riscv.vfmv.v.f.nxv32f16(
   half,
   i32);
 
 define <vscale x 32 x half> @intrinsic_vfmv.v.f_f_nxv32f16(half %0, i32 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv32f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m8,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 32 x half> @llvm.riscv.vfmv.v.f.nxv32f16(
     half %0,
     i32 %1)
 

llvm/test/CodeGen/RISCV/rvv/vfmv.v.f-rv64.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+d,+experimental-zfh -target-abi lp64d -verify-machineinstrs \
 ; RUN:   --riscv-no-aliases < %s | FileCheck %s
 declare <vscale x 1 x half> @llvm.riscv.vfmv.v.f.nxv1f16(
   half,
   i64);
 
 define <vscale x 1 x half> @intrinsic_vfmv.v.f_f_nxv1f16(half %0, i64 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv1f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 1 x half> @llvm.riscv.vfmv.v.f.nxv1f16(
     half %0,
     i64 %1)
 
   ret <vscale x 1 x half> %a
 }
 
 declare <vscale x 2 x half> @llvm.riscv.vfmv.v.f.nxv2f16(
   half,
   i64);
 
 define <vscale x 2 x half> @intrinsic_vfmv.v.f_f_nxv2f16(half %0, i64 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv2f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 2 x half> @llvm.riscv.vfmv.v.f.nxv2f16(
     half %0,
     i64 %1)
 
   ret <vscale x 2 x half> %a
 }
 
 declare <vscale x 4 x half> @llvm.riscv.vfmv.v.f.nxv4f16(
   half,
   i64);
 
 define <vscale x 4 x half> @intrinsic_vfmv.v.f_f_nxv4f16(half %0, i64 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv4f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 4 x half> @llvm.riscv.vfmv.v.f.nxv4f16(
     half %0,
     i64 %1)
 
   ret <vscale x 4 x half> %a
 }
 
 declare <vscale x 8 x half> @llvm.riscv.vfmv.v.f.nxv8f16(
   half,
   i64);
 
 define <vscale x 8 x half> @intrinsic_vfmv.v.f_f_nxv8f16(half %0, i64 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv8f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 8 x half> @llvm.riscv.vfmv.v.f.nxv8f16(
     half %0,
     i64 %1)
 
   ret <vscale x 8 x half> %a
 }
 
 declare <vscale x 16 x half> @llvm.riscv.vfmv.v.f.nxv16f16(
   half,
   i64);
 
 define <vscale x 16 x half> @intrinsic_vfmv.v.f_f_nxv16f16(half %0, i64 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv16f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m4,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 16 x half> @llvm.riscv.vfmv.v.f.nxv16f16(
     half %0,
     i64 %1)
 
   ret <vscale x 16 x half> %a
 }
 
 declare <vscale x 32 x half> @llvm.riscv.vfmv.v.f.nxv32f16(
   half,
   i64);
 
 define <vscale x 32 x half> @intrinsic_vfmv.v.f_f_nxv32f16(half %0, i64 %1) nounwind {
 ; CHECK-LABEL: intrinsic_vfmv.v.f_f_nxv32f16:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    # kill: def $f10_h killed $f10_h def $f10_f
 ; CHECK-NEXT:    vsetvli a0, a0, e16,m8,ta,mu
 ; CHECK-NEXT:    vfmv.v.f v8, fa0
 ; CHECK-NEXT:    jalr zero, 0(ra)
 entry:
   %a = call <vscale x 32 x half> @llvm.riscv.vfmv.v.f.nxv32f16(
     half %0,
     i64 %1)