diff --git a/llvm/test/CodeGen/RISCV/rvv/vfadd-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/vfadd-sdnode.ll
index d5786660f7030..68226d2bbd16c 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vfadd-sdnode.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vfadd-sdnode.ll
@@ -11,10 +11,10 @@
 ; RUN: llc -mtriple=riscv64 -mattr=+d,+zfhmin,+zvfhmin,+zfbfmin,+zvfbfmin,+v \
 ; RUN:     -target-abi=lp64d -verify-machineinstrs < %s | FileCheck %s \
 ; RUN:     --check-prefixes=CHECK,ZVFHMIN
-; RUN: llc -mtriple=riscv32 -mattr=+d,+zvfh,+experimental-zvfbfa,+v \
+; RUN: llc -mtriple=riscv32 -mattr=+d,+zvfhmin,+experimental-zvfbfa,+v \
 ; RUN:     -verify-machineinstrs < %s | FileCheck %s \
 ; RUN:     --check-prefixes=CHECK,ZVFBFA
-; RUN: llc -mtriple=riscv64 -mattr=+d,+zvfh,+experimental-zvfbfa,+v \
+; RUN: llc -mtriple=riscv64 -mattr=+d,+zvfhmin,+experimental-zvfbfa,+v \
 ; RUN:     -verify-machineinstrs < %s | FileCheck %s \
 ; RUN:     --check-prefixes=CHECK,ZVFBFA
 
@@ -563,7 +563,12 @@ define <vscale x 1 x half> @vfadd_vv_nxv1f16(<vscale x 1 x half> %va, <vscale x
 ; ZVFBFA-LABEL: vfadd_vv_nxv1f16:
 ; ZVFBFA:       # %bb.0:
 ; ZVFBFA-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
-; ZVFBFA-NEXT:    vfadd.vv v8, v8, v9
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v10, v9
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v9, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v9, v9, v10
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v9
 ; ZVFBFA-NEXT:    ret
   %vc = fadd <vscale x 1 x half> %va, %vb
   ret <vscale x 1 x half> %vc
@@ -589,8 +594,15 @@ define <vscale x 1 x half> @vfadd_vf_nxv1f16(<vscale x 1 x half> %va, half %b) {
 ;
 ; ZVFBFA-LABEL: vfadd_vf_nxv1f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, mf4, ta, ma
-; ZVFBFA-NEXT:    vfadd.vf v8, v8, fa0
+; ZVFBFA-NEXT:    fmv.x.w a0, fa0
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, mf4, ta, ma
+; ZVFBFA-NEXT:    vmv.v.x v9, a0
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v10, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v8, v9
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, mf2, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v9, v10, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, mf4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v9
 ; ZVFBFA-NEXT:    ret
   %head = insertelement <vscale x 1 x half> poison, half %b, i32 0
   %splat = shufflevector <vscale x 1 x half> %head, <vscale x 1 x half> poison, <vscale x 1 x i32> zeroinitializer
@@ -619,7 +631,12 @@ define <vscale x 2 x half> @vfadd_vv_nxv2f16(<vscale x 2 x half> %va, <vscale x
 ; ZVFBFA-LABEL: vfadd_vv_nxv2f16:
 ; ZVFBFA:       # %bb.0:
 ; ZVFBFA-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
-; ZVFBFA-NEXT:    vfadd.vv v8, v8, v9
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v10, v9
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v9, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v9, v9, v10
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v9
 ; ZVFBFA-NEXT:    ret
   %vc = fadd <vscale x 2 x half> %va, %vb
   ret <vscale x 2 x half> %vc
@@ -645,8 +662,15 @@ define <vscale x 2 x half> @vfadd_vf_nxv2f16(<vscale x 2 x half> %va, half %b) {
 ;
 ; ZVFBFA-LABEL: vfadd_vf_nxv2f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, mf2, ta, ma
-; ZVFBFA-NEXT:    vfadd.vf v8, v8, fa0
+; ZVFBFA-NEXT:    fmv.x.w a0, fa0
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, mf2, ta, ma
+; ZVFBFA-NEXT:    vmv.v.x v9, a0
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v10, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v8, v9
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m1, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v9, v10, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, mf2, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v9
 ; ZVFBFA-NEXT:    ret
   %head = insertelement <vscale x 2 x half> poison, half %b, i32 0
   %splat = shufflevector <vscale x 2 x half> %head, <vscale x 2 x half> poison, <vscale x 2 x i32> zeroinitializer
@@ -675,7 +699,12 @@ define <vscale x 4 x half> @vfadd_vv_nxv4f16(<vscale x 4 x half> %va, <vscale x
 ; ZVFBFA-LABEL: vfadd_vv_nxv4f16:
 ; ZVFBFA:       # %bb.0:
 ; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
-; ZVFBFA-NEXT:    vfadd.vv v8, v8, v9
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v10, v9
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v12, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v10, v12, v10
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v10
 ; ZVFBFA-NEXT:    ret
   %vc = fadd <vscale x 4 x half> %va, %vb
   ret <vscale x 4 x half> %vc
@@ -701,8 +730,15 @@ define <vscale x 4 x half> @vfadd_vf_nxv4f16(<vscale x 4 x half> %va, half %b) {
 ;
 ; ZVFBFA-LABEL: vfadd_vf_nxv4f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m1, ta, ma
-; ZVFBFA-NEXT:    vfadd.vf v8, v8, fa0
+; ZVFBFA-NEXT:    fmv.x.w a0, fa0
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, m1, ta, ma
+; ZVFBFA-NEXT:    vmv.v.x v12, a0
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v10, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v8, v12
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m2, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v10, v10, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m1, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v10
 ; ZVFBFA-NEXT:    ret
   %head = insertelement <vscale x 4 x half> poison, half %b, i32 0
   %splat = shufflevector <vscale x 4 x half> %head, <vscale x 4 x half> poison, <vscale x 4 x i32> zeroinitializer
@@ -731,7 +767,12 @@ define <vscale x 8 x half> @vfadd_vv_nxv8f16(<vscale x 8 x half> %va, <vscale x
 ; ZVFBFA-LABEL: vfadd_vv_nxv8f16:
 ; ZVFBFA:       # %bb.0:
 ; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
-; ZVFBFA-NEXT:    vfadd.vv v8, v8, v10
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v12, v10
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v16, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v12, v16, v12
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v12
 ; ZVFBFA-NEXT:    ret
   %vc = fadd <vscale x 8 x half> %va, %vb
   ret <vscale x 8 x half> %vc
@@ -757,8 +798,15 @@ define <vscale x 8 x half> @vfadd_vf_nxv8f16(<vscale x 8 x half> %va, half %b) {
 ;
 ; ZVFBFA-LABEL: vfadd_vf_nxv8f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
-; ZVFBFA-NEXT:    vfadd.vf v8, v8, fa0
+; ZVFBFA-NEXT:    fmv.x.w a0, fa0
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; ZVFBFA-NEXT:    vmv.v.x v16, a0
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v12, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v8, v16
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v12, v12, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v12
 ; ZVFBFA-NEXT:    ret
   %head = insertelement <vscale x 8 x half> poison, half %b, i32 0
   %splat = shufflevector <vscale x 8 x half> %head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
@@ -786,8 +834,15 @@ define <vscale x 8 x half> @vfadd_fv_nxv8f16(<vscale x 8 x half> %va, half %b) {
 ;
 ; ZVFBFA-LABEL: vfadd_fv_nxv8f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m2, ta, ma
-; ZVFBFA-NEXT:    vfadd.vf v8, v8, fa0
+; ZVFBFA-NEXT:    fmv.x.w a0, fa0
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, m2, ta, ma
+; ZVFBFA-NEXT:    vmv.v.x v16, a0
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v12, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v8, v16
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m4, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v12, v8, v12
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m2, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v12
 ; ZVFBFA-NEXT:    ret
   %head = insertelement <vscale x 8 x half> poison, half %b, i32 0
   %splat = shufflevector <vscale x 8 x half> %head, <vscale x 8 x half> poison, <vscale x 8 x i32> zeroinitializer
@@ -816,7 +871,12 @@ define <vscale x 16 x half> @vfadd_vv_nxv16f16(<vscale x 16 x half> %va, <vscale
 ; ZVFBFA-LABEL: vfadd_vv_nxv16f16:
 ; ZVFBFA:       # %bb.0:
 ; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
-; ZVFBFA-NEXT:    vfadd.vv v8, v8, v12
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v16, v12
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v24, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v16, v24, v16
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v16
 ; ZVFBFA-NEXT:    ret
   %vc = fadd <vscale x 16 x half> %va, %vb
   ret <vscale x 16 x half> %vc
@@ -842,8 +902,15 @@ define <vscale x 16 x half> @vfadd_vf_nxv16f16(<vscale x 16 x half> %va, half %b
 ;
 ; ZVFBFA-LABEL: vfadd_vf_nxv16f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
-; ZVFBFA-NEXT:    vfadd.vf v8, v8, fa0
+; ZVFBFA-NEXT:    fmv.x.w a0, fa0
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vmv.v.x v24, a0
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v16, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v8, v24
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v16, v16, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v16
 ; ZVFBFA-NEXT:    ret
   %head = insertelement <vscale x 16 x half> poison, half %b, i32 0
   %splat = shufflevector <vscale x 16 x half> %head, <vscale x 16 x half> poison, <vscale x 16 x i32> zeroinitializer
@@ -892,8 +959,34 @@ define <vscale x 32 x half> @vfadd_vv_nxv32f16(<vscale x 32 x half> %va, <vscale
 ;
 ; ZVFBFA-LABEL: vfadd_vv_nxv32f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m8, ta, ma
-; ZVFBFA-NEXT:    vfadd.vv v8, v8, v16
+; ZVFBFA-NEXT:    addi sp, sp, -16
+; ZVFBFA-NEXT:    .cfi_def_cfa_offset 16
+; ZVFBFA-NEXT:    csrr a0, vlenb
+; ZVFBFA-NEXT:    slli a0, a0, 3
+; ZVFBFA-NEXT:    sub sp, sp, a0
+; ZVFBFA-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 8 * vlenb
+; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v24, v16
+; ZVFBFA-NEXT:    addi a0, sp, 16
+; ZVFBFA-NEXT:    vs8r.v v24, (a0) # vscale x 64-byte Folded Spill
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v0, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v24, v20
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v16, v12
+; ZVFBFA-NEXT:    vl8r.v v8, (a0) # vscale x 64-byte Folded Reload
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v0, v0, v8
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v0
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v16, v16, v24
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v12, v16
+; ZVFBFA-NEXT:    csrr a0, vlenb
+; ZVFBFA-NEXT:    slli a0, a0, 3
+; ZVFBFA-NEXT:    add sp, sp, a0
+; ZVFBFA-NEXT:    .cfi_def_cfa sp, 16
+; ZVFBFA-NEXT:    addi sp, sp, 16
+; ZVFBFA-NEXT:    .cfi_def_cfa_offset 0
 ; ZVFBFA-NEXT:    ret
   %vc = fadd <vscale x 32 x half> %va, %vb
   ret <vscale x 32 x half> %vc
@@ -945,8 +1038,39 @@ define <vscale x 32 x half> @vfadd_vf_nxv32f16(<vscale x 32 x half> %va, half %b
 ;
 ; ZVFBFA-LABEL: vfadd_vf_nxv32f16:
 ; ZVFBFA:       # %bb.0:
-; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m8, ta, ma
-; ZVFBFA-NEXT:    vfadd.vf v8, v8, fa0
+; ZVFBFA-NEXT:    addi sp, sp, -16
+; ZVFBFA-NEXT:    .cfi_def_cfa_offset 16
+; ZVFBFA-NEXT:    csrr a0, vlenb
+; ZVFBFA-NEXT:    slli a0, a0, 3
+; ZVFBFA-NEXT:    sub sp, sp, a0
+; ZVFBFA-NEXT:    .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x08, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 8 * vlenb
+; ZVFBFA-NEXT:    fmv.x.w a0, fa0
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v16, v8
+; ZVFBFA-NEXT:    addi a1, sp, 16
+; ZVFBFA-NEXT:    vs8r.v v16, (a1) # vscale x 64-byte Folded Spill
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v24, v12
+; ZVFBFA-NEXT:    vsetvli a1, zero, e16, m8, ta, ma
+; ZVFBFA-NEXT:    vmv.v.x v8, a0
+; ZVFBFA-NEXT:    vsetvli a0, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v0, v8
+; ZVFBFA-NEXT:    vfwcvt.f.f.v v16, v12
+; ZVFBFA-NEXT:    addi a0, sp, 16
+; ZVFBFA-NEXT:    vl8r.v v8, (a0) # vscale x 64-byte Folded Reload
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v0, v8, v0
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v8, v0
+; ZVFBFA-NEXT:    vsetvli zero, zero, e32, m8, ta, ma
+; ZVFBFA-NEXT:    vfadd.vv v16, v24, v16
+; ZVFBFA-NEXT:    vsetvli zero, zero, e16, m4, ta, ma
+; ZVFBFA-NEXT:    vfncvt.f.f.w v12, v16
+; ZVFBFA-NEXT:    csrr a0, vlenb
+; ZVFBFA-NEXT:    slli a0, a0, 3
+; ZVFBFA-NEXT:    add sp, sp, a0
+; ZVFBFA-NEXT:    .cfi_def_cfa sp, 16
+; ZVFBFA-NEXT:    addi sp, sp, 16
+; ZVFBFA-NEXT:    .cfi_def_cfa_offset 0
 ; ZVFBFA-NEXT:    ret
   %head = insertelement <vscale x 32 x half> poison, half %b, i32 0
   %splat = shufflevector <vscale x 32 x half> %head, <vscale x 32 x half> poison, <vscale x 32 x i32> zeroinitializer