[RISCV] Add intrinsics for vfwmacc, vfwnmacc, vfwmsac, vfwnmsac instructions

[arcbbb]

Summary

This patch defines vfwmacc, vfwnmacc, vfwmsac, vfwnmsac intrinsics and lower to V instructions.
We work with @rogfer01 from BSC to come out this patch.

Authored-by: Roger Ferrer Ibanez <[e-mail]>
Co-Authored-by: ShihPo Hung <[e-mail]>

[arcbbb]

Updates

1. rebased
2. remove unneeded !eq
3. use true/false instead of 1/0

[craig.topper]

LGTM

llvm/include/llvm/IR/IntrinsicsRISCV.td

let TargetPrefix = "riscv" in {
   defm vfnmacc : RISCVTernaryAAXA;
   defm vfmsac : RISCVTernaryAAXA;
   defm vfnmsac : RISCVTernaryAAXA;
   defm vfmadd : RISCVTernaryAAXA;
   defm vfnmadd : RISCVTernaryAAXA;
   defm vfmsub : RISCVTernaryAAXA;
   defm vfnmsub : RISCVTernaryAAXA;
 
+  defm vfwmacc : RISCVTernaryWide;
+  defm vfwnmacc : RISCVTernaryWide;
+  defm vfwmsac : RISCVTernaryWide;
+  defm vfwnmsac : RISCVTernaryWide;
+
   defm vfmin : RISCVBinaryAAX;
   defm vfmax : RISCVBinaryAAX;
 
   defm vfsgnj : RISCVBinaryAAX;
   defm vfsgnjn : RISCVBinaryAAX;
   defm vfsgnjx : RISCVBinaryAAX;
 
   defm vfmerge : RISCVBinaryWithV0;

llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td

 }
 
 multiclass VPseudoTernaryW_VV {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxList.m in
     defm _VV : VPseudoTernary<m.wvrclass, m.vrclass, m.vrclass, m, constraint>;
 }
 
-multiclass VPseudoTernaryW_VX {
+multiclass VPseudoTernaryW_VX<bit IsFloat> {
   defvar constraint = "@earlyclobber $rd";
   foreach m = MxList.m in
-    defm _VX : VPseudoTernary<m.wvrclass, GPR, m.vrclass, m, constraint>;
+    defm !if(IsFloat, "_VF", "_VX")  : VPseudoTernary<m.wvrclass,
+                             !if(IsFloat, FPR32, GPR), 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 = ""> {
   defm "" : VPseudoTernaryV_VV<Constraint>;
   defm "" : VPseudoTernaryV_VX_AAXA<IsFloat, Constraint>;
 }
 
 multiclass VPseudoTernaryV_VX_VI<Operand ImmType = simm5, string Constraint = ""> {
   defm "" : VPseudoTernaryV_VX<Constraint>;
   defm "" : VPseudoTernaryV_VI<ImmType, Constraint>;
 }
 
-multiclass VPseudoTernaryW_VV_VX {
+multiclass VPseudoTernaryW_VV_VX<bit IsFloat = false> {
   defm "" : VPseudoTernaryW_VV;
-  defm "" : VPseudoTernaryW_VX;
+  defm "" : VPseudoTernaryW_VX<IsFloat>;
 }
 
 multiclass VPseudoBinaryM_VV_VX_VI {
   defm "" : VPseudoBinaryM_VV;
   defm "" : VPseudoBinaryM_VX</*IsFloat=*/0>;
   defm "" : VPseudoBinaryM_VI;
 }
 
multiclass VPatTernaryW_VV<string intrinsic, string instruction,
   }
 }
 
 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, "VX",
-                      wti.Vector, XLenVT, vti.Vector,
+    defm : VPatTernary<intrinsic, instruction,
+                       !if(!eq(vti.Scalar, XLenVT), "VX", "VF"),
+                       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 "" : VPatTernaryV_VV<intrinsic, instruction, vtilist>;
   defm "" : VPatTernaryV_VX_AAXA<intrinsic, instruction, 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;
+defm PseudoVWMACCUS    : VPseudoTernaryW_VX</*IsFloat*/false>;
 
 //===----------------------------------------------------------------------===//
 // 12.15. Vector Integer Merge Instructions
 //===----------------------------------------------------------------------===//
 defm PseudoVMERGE      : VPseudoBinaryV_VM_XM_IM;
 
 //===----------------------------------------------------------------------===//
 // 12.17. Vector Integer Move Instructions

 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>;
 
 //===----------------------------------------------------------------------===//
+// 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>;
+
+//===----------------------------------------------------------------------===//
 // 14.9. Vector Floating-Point Min/Max Instructions
 //===----------------------------------------------------------------------===//
 defm PseudoVFMIN       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
 defm PseudoVFMAX       : VPseudoBinaryV_VV_VX</*IsFloat=*/1>;
 
 //===----------------------------------------------------------------------===//
 // 14.12. Vector Floating-Point Sign-Injection Instructions
 //===----------------------------------------------------------------------===//

 defm "" : VPatTernaryV_VV_VX_AAXA<"int_riscv_vfmsac", "PseudoVFMSAC", AllFloatVectors>;
 defm "" : VPatTernaryV_VV_VX_AAXA<"int_riscv_vfnmsac", "PseudoVFNMSAC", AllFloatVectors>;
 defm "" : VPatTernaryV_VV_VX_AAXA<"int_riscv_vfmadd", "PseudoVFMADD", AllFloatVectors>;
 defm "" : VPatTernaryV_VV_VX_AAXA<"int_riscv_vfnmadd", "PseudoVFNMADD", AllFloatVectors>;
 defm "" : VPatTernaryV_VV_VX_AAXA<"int_riscv_vfmsub", "PseudoVFMSUB", AllFloatVectors>;
 defm "" : VPatTernaryV_VV_VX_AAXA<"int_riscv_vfnmsub", "PseudoVFNMSUB", AllFloatVectors>;
 
 //===----------------------------------------------------------------------===//
+// 14.7. Vector Widening Floating-Point Fused Multiply-Add Instructions
+//===----------------------------------------------------------------------===//
+defm "" : VPatTernaryW_VV_VX<"int_riscv_vfwmacc", "PseudoVFWMACC", AllWidenableFloatVectors>;
+defm "" : VPatTernaryW_VV_VX<"int_riscv_vfwnmacc", "PseudoVFWNMACC", AllWidenableFloatVectors>;
+defm "" : VPatTernaryW_VV_VX<"int_riscv_vfwmsac", "PseudoVFWMSAC", AllWidenableFloatVectors>;
+defm "" : VPatTernaryW_VV_VX<"int_riscv_vfwnmsac", "PseudoVFWNMSAC", AllWidenableFloatVectors>;
+
+//===----------------------------------------------------------------------===//
 // 14.9. Vector Floating-Point Min/Max Instructions
 //===----------------------------------------------------------------------===//
 defm "" : VPatBinaryV_VV_VX<"int_riscv_vfmin", "PseudoVFMIN", AllFloatVectors>;
 defm "" : VPatBinaryV_VV_VX<"int_riscv_vfmax", "PseudoVFMAX", AllFloatVectors>;
 
 //===----------------------------------------------------------------------===//
 // 14.12. Vector Floating-Point Sign-Injection Instructions
 //===----------------------------------------------------------------------===//

llvm/test/CodeGen/RISCV/rvv/vfwmacc-rv32.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v,+f,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwmacc_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwmacc_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwmacc_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwmacc_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwmacc_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwmacc_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float> @intrinsic_vfwmacc_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwmacc_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float> @intrinsic_vfwmacc_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwmacc_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float> @intrinsic_vfwmacc_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwmacc_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float> @intrinsic_vfwmacc_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwmacc_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float> @intrinsic_vfwmacc_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}

llvm/test/CodeGen/RISCV/rvv/vfwmacc-rv64.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+d,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwmacc_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwmacc_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwmacc_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwmacc_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwmacc_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmacc.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwmacc_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmacc.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmacc.mask.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwmacc_mask_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmacc.mask.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmacc.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwmacc_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmacc.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmacc.mask.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwmacc_mask_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmacc.mask.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmacc.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwmacc_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmacc.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmacc.mask.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwmacc_mask_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmacc.mask.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmacc.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwmacc_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmacc.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmacc.mask.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwmacc_mask_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmacc.mask.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwmacc_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float> @intrinsic_vfwmacc_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmacc.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwmacc_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float> @intrinsic_vfwmacc_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmacc.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwmacc_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float> @intrinsic_vfwmacc_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmacc.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwmacc_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float> @intrinsic_vfwmacc_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmacc.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwmacc_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float> @intrinsic_vfwmacc_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmacc.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmacc.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwmacc_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmacc.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmacc.mask.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double> @intrinsic_vfwmacc_mask_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmacc.mask.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmacc.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwmacc_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmacc.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmacc.mask.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double> @intrinsic_vfwmacc_mask_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmacc.mask.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmacc.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwmacc_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmacc.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmacc.mask.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double> @intrinsic_vfwmacc_mask_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmacc.mask.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmacc.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwmacc_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmacc.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmacc.mask.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double> @intrinsic_vfwmacc_mask_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmacc_mask_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmacc.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmacc.mask.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}

llvm/test/CodeGen/RISCV/rvv/vfwmsac-rv32.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v,+f,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwmsac_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwmsac_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwmsac_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwmsac_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwmsac_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwmsac_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float> @intrinsic_vfwmsac_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwmsac_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float> @intrinsic_vfwmsac_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwmsac_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float> @intrinsic_vfwmsac_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwmsac_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float> @intrinsic_vfwmsac_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwmsac_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float> @intrinsic_vfwmsac_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}

llvm/test/CodeGen/RISCV/rvv/vfwmsac-rv64.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+d,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwmsac_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwmsac_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwmsac_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwmsac_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwmsac_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmsac.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwmsac_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmsac.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmsac.mask.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwmsac_mask_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmsac.mask.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmsac.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwmsac_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmsac.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmsac.mask.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwmsac_mask_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmsac.mask.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmsac.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwmsac_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmsac.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmsac.mask.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwmsac_mask_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmsac.mask.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmsac.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwmsac_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmsac.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmsac.mask.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwmsac_mask_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmsac.mask.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwmsac_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float> @intrinsic_vfwmsac_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwmsac.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwmsac_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float> @intrinsic_vfwmsac_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwmsac.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwmsac_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float> @intrinsic_vfwmsac_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwmsac.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwmsac_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float> @intrinsic_vfwmsac_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwmsac.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwmsac_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float> @intrinsic_vfwmsac_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwmsac.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmsac.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwmsac_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmsac.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwmsac.mask.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double> @intrinsic_vfwmsac_mask_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwmsac.mask.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmsac.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwmsac_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmsac.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwmsac.mask.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double> @intrinsic_vfwmsac_mask_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwmsac.mask.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmsac.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwmsac_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmsac.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwmsac.mask.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double> @intrinsic_vfwmsac_mask_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwmsac.mask.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmsac.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwmsac_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmsac.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwmsac.mask.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double> @intrinsic_vfwmsac_mask_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwmsac_mask_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwmsac.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwmsac.mask.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}

llvm/test/CodeGen/RISCV/rvv/vfwnmacc-rv32.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v,+f,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmacc_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmacc_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmacc_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmacc_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmacc_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmacc_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float> @intrinsic_vfwnmacc_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmacc_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float> @intrinsic_vfwnmacc_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmacc_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float> @intrinsic_vfwnmacc_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmacc_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float> @intrinsic_vfwnmacc_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmacc_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float> @intrinsic_vfwnmacc_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}

llvm/test/CodeGen/RISCV/rvv/vfwnmacc-rv64.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+d,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmacc_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmacc_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmacc_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmacc_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmacc_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmacc.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwnmacc_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmacc.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmacc.mask.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwnmacc_mask_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmacc.mask.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmacc.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwnmacc_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmacc.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmacc.mask.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwnmacc_mask_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmacc.mask.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmacc.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwnmacc_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmacc.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmacc.mask.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwnmacc_mask_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmacc.mask.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmacc.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwnmacc_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmacc.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmacc.mask.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwnmacc_mask_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmacc.mask.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmacc_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float> @intrinsic_vfwnmacc_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmacc.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmacc_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float> @intrinsic_vfwnmacc_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmacc.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmacc_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float> @intrinsic_vfwnmacc_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmacc.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmacc_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float> @intrinsic_vfwnmacc_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmacc.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmacc_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float> @intrinsic_vfwnmacc_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmacc.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmacc.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwnmacc_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmacc.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmacc.mask.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double> @intrinsic_vfwnmacc_mask_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmacc.mask.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmacc.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwnmacc_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmacc.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmacc.mask.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double> @intrinsic_vfwnmacc_mask_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmacc.mask.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmacc.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwnmacc_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmacc.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmacc.mask.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double> @intrinsic_vfwnmacc_mask_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmacc.mask.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmacc.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwnmacc_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmacc.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmacc.mask.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double> @intrinsic_vfwnmacc_mask_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmacc_mask_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmacc.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmacc.mask.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}

llvm/test/CodeGen/RISCV/rvv/vfwnmsac-rv32.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v,+f,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmsac_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmsac_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmsac_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmsac_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmsac_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i32);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmsac_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i32 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i32);
+
+define <vscale x 1 x float> @intrinsic_vfwnmsac_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i32);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmsac_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i32 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i32);
+
+define <vscale x 2 x float> @intrinsic_vfwnmsac_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i32);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmsac_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i32 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i32);
+
+define <vscale x 4 x float> @intrinsic_vfwnmsac_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i32);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmsac_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i32 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i32);
+
+define <vscale x 8 x float> @intrinsic_vfwnmsac_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i32);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmsac_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i32 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i32 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i32);
+
+define <vscale x 16 x float> @intrinsic_vfwnmsac_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i32 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i32 %4)
+
+  ret <vscale x 16 x float> %a
+}

llvm/test/CodeGen/RISCV/rvv/vfwnmsac-rv64.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v,+d,+experimental-zfh -verify-machineinstrs \
+; RUN:   --riscv-no-aliases < %s | FileCheck %s
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmsac_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.nxv1f16(
+  <vscale x 1 x float>,
+  <vscale x 1 x half>,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16(<vscale x 1 x float> %0, <vscale x 1 x half> %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv1f32_nxv1f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.nxv1f16(
+    <vscale x 1 x float> %0,
+    <vscale x 1 x half> %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmsac_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.nxv2f16(
+  <vscale x 2 x float>,
+  <vscale x 2 x half>,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16(<vscale x 2 x float> %0, <vscale x 2 x half> %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv2f32_nxv2f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.nxv2f16(
+    <vscale x 2 x float> %0,
+    <vscale x 2 x half> %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmsac_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.nxv4f16(
+  <vscale x 4 x float>,
+  <vscale x 4 x half>,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16(<vscale x 4 x float> %0, <vscale x 4 x half> %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv4f32_nxv4f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.nxv4f16(
+    <vscale x 4 x float> %0,
+    <vscale x 4 x half> %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmsac_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.nxv8f16(
+  <vscale x 8 x float>,
+  <vscale x 8 x half>,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16(<vscale x 8 x float> %0, <vscale x 8 x half> %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv8f32_nxv8f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.nxv8f16(
+    <vscale x 8 x float> %0,
+    <vscale x 8 x half> %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmsac_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.nxv16f16(
+  <vscale x 16 x float>,
+  <vscale x 16 x half>,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16(<vscale x 16 x float> %0, <vscale x 16 x half> %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv16f32_nxv16f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    vle16.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.nxv16f16(
+    <vscale x 16 x float> %0,
+    <vscale x 16 x half> %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmsac.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwnmsac_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmsac.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmsac.mask.nxv1f64.nxv1f32(
+  <vscale x 1 x double>,
+  <vscale x 1 x float>,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwnmsac_mask_vv_nxv1f64_nxv1f32_nxv1f32(<vscale x 1 x double> %0, <vscale x 1 x float> %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv1f64_nxv1f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v17, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmsac.mask.nxv1f64.nxv1f32(
+    <vscale x 1 x double> %0,
+    <vscale x 1 x float> %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmsac.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwnmsac_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v18, v19
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmsac.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmsac.mask.nxv2f64.nxv2f32(
+  <vscale x 2 x double>,
+  <vscale x 2 x float>,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwnmsac_mask_vv_nxv2f64_nxv2f32_nxv2f32(<vscale x 2 x double> %0, <vscale x 2 x float> %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv2f64_nxv2f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v18, v19, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmsac.mask.nxv2f64.nxv2f32(
+    <vscale x 2 x double> %0,
+    <vscale x 2 x float> %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmsac.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwnmsac_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v20, v22
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmsac.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmsac.mask.nxv4f64.nxv4f32(
+  <vscale x 4 x double>,
+  <vscale x 4 x float>,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwnmsac_mask_vv_nxv4f64_nxv4f32_nxv4f32(<vscale x 4 x double> %0, <vscale x 4 x float> %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv4f64_nxv4f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a0, a0, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v20, v22, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmsac.mask.nxv4f64.nxv4f32(
+    <vscale x 4 x double> %0,
+    <vscale x 4 x float> %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmsac.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwnmsac_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v8, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmsac.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmsac.mask.nxv8f64.nxv8f32(
+  <vscale x 8 x double>,
+  <vscale x 8 x float>,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwnmsac_mask_vv_nxv8f64_nxv8f32_nxv8f32(<vscale x 8 x double> %0, <vscale x 8 x float> %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vv_nxv8f64_nxv8f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    vle32.v v8, (a0)
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vv v16, v8, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmsac.mask.nxv8f64.nxv8f32(
+    <vscale x 8 x double> %0,
+    <vscale x 8 x float> %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  i64);
+
+define <vscale x 1 x float>  @intrinsic_vfwnmsac_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    i64 %3)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.f16(
+  <vscale x 1 x float>,
+  half,
+  <vscale x 1 x half>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x float> @intrinsic_vfwnmsac_mask_vf_nxv1f32_f16_nxv1f16(<vscale x 1 x float> %0, half %1, <vscale x 1 x half> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv1f32_f16_nxv1f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x float> @llvm.riscv.vfwnmsac.mask.nxv1f32.f16(
+    <vscale x 1 x float> %0,
+    half %1,
+    <vscale x 1 x half> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  i64);
+
+define <vscale x 2 x float>  @intrinsic_vfwnmsac_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    i64 %3)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.f16(
+  <vscale x 2 x float>,
+  half,
+  <vscale x 2 x half>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x float> @intrinsic_vfwnmsac_mask_vf_nxv2f32_f16_nxv2f16(<vscale x 2 x float> %0, half %1, <vscale x 2 x half> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv2f32_f16_nxv2f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x float> @llvm.riscv.vfwnmsac.mask.nxv2f32.f16(
+    <vscale x 2 x float> %0,
+    half %1,
+    <vscale x 2 x half> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  i64);
+
+define <vscale x 4 x float>  @intrinsic_vfwnmsac_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    i64 %3)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.f16(
+  <vscale x 4 x float>,
+  half,
+  <vscale x 4 x half>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x float> @intrinsic_vfwnmsac_mask_vf_nxv4f32_f16_nxv4f16(<vscale x 4 x float> %0, half %1, <vscale x 4 x half> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv4f32_f16_nxv4f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x float> @llvm.riscv.vfwnmsac.mask.nxv4f32.f16(
+    <vscale x 4 x float> %0,
+    half %1,
+    <vscale x 4 x half> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  i64);
+
+define <vscale x 8 x float>  @intrinsic_vfwnmsac_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    i64 %3)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.f16(
+  <vscale x 8 x float>,
+  half,
+  <vscale x 8 x half>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x float> @intrinsic_vfwnmsac_mask_vf_nxv8f32_f16_nxv8f16(<vscale x 8 x float> %0, half %1, <vscale x 8 x half> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv8f32_f16_nxv8f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e16,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x float> @llvm.riscv.vfwnmsac.mask.nxv8f32.f16(
+    <vscale x 8 x float> %0,
+    half %1,
+    <vscale x 8 x half> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  i64);
+
+define <vscale x 16 x float>  @intrinsic_vfwnmsac_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    i64 %3)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.f16(
+  <vscale x 16 x float>,
+  half,
+  <vscale x 16 x half>,
+  <vscale x 16 x i1>,
+  i64);
+
+define <vscale x 16 x float> @intrinsic_vfwnmsac_mask_vf_nxv16f32_f16_nxv16f16(<vscale x 16 x float> %0, half %1, <vscale x 16 x half> %2, <vscale x 16 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv16f32_f16_nxv16f16:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e16,m4,ta,mu
+; CHECK-NEXT:    vle16.v v28, (a1)
+; CHECK-NEXT:    fmv.h.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e16,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 16 x float> @llvm.riscv.vfwnmsac.mask.nxv16f32.f16(
+    <vscale x 16 x float> %0,
+    half %1,
+    <vscale x 16 x half> %2,
+    <vscale x 16 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 16 x float> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmsac.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  i64);
+
+define <vscale x 1 x double>  @intrinsic_vfwnmsac_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmsac.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    i64 %3)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 1 x double> @llvm.riscv.vfwnmsac.mask.nxv1f64.f32(
+  <vscale x 1 x double>,
+  float,
+  <vscale x 1 x float>,
+  <vscale x 1 x i1>,
+  i64);
+
+define <vscale x 1 x double> @intrinsic_vfwnmsac_mask_vf_nxv1f64_f32_nxv1f32(<vscale x 1 x double> %0, float %1, <vscale x 1 x float> %2, <vscale x 1 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv1f64_f32_nxv1f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,mf2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v17, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 1 x double> @llvm.riscv.vfwnmsac.mask.nxv1f64.f32(
+    <vscale x 1 x double> %0,
+    float %1,
+    <vscale x 1 x float> %2,
+    <vscale x 1 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 1 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmsac.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  i64);
+
+define <vscale x 2 x double>  @intrinsic_vfwnmsac_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v18
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmsac.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    i64 %3)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 2 x double> @llvm.riscv.vfwnmsac.mask.nxv2f64.f32(
+  <vscale x 2 x double>,
+  float,
+  <vscale x 2 x float>,
+  <vscale x 2 x i1>,
+  i64);
+
+define <vscale x 2 x double> @intrinsic_vfwnmsac_mask_vf_nxv2f64_f32_nxv2f32(<vscale x 2 x double> %0, float %1, <vscale x 2 x float> %2, <vscale x 2 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv2f64_f32_nxv2f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m1,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v18, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 2 x double> @llvm.riscv.vfwnmsac.mask.nxv2f64.f32(
+    <vscale x 2 x double> %0,
+    float %1,
+    <vscale x 2 x float> %2,
+    <vscale x 2 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 2 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmsac.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  i64);
+
+define <vscale x 4 x double>  @intrinsic_vfwnmsac_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v20
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmsac.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    i64 %3)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 4 x double> @llvm.riscv.vfwnmsac.mask.nxv4f64.f32(
+  <vscale x 4 x double>,
+  float,
+  <vscale x 4 x float>,
+  <vscale x 4 x i1>,
+  i64);
+
+define <vscale x 4 x double> @intrinsic_vfwnmsac_mask_vf_nxv4f64_f32_nxv4f32(<vscale x 4 x double> %0, float %1, <vscale x 4 x float> %2, <vscale x 4 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv4f64_f32_nxv4f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a1, e32,m2,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v20, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 4 x double> @llvm.riscv.vfwnmsac.mask.nxv4f64.f32(
+    <vscale x 4 x double> %0,
+    float %1,
+    <vscale x 4 x float> %2,
+    <vscale x 4 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 4 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmsac.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  i64);
+
+define <vscale x 8 x double>  @intrinsic_vfwnmsac_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, i64 %3) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v28
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmsac.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    i64 %3)
+
+  ret <vscale x 8 x double> %a
+}
+
+declare <vscale x 8 x double> @llvm.riscv.vfwnmsac.mask.nxv8f64.f32(
+  <vscale x 8 x double>,
+  float,
+  <vscale x 8 x float>,
+  <vscale x 8 x i1>,
+  i64);
+
+define <vscale x 8 x double> @intrinsic_vfwnmsac_mask_vf_nxv8f64_f32_nxv8f32(<vscale x 8 x double> %0, float %1, <vscale x 8 x float> %2, <vscale x 8 x i1> %3, i64 %4) nounwind {
+; CHECK-LABEL: intrinsic_vfwnmsac_mask_vf_nxv8f64_f32_nxv8f32:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetvli a3, zero, e32,m4,ta,mu
+; CHECK-NEXT:    vle32.v v28, (a1)
+; CHECK-NEXT:    fmv.w.x ft0, a0
+; CHECK-NEXT:    vsetvli a0, a2, e32,m4,ta,mu
+; CHECK-NEXT:    vfwnmsac.vf v16, ft0, v28, v0.t
+; CHECK-NEXT:    jalr zero, 0(ra)
+entry:
+  %a = call <vscale x 8 x double> @llvm.riscv.vfwnmsac.mask.nxv8f64.f32(
+    <vscale x 8 x double> %0,
+    float %1,
+    <vscale x 8 x float> %2,
+    <vscale x 8 x i1> %3,
+    i64 %4)
+
+  ret <vscale x 8 x double> %a
+}