diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -1170,7 +1170,13 @@
   case CASE_VFMA_OPCODE_LMULS(FMACC, VV):
   case CASE_VFMA_OPCODE_LMULS(FMSAC, VV):
   case CASE_VFMA_OPCODE_LMULS(FNMACC, VV):
-  case CASE_VFMA_OPCODE_LMULS(FNMSAC, VV): {
+  case CASE_VFMA_OPCODE_LMULS(FNMSAC, VV):
+  case CASE_VFMA_OPCODE_LMULS(MADD, VX):
+  case CASE_VFMA_OPCODE_LMULS(NMSUB, VX):
+  case CASE_VFMA_OPCODE_LMULS(MACC, VX):
+  case CASE_VFMA_OPCODE_LMULS(NMSAC, VX):
+  case CASE_VFMA_OPCODE_LMULS(MACC, VV):
+  case CASE_VFMA_OPCODE_LMULS(NMSAC, VV): {
     // For these instructions we can only swap operand 1 and operand 3 by
     // changing the opcode.
     unsigned CommutableOpIdx1 = 1;
@@ -1183,7 +1189,9 @@
   case CASE_VFMA_OPCODE_LMULS(FMADD, VV):
   case CASE_VFMA_OPCODE_LMULS(FMSUB, VV):
   case CASE_VFMA_OPCODE_LMULS(FNMADD, VV):
-  case CASE_VFMA_OPCODE_LMULS(FNMSUB, VV): {
+  case CASE_VFMA_OPCODE_LMULS(FNMSUB, VV):
+  case CASE_VFMA_OPCODE_LMULS(MADD, VV):
+  case CASE_VFMA_OPCODE_LMULS(NMSUB, VV): {
     // For these instructions we have more freedom. We can commute with the
     // other multiplicand or with the addend/subtrahend/minuend.
 
@@ -1288,7 +1296,13 @@
   case CASE_VFMA_OPCODE_LMULS(FMACC, VV):
   case CASE_VFMA_OPCODE_LMULS(FMSAC, VV):
   case CASE_VFMA_OPCODE_LMULS(FNMACC, VV):
-  case CASE_VFMA_OPCODE_LMULS(FNMSAC, VV): {
+  case CASE_VFMA_OPCODE_LMULS(FNMSAC, VV):
+  case CASE_VFMA_OPCODE_LMULS(MADD, VX):
+  case CASE_VFMA_OPCODE_LMULS(NMSUB, VX):
+  case CASE_VFMA_OPCODE_LMULS(MACC, VX):
+  case CASE_VFMA_OPCODE_LMULS(NMSAC, VX):
+  case CASE_VFMA_OPCODE_LMULS(MACC, VV):
+  case CASE_VFMA_OPCODE_LMULS(NMSAC, VV): {
     // It only make sense to toggle these between clobbering the
     // addend/subtrahend/minuend one of the multiplicands.
     assert((OpIdx1 == 1 || OpIdx2 == 1) && "Unexpected opcode index");
@@ -1309,6 +1323,12 @@
       CASE_VFMA_CHANGE_OPCODE_LMULS(FMSAC, FMSUB, VV)
       CASE_VFMA_CHANGE_OPCODE_LMULS(FNMACC, FNMADD, VV)
       CASE_VFMA_CHANGE_OPCODE_LMULS(FNMSAC, FNMSUB, VV)
+      CASE_VFMA_CHANGE_OPCODE_LMULS(MACC, MADD, VX)
+      CASE_VFMA_CHANGE_OPCODE_LMULS(MADD, MACC, VX)
+      CASE_VFMA_CHANGE_OPCODE_LMULS(NMSAC, NMSUB, VX)
+      CASE_VFMA_CHANGE_OPCODE_LMULS(NMSUB, NMSAC, VX)
+      CASE_VFMA_CHANGE_OPCODE_LMULS(MACC, MADD, VV)
+      CASE_VFMA_CHANGE_OPCODE_LMULS(NMSAC, NMSUB, VV)
     }
 
     auto &WorkingMI = cloneIfNew(MI);
@@ -1319,7 +1339,9 @@
   case CASE_VFMA_OPCODE_LMULS(FMADD, VV):
   case CASE_VFMA_OPCODE_LMULS(FMSUB, VV):
   case CASE_VFMA_OPCODE_LMULS(FNMADD, VV):
-  case CASE_VFMA_OPCODE_LMULS(FNMSUB, VV): {
+  case CASE_VFMA_OPCODE_LMULS(FNMSUB, VV):
+  case CASE_VFMA_OPCODE_LMULS(MADD, VV):
+  case CASE_VFMA_OPCODE_LMULS(NMSUB, VV): {
     assert((OpIdx1 == 1 || OpIdx2 == 1) && "Unexpected opcode index");
     // If one of the operands, is the addend we need to change opcode.
     // Otherwise we're just swapping 2 of the multiplicands.
@@ -1332,6 +1354,8 @@
         CASE_VFMA_CHANGE_OPCODE_LMULS(FMSUB, FMSAC, VV)
         CASE_VFMA_CHANGE_OPCODE_LMULS(FNMADD, FNMACC, VV)
         CASE_VFMA_CHANGE_OPCODE_LMULS(FNMSUB, FNMSAC, VV)
+        CASE_VFMA_CHANGE_OPCODE_LMULS(MADD, MACC, VV)
+        CASE_VFMA_CHANGE_OPCODE_LMULS(NMSUB, NMSAC, VV)
       }
 
       auto &WorkingMI = cloneIfNew(MI);
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td
@@ -1936,6 +1936,22 @@
 multiclass VPseudoTernaryV_VV_VX_AAXA<string Constraint = ""> {
   defm "" : VPseudoTernaryV_VV<Constraint>;
   defm "" : VPseudoTernaryV_VX_AAXA<Constraint>;
+
+  foreach m = MxList.m in {
+    // Add a commutable version for use by IR fma.
+    // NOTE: We need this because we use a tail undisturbed policy on the
+    // intrinsic version so we can't commute those instructions since it would
+    // change which input operand is tied to the destination. That would
+    // remove user control of the tail elements.
+    let isCommutable = 1, ForceTailAgnostic = true, VLMul = m.value in {
+      def "_VV_" # m.MX # "_COMMUTABLE" : VPseudoTernaryNoMask<m.vrclass,
+                                                               m.vrclass,
+                                                               m.vrclass,
+                                                               Constraint>;
+      def "_VX_" # m.MX # "_COMMUTABLE" :
+         VPseudoTernaryNoMask<m.vrclass, GPR, m.vrclass, Constraint>;
+    }
+  }
 }
 
 multiclass VPseudoTernaryV_VV_VF_AAXA<string Constraint = ""> {
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVSDPatterns.td
@@ -472,6 +472,36 @@
 defm : VPatBinarySDNode_VV_VX<urem, "PseudoVREMU">;
 defm : VPatBinarySDNode_VV_VX<srem, "PseudoVREM">;
 
+// 12.13 Vector Single-Width Integer Multiply-Add Instructions.
+foreach vti = AllIntegerVectors in {
+  // NOTE: We choose VMADD because it has the most commuting freedom. So it
+  // works best with how TwoAddressInstructionPass tries commuting.
+  defvar suffix = vti.LMul.MX # "_COMMUTABLE";
+  def : Pat<(vti.Vector (add vti.RegClass:$rs2,
+                              (mul vti.RegClass:$rs1, vti.RegClass:$rd))),
+            (!cast<Instruction>("PseudoVMADD_VV_"# suffix)
+                 vti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
+                 vti.AVL, vti.Log2SEW)>;
+  def : Pat<(vti.Vector (sub vti.RegClass:$rs2,
+                              (mul vti.RegClass:$rs1, vti.RegClass:$rd))),
+            (!cast<Instruction>("PseudoVNMSUB_VV_"# suffix)
+                 vti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
+                 vti.AVL, vti.Log2SEW)>;
+
+  // The choice of VMADD here is arbitrary, vfmadd.vf and vfmacc.vf are equally
+  // commutable.
+  def : Pat<(vti.Vector (add vti.RegClass:$rs2,
+                              (mul (SplatPat XLenVT:$rs1), vti.RegClass:$rd))),
+            (!cast<Instruction>("PseudoVMADD_VX_" # suffix)
+                 vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
+                 vti.AVL, vti.Log2SEW)>;
+  def : Pat<(vti.Vector (sub vti.RegClass:$rs2,
+                              (mul (SplatPat XLenVT:$rs1), vti.RegClass:$rd))),
+            (!cast<Instruction>("PseudoVNMSUB_VX_" # suffix)
+                 vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
+                 vti.AVL, vti.Log2SEW)>;
+}
+
 // 12.15. Vector Integer Merge Instructions
 foreach vti = AllIntegerVectors in {
   def : Pat<(vti.Vector (vselect (vti.Mask VMV0:$vm), vti.RegClass:$rs1,
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVVLPatterns.td
@@ -731,6 +731,50 @@
 defm : VPatBinaryVL_VV_VX<riscv_urem_vl, "PseudoVREMU">;
 defm : VPatBinaryVL_VV_VX<riscv_srem_vl, "PseudoVREM">;
 
+// 12.13 Vector Single-Width Integer Multiply-Add Instructions
+foreach vti = AllIntegerVectors in {
+  // NOTE: We choose VMADD because it has the most commuting freedom. So it
+  // works best with how TwoAddressInstructionPass tries commuting.
+  defvar suffix = vti.LMul.MX # "_COMMUTABLE";
+  def : Pat<(vti.Vector
+             (riscv_add_vl vti.RegClass:$rs2,
+                           (riscv_mul_vl vti.RegClass:$rs1, vti.RegClass:$rd,
+                                         (vti.Mask true_mask), VLOpFrag),
+                           (vti.Mask true_mask), VLOpFrag)),
+            (!cast<Instruction>("PseudoVMADD_VV_"# suffix)
+                 vti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
+                 GPR:$vl, vti.Log2SEW)>;
+  def : Pat<(vti.Vector
+             (riscv_sub_vl vti.RegClass:$rs2,
+                           (riscv_mul_vl vti.RegClass:$rs1, vti.RegClass:$rd,
+                                         (vti.Mask true_mask), VLOpFrag),
+                           (vti.Mask true_mask), VLOpFrag)),
+            (!cast<Instruction>("PseudoVNMSUB_VV_"# suffix)
+                 vti.RegClass:$rd, vti.RegClass:$rs1, vti.RegClass:$rs2,
+                 GPR:$vl, vti.Log2SEW)>;
+
+  // The choice of VMADD here is arbitrary, vmadd.vx and vmacc.vx are equally
+  // commutable.
+  def : Pat<(vti.Vector
+             (riscv_add_vl vti.RegClass:$rs2,
+                           (riscv_mul_vl (SplatPat XLenVT:$rs1),
+                                         vti.RegClass:$rd, (vti.Mask true_mask),
+                                         VLOpFrag),
+                           (vti.Mask true_mask), VLOpFrag)),
+            (!cast<Instruction>("PseudoVMADD_VX_" # suffix)
+                 vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
+                 GPR:$vl, vti.Log2SEW)>;
+  def : Pat<(vti.Vector
+             (riscv_sub_vl vti.RegClass:$rs2,
+                           (riscv_mul_vl (SplatPat XLenVT:$rs1),
+                                         vti.RegClass:$rd, (vti.Mask true_mask),
+                                         VLOpFrag),
+                           (vti.Mask true_mask), VLOpFrag)),
+            (!cast<Instruction>("PseudoVNMSUB_VX_" # suffix)
+                 vti.RegClass:$rd, vti.ScalarRegClass:$rs1, vti.RegClass:$rs2,
+                 GPR:$vl, vti.Log2SEW)>;
+}
+
 // 12.15. Vector Integer Merge Instructions
 foreach vti = AllIntegerVectors in {
   def : Pat<(vti.Vector (riscv_vselect_vl (vti.Mask VMV0:$vm),
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int.ll
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-int.ll
@@ -1233,19 +1233,18 @@
 ; RV32-NEXT:    vmv.v.i v27, -1
 ; RV32-NEXT:    vmerge.vim v27, v27, 0, v0
 ; RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, mu
-; RV32-NEXT:    vmul.vv v25, v25, v27
-; RV32-NEXT:    vadd.vv v25, v26, v25
+; RV32-NEXT:    vmadd.vv v27, v25, v26
 ; RV32-NEXT:    addi a1, zero, 63
-; RV32-NEXT:    vsrl.vx v26, v25, a1
+; RV32-NEXT:    vsrl.vx v25, v27, a1
 ; RV32-NEXT:    addi a1, zero, 1
 ; RV32-NEXT:    vsetivli zero, 4, e32, m1, ta, mu
-; RV32-NEXT:    vmv.s.x v27, a1
+; RV32-NEXT:    vmv.s.x v26, a1
 ; RV32-NEXT:    vmv.v.i v28, 0
 ; RV32-NEXT:    vsetivli zero, 3, e32, m1, tu, mu
-; RV32-NEXT:    vslideup.vi v28, v27, 2
+; RV32-NEXT:    vslideup.vi v28, v26, 2
 ; RV32-NEXT:    vsetivli zero, 2, e64, m1, ta, mu
-; RV32-NEXT:    vsra.vv v25, v25, v28
-; RV32-NEXT:    vadd.vv v25, v25, v26
+; RV32-NEXT:    vsra.vv v26, v27, v28
+; RV32-NEXT:    vadd.vv v25, v26, v25
 ; RV32-NEXT:    vse64.v v25, (a0)
 ; RV32-NEXT:    ret
 ;
@@ -1256,8 +1255,6 @@
 ; RV64-NEXT:    vmv.v.i v26, -1
 ; RV64-NEXT:    vsetvli zero, zero, e64, m1, tu, mu
 ; RV64-NEXT:    vmv.s.x v26, zero
-; RV64-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
-; RV64-NEXT:    vmul.vv v26, v25, v26
 ; RV64-NEXT:    lui a1, 21845
 ; RV64-NEXT:    addiw a1, a1, 1365
 ; RV64-NEXT:    slli a1, a1, 12
@@ -1266,18 +1263,19 @@
 ; RV64-NEXT:    addi a1, a1, 1365
 ; RV64-NEXT:    slli a1, a1, 12
 ; RV64-NEXT:    addi a2, a1, 1365
+; RV64-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
 ; RV64-NEXT:    vmv.v.x v27, a2
 ; RV64-NEXT:    addi a1, a1, 1366
 ; RV64-NEXT:    vsetvli zero, zero, e64, m1, tu, mu
 ; RV64-NEXT:    vmv.s.x v27, a1
 ; RV64-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
-; RV64-NEXT:    vmulh.vv v25, v25, v27
-; RV64-NEXT:    vadd.vv v25, v25, v26
+; RV64-NEXT:    vmulh.vv v27, v25, v27
+; RV64-NEXT:    vmacc.vv v27, v25, v26
 ; RV64-NEXT:    addi a1, zero, 63
-; RV64-NEXT:    vsrl.vx v26, v25, a1
-; RV64-NEXT:    vid.v v27
-; RV64-NEXT:    vsra.vv v25, v25, v27
-; RV64-NEXT:    vadd.vv v25, v25, v26
+; RV64-NEXT:    vsrl.vx v25, v27, a1
+; RV64-NEXT:    vid.v v26
+; RV64-NEXT:    vsra.vv v26, v27, v26
+; RV64-NEXT:    vadd.vv v25, v26, v25
 ; RV64-NEXT:    vse64.v v25, (a0)
 ; RV64-NEXT:    ret
   %a = load <2 x i64>, <2 x i64>* %x
@@ -4709,37 +4707,36 @@
 ; LMULMAX2-RV32:       # %bb.0:
 ; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
 ; LMULMAX2-RV32-NEXT:    vle64.v v26, (a0)
-; LMULMAX2-RV32-NEXT:    addi a1, zero, 51
-; LMULMAX2-RV32-NEXT:    vsetivli zero, 1, e8, mf8, ta, mu
-; LMULMAX2-RV32-NEXT:    vmv.s.x v0, a1
-; LMULMAX2-RV32-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
-; LMULMAX2-RV32-NEXT:    vmv.v.i v28, -1
-; LMULMAX2-RV32-NEXT:    vmerge.vim v28, v28, 0, v0
-; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
-; LMULMAX2-RV32-NEXT:    vmul.vv v28, v26, v28
 ; LMULMAX2-RV32-NEXT:    addi a1, zero, 17
 ; LMULMAX2-RV32-NEXT:    vsetivli zero, 1, e8, mf8, ta, mu
 ; LMULMAX2-RV32-NEXT:    vmv.s.x v0, a1
 ; LMULMAX2-RV32-NEXT:    lui a1, 349525
 ; LMULMAX2-RV32-NEXT:    addi a2, a1, 1365
 ; LMULMAX2-RV32-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
-; LMULMAX2-RV32-NEXT:    vmv.v.x v30, a2
+; LMULMAX2-RV32-NEXT:    vmv.v.x v28, a2
 ; LMULMAX2-RV32-NEXT:    addi a1, a1, 1366
-; LMULMAX2-RV32-NEXT:    vmerge.vxm v30, v30, a1, v0
+; LMULMAX2-RV32-NEXT:    vmerge.vxm v28, v28, a1, v0
 ; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
-; LMULMAX2-RV32-NEXT:    vmulh.vv v26, v26, v30
-; LMULMAX2-RV32-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-RV32-NEXT:    vmulh.vv v28, v26, v28
+; LMULMAX2-RV32-NEXT:    addi a1, zero, 51
+; LMULMAX2-RV32-NEXT:    vsetivli zero, 1, e8, mf8, ta, mu
+; LMULMAX2-RV32-NEXT:    vmv.s.x v0, a1
+; LMULMAX2-RV32-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
+; LMULMAX2-RV32-NEXT:    vmv.v.i v30, -1
+; LMULMAX2-RV32-NEXT:    vmerge.vim v30, v30, 0, v0
+; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
+; LMULMAX2-RV32-NEXT:    vmadd.vv v30, v26, v28
 ; LMULMAX2-RV32-NEXT:    addi a1, zero, 63
-; LMULMAX2-RV32-NEXT:    vsrl.vx v28, v26, a1
+; LMULMAX2-RV32-NEXT:    vsrl.vx v26, v30, a1
 ; LMULMAX2-RV32-NEXT:    addi a1, zero, 68
 ; LMULMAX2-RV32-NEXT:    vsetivli zero, 1, e8, mf8, ta, mu
 ; LMULMAX2-RV32-NEXT:    vmv.s.x v0, a1
 ; LMULMAX2-RV32-NEXT:    vsetivli zero, 8, e32, m2, ta, mu
-; LMULMAX2-RV32-NEXT:    vmv.v.i v30, 0
-; LMULMAX2-RV32-NEXT:    vmerge.vim v30, v30, 1, v0
+; LMULMAX2-RV32-NEXT:    vmv.v.i v28, 0
+; LMULMAX2-RV32-NEXT:    vmerge.vim v28, v28, 1, v0
 ; LMULMAX2-RV32-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
-; LMULMAX2-RV32-NEXT:    vsra.vv v26, v26, v30
-; LMULMAX2-RV32-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-RV32-NEXT:    vsra.vv v28, v30, v28
+; LMULMAX2-RV32-NEXT:    vadd.vv v26, v28, v26
 ; LMULMAX2-RV32-NEXT:    vse64.v v26, (a0)
 ; LMULMAX2-RV32-NEXT:    ret
 ;
@@ -4753,7 +4750,6 @@
 ; LMULMAX2-RV64-NEXT:    vsetivli zero, 4, e64, m2, ta, mu
 ; LMULMAX2-RV64-NEXT:    vmv.v.i v28, -1
 ; LMULMAX2-RV64-NEXT:    vmerge.vim v28, v28, 0, v0
-; LMULMAX2-RV64-NEXT:    vmul.vv v28, v26, v28
 ; LMULMAX2-RV64-NEXT:    lui a1, 21845
 ; LMULMAX2-RV64-NEXT:    addiw a1, a1, 1365
 ; LMULMAX2-RV64-NEXT:    slli a1, a1, 12
@@ -4765,14 +4761,14 @@
 ; LMULMAX2-RV64-NEXT:    vmv.v.x v30, a2
 ; LMULMAX2-RV64-NEXT:    addi a1, a1, 1366
 ; LMULMAX2-RV64-NEXT:    vmerge.vxm v30, v30, a1, v0
-; LMULMAX2-RV64-NEXT:    vmulh.vv v26, v26, v30
-; LMULMAX2-RV64-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-RV64-NEXT:    vmulh.vv v30, v26, v30
+; LMULMAX2-RV64-NEXT:    vmacc.vv v30, v26, v28
 ; LMULMAX2-RV64-NEXT:    addi a1, zero, 63
-; LMULMAX2-RV64-NEXT:    vsrl.vx v28, v26, a1
-; LMULMAX2-RV64-NEXT:    vmv.v.i v30, 1
-; LMULMAX2-RV64-NEXT:    vmerge.vim v30, v30, 0, v0
-; LMULMAX2-RV64-NEXT:    vsra.vv v26, v26, v30
-; LMULMAX2-RV64-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX2-RV64-NEXT:    vsrl.vx v26, v30, a1
+; LMULMAX2-RV64-NEXT:    vmv.v.i v28, 1
+; LMULMAX2-RV64-NEXT:    vmerge.vim v28, v28, 0, v0
+; LMULMAX2-RV64-NEXT:    vsra.vv v28, v30, v28
+; LMULMAX2-RV64-NEXT:    vadd.vv v26, v28, v26
 ; LMULMAX2-RV64-NEXT:    vse64.v v26, (a0)
 ; LMULMAX2-RV64-NEXT:    ret
 ;
@@ -4802,8 +4798,6 @@
 ; LMULMAX1-RV64-NEXT:    vmv.v.i v27, -1
 ; LMULMAX1-RV64-NEXT:    vsetvli zero, zero, e64, m1, tu, mu
 ; LMULMAX1-RV64-NEXT:    vmv.s.x v27, zero
-; LMULMAX1-RV64-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
-; LMULMAX1-RV64-NEXT:    vmul.vv v28, v26, v27
 ; LMULMAX1-RV64-NEXT:    lui a2, 21845
 ; LMULMAX1-RV64-NEXT:    addiw a2, a2, 1365
 ; LMULMAX1-RV64-NEXT:    slli a2, a2, 12
@@ -4812,24 +4806,24 @@
 ; LMULMAX1-RV64-NEXT:    addi a2, a2, 1365
 ; LMULMAX1-RV64-NEXT:    slli a2, a2, 12
 ; LMULMAX1-RV64-NEXT:    addi a3, a2, 1365
-; LMULMAX1-RV64-NEXT:    vmv.v.x v29, a3
+; LMULMAX1-RV64-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
+; LMULMAX1-RV64-NEXT:    vmv.v.x v28, a3
 ; LMULMAX1-RV64-NEXT:    addi a2, a2, 1366
 ; LMULMAX1-RV64-NEXT:    vsetvli zero, zero, e64, m1, tu, mu
-; LMULMAX1-RV64-NEXT:    vmv.s.x v29, a2
+; LMULMAX1-RV64-NEXT:    vmv.s.x v28, a2
 ; LMULMAX1-RV64-NEXT:    vsetvli zero, zero, e64, m1, ta, mu
-; LMULMAX1-RV64-NEXT:    vmulh.vv v26, v26, v29
-; LMULMAX1-RV64-NEXT:    vadd.vv v26, v26, v28
+; LMULMAX1-RV64-NEXT:    vmulh.vv v29, v26, v28
+; LMULMAX1-RV64-NEXT:    vmacc.vv v29, v27, v26
 ; LMULMAX1-RV64-NEXT:    addi a2, zero, 63
-; LMULMAX1-RV64-NEXT:    vsrl.vx v28, v26, a2
+; LMULMAX1-RV64-NEXT:    vsrl.vx v26, v29, a2
 ; LMULMAX1-RV64-NEXT:    vid.v v30
-; LMULMAX1-RV64-NEXT:    vsra.vv v26, v26, v30
-; LMULMAX1-RV64-NEXT:    vadd.vv v26, v26, v28
-; LMULMAX1-RV64-NEXT:    vmul.vv v27, v25, v27
-; LMULMAX1-RV64-NEXT:    vmulh.vv v25, v25, v29
-; LMULMAX1-RV64-NEXT:    vadd.vv v25, v25, v27
-; LMULMAX1-RV64-NEXT:    vsrl.vx v27, v25, a2
-; LMULMAX1-RV64-NEXT:    vsra.vv v25, v25, v30
-; LMULMAX1-RV64-NEXT:    vadd.vv v25, v25, v27
+; LMULMAX1-RV64-NEXT:    vsra.vv v29, v29, v30
+; LMULMAX1-RV64-NEXT:    vadd.vv v26, v29, v26
+; LMULMAX1-RV64-NEXT:    vmulh.vv v28, v25, v28
+; LMULMAX1-RV64-NEXT:    vmacc.vv v28, v25, v27
+; LMULMAX1-RV64-NEXT:    vsrl.vx v25, v28, a2
+; LMULMAX1-RV64-NEXT:    vsra.vv v27, v28, v30
+; LMULMAX1-RV64-NEXT:    vadd.vv v25, v27, v25
 ; LMULMAX1-RV64-NEXT:    vse64.v v25, (a0)
 ; LMULMAX1-RV64-NEXT:    vse64.v v26, (a1)
 ; LMULMAX1-RV64-NEXT:    ret
diff --git a/llvm/test/CodeGen/RISCV/rvv/stepvector.ll b/llvm/test/CodeGen/RISCV/rvv/stepvector.ll
--- a/llvm/test/CodeGen/RISCV/rvv/stepvector.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/stepvector.ll
@@ -480,15 +480,16 @@
 define <vscale x 16 x i64> @mul_stepvector_nxv16i64() {
 ; CHECK-LABEL: mul_stepvector_nxv16i64:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
-; CHECK-NEXT:    vid.v v8
-; CHECK-NEXT:    addi a0, zero, 3
-; CHECK-NEXT:    vmul.vx v8, v8, a0
 ; CHECK-NEXT:    csrr a0, vlenb
 ; CHECK-NEXT:    srli a0, a0, 3
 ; CHECK-NEXT:    addi a1, zero, 24
 ; CHECK-NEXT:    mul a0, a0, a1
-; CHECK-NEXT:    vadd.vx v16, v8, a0
+; CHECK-NEXT:    vsetvli a1, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vmv.v.x v16, a0
+; CHECK-NEXT:    vid.v v8
+; CHECK-NEXT:    addi a0, zero, 3
+; CHECK-NEXT:    vmacc.vx v16, a0, v8
+; CHECK-NEXT:    vmul.vx v8, v8, a0
 ; CHECK-NEXT:    ret
 entry:
   %0 = insertelement <vscale x 16 x i64> poison, i64 3, i32 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/vmadd-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/vmadd-sdnode.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vmadd-sdnode.ll
@@ -0,0 +1,592 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v -target-abi=ilp32 \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v -target-abi=lp64 \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64
+
+; This tests a mix of vmacc and vmadd by using different operand orders to
+; trigger commuting in TwoAddressInstructionPass.
+
+define <vscale x 1 x i8> @vmadd_vv_nxv1i8(<vscale x 1 x i8> %va, <vscale x 1 x i8> %vb, <vscale x 1 x i8> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv1i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf8, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i8> %va, %vb
+  %y = add <vscale x 1 x i8> %x, %vc
+  ret <vscale x 1 x i8> %y
+}
+
+define <vscale x 1 x i8> @vmadd_vx_nxv1i8(<vscale x 1 x i8> %va, <vscale x 1 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vmadd_vx_nxv1i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf8, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 1 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i8> %head, <vscale x 1 x i8> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i8> %va, %splat
+  %y = add <vscale x 1 x i8> %x, %vb
+  ret <vscale x 1 x i8> %y
+}
+
+define <vscale x 2 x i8> @vmadd_vv_nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i8> %vb, <vscale x 2 x i8> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i8> %va, %vc
+  %y = add <vscale x 2 x i8> %x, %vb
+  ret <vscale x 2 x i8> %y
+}
+
+define <vscale x 2 x i8> @vmadd_vx_nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vmadd_vx_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 2 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i8> %head, <vscale x 2 x i8> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i8> %vb, %splat
+  %y = add <vscale x 2 x i8> %x, %va
+  ret <vscale x 2 x i8> %y
+}
+
+define <vscale x 4 x i8> @vmadd_vv_nxv4i8(<vscale x 4 x i8> %va, <vscale x 4 x i8> %vb, <vscale x 4 x i8> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf2, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i8> %vb, %va
+  %y = add <vscale x 4 x i8> %x, %vc
+  ret <vscale x 4 x i8> %y
+}
+
+define <vscale x 4 x i8> @vmadd_vx_nxv4i8(<vscale x 4 x i8> %va, <vscale x 4 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vmadd_vx_nxv4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf2, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 4 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i8> %head, <vscale x 4 x i8> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i8> %va, %splat
+  %y = add <vscale x 4 x i8> %x, %vb
+  ret <vscale x 4 x i8> %y
+}
+
+define <vscale x 8 x i8> @vmadd_vv_nxv8i8(<vscale x 8 x i8> %va, <vscale x 8 x i8> %vb, <vscale x 8 x i8> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv8i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, m1, ta, mu
+; CHECK-NEXT:    vmacc.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i8> %vb, %vc
+  %y = add <vscale x 8 x i8> %x, %va
+  ret <vscale x 8 x i8> %y
+}
+
+define <vscale x 8 x i8> @vmadd_vx_nxv8i8(<vscale x 8 x i8> %va, <vscale x 8 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vmadd_vx_nxv8i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m1, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 8 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i8> %head, <vscale x 8 x i8> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i8> %vb, %splat
+  %y = add <vscale x 8 x i8> %x, %va
+  ret <vscale x 8 x i8> %y
+}
+
+define <vscale x 16 x i8> @vmadd_vv_nxv16i8(<vscale x 16 x i8> %va, <vscale x 16 x i8> %vb, <vscale x 16 x i8> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, m2, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v12, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 16 x i8> %vc, %va
+  %y = add <vscale x 16 x i8> %x, %vb
+  ret <vscale x 16 x i8> %y
+}
+
+define <vscale x 16 x i8> @vmadd_vx_nxv16i8(<vscale x 16 x i8> %va, <vscale x 16 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vmadd_vx_nxv16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m2, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v10
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 16 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %head, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %x = mul <vscale x 16 x i8> %va, %splat
+  %y = add <vscale x 16 x i8> %x, %vb
+  ret <vscale x 16 x i8> %y
+}
+
+define <vscale x 32 x i8> @vmadd_vv_nxv32i8(<vscale x 32 x i8> %va, <vscale x 32 x i8> %vb, <vscale x 32 x i8> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv32i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, m4, ta, mu
+; CHECK-NEXT:    vmacc.vv v8, v16, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 32 x i8> %vc, %vb
+  %y = add <vscale x 32 x i8> %x, %va
+  ret <vscale x 32 x i8> %y
+}
+
+define <vscale x 32 x i8> @vmadd_vx_nxv32i8(<vscale x 32 x i8> %va, <vscale x 32 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vmadd_vx_nxv32i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m4, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v12
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 32 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 32 x i8> %head, <vscale x 32 x i8> undef, <vscale x 32 x i32> zeroinitializer
+  %x = mul <vscale x 32 x i8> %vb, %splat
+  %y = add <vscale x 32 x i8> %x, %va
+  ret <vscale x 32 x i8> %y
+}
+
+define <vscale x 64 x i8> @vmadd_vv_nxv64i8(<vscale x 64 x i8> %va, <vscale x 64 x i8> %vb, <vscale x 64 x i8> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv64i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8r.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e8, m8, ta, mu
+; CHECK-NEXT:    vmacc.vv v8, v16, v24
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 64 x i8> %vc, %vb
+  %y = add <vscale x 64 x i8> %x, %va
+  ret <vscale x 64 x i8> %y
+}
+
+define <vscale x 64 x i8> @vmadd_vx_nxv64i8(<vscale x 64 x i8> %va, <vscale x 64 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vmadd_vx_nxv64i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m8, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v16
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 64 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 64 x i8> %head, <vscale x 64 x i8> undef, <vscale x 64 x i32> zeroinitializer
+  %x = mul <vscale x 64 x i8> %vb, %splat
+  %y = add <vscale x 64 x i8> %x, %va
+  ret <vscale x 64 x i8> %y
+}
+
+define <vscale x 1 x i16> @vmadd_vv_nxv1i16(<vscale x 1 x i16> %va, <vscale x 1 x i16> %vb, <vscale x 1 x i16> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv1i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i16> %va, %vb
+  %y = add <vscale x 1 x i16> %x, %vc
+  ret <vscale x 1 x i16> %y
+}
+
+define <vscale x 1 x i16> @vmadd_vx_nxv1i16(<vscale x 1 x i16> %va, <vscale x 1 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vmadd_vx_nxv1i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 1 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i16> %head, <vscale x 1 x i16> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i16> %va, %splat
+  %y = add <vscale x 1 x i16> %x, %vb
+  ret <vscale x 1 x i16> %y
+}
+
+define <vscale x 2 x i16> @vmadd_vv_nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i16> %vb, <vscale x 2 x i16> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i16> %va, %vc
+  %y = add <vscale x 2 x i16> %x, %vb
+  ret <vscale x 2 x i16> %y
+}
+
+define <vscale x 2 x i16> @vmadd_vx_nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vmadd_vx_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 2 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i16> %head, <vscale x 2 x i16> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i16> %vb, %splat
+  %y = add <vscale x 2 x i16> %x, %va
+  ret <vscale x 2 x i16> %y
+}
+
+define <vscale x 4 x i16> @vmadd_vv_nxv4i16(<vscale x 4 x i16> %va, <vscale x 4 x i16> %vb, <vscale x 4 x i16> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i16> %vb, %va
+  %y = add <vscale x 4 x i16> %x, %vc
+  ret <vscale x 4 x i16> %y
+}
+
+define <vscale x 4 x i16> @vmadd_vx_nxv4i16(<vscale x 4 x i16> %va, <vscale x 4 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vmadd_vx_nxv4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 4 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i16> %head, <vscale x 4 x i16> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i16> %va, %splat
+  %y = add <vscale x 4 x i16> %x, %vb
+  ret <vscale x 4 x i16> %y
+}
+
+define <vscale x 8 x i16> @vmadd_vv_nxv8i16(<vscale x 8 x i16> %va, <vscale x 8 x i16> %vb, <vscale x 8 x i16> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vmacc.vv v8, v12, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i16> %vb, %vc
+  %y = add <vscale x 8 x i16> %x, %va
+  ret <vscale x 8 x i16> %y
+}
+
+define <vscale x 8 x i16> @vmadd_vx_nxv8i16(<vscale x 8 x i16> %va, <vscale x 8 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vmadd_vx_nxv8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v10
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 8 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %head, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i16> %vb, %splat
+  %y = add <vscale x 8 x i16> %x, %va
+  ret <vscale x 8 x i16> %y
+}
+
+define <vscale x 16 x i16> @vmadd_vv_nxv16i16(<vscale x 16 x i16> %va, <vscale x 16 x i16> %vb, <vscale x 16 x i16> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v16, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 16 x i16> %vc, %va
+  %y = add <vscale x 16 x i16> %x, %vb
+  ret <vscale x 16 x i16> %y
+}
+
+define <vscale x 16 x i16> @vmadd_vx_nxv16i16(<vscale x 16 x i16> %va, <vscale x 16 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vmadd_vx_nxv16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v12
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 16 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 16 x i16> %head, <vscale x 16 x i16> undef, <vscale x 16 x i32> zeroinitializer
+  %x = mul <vscale x 16 x i16> %va, %splat
+  %y = add <vscale x 16 x i16> %x, %vb
+  ret <vscale x 16 x i16> %y
+}
+
+define <vscale x 32 x i16> @vmadd_vv_nxv32i16(<vscale x 32 x i16> %va, <vscale x 32 x i16> %vb, <vscale x 32 x i16> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv32i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8re16.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vmacc.vv v8, v16, v24
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 32 x i16> %vc, %vb
+  %y = add <vscale x 32 x i16> %x, %va
+  ret <vscale x 32 x i16> %y
+}
+
+define <vscale x 32 x i16> @vmadd_vx_nxv32i16(<vscale x 32 x i16> %va, <vscale x 32 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vmadd_vx_nxv32i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v16
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 32 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 32 x i16> %head, <vscale x 32 x i16> undef, <vscale x 32 x i32> zeroinitializer
+  %x = mul <vscale x 32 x i16> %vb, %splat
+  %y = add <vscale x 32 x i16> %x, %va
+  ret <vscale x 32 x i16> %y
+}
+
+define <vscale x 1 x i32> @vmadd_vv_nxv1i32(<vscale x 1 x i32> %va, <vscale x 1 x i32> %vb, <vscale x 1 x i32> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv1i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i32> %va, %vb
+  %y = add <vscale x 1 x i32> %x, %vc
+  ret <vscale x 1 x i32> %y
+}
+
+define <vscale x 1 x i32> @vmadd_vx_nxv1i32(<vscale x 1 x i32> %va, <vscale x 1 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vmadd_vx_nxv1i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 1 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i32> %head, <vscale x 1 x i32> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i32> %va, %splat
+  %y = add <vscale x 1 x i32> %x, %vb
+  ret <vscale x 1 x i32> %y
+}
+
+define <vscale x 2 x i32> @vmadd_vv_nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i32> %vb, <vscale x 2 x i32> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i32> %va, %vc
+  %y = add <vscale x 2 x i32> %x, %vb
+  ret <vscale x 2 x i32> %y
+}
+
+define <vscale x 2 x i32> @vmadd_vx_nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vmadd_vx_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 2 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i32> %head, <vscale x 2 x i32> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i32> %vb, %splat
+  %y = add <vscale x 2 x i32> %x, %va
+  ret <vscale x 2 x i32> %y
+}
+
+define <vscale x 4 x i32> @vmadd_vv_nxv4i32(<vscale x 4 x i32> %va, <vscale x 4 x i32> %vb, <vscale x 4 x i32> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v10, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i32> %vb, %va
+  %y = add <vscale x 4 x i32> %x, %vc
+  ret <vscale x 4 x i32> %y
+}
+
+define <vscale x 4 x i32> @vmadd_vx_nxv4i32(<vscale x 4 x i32> %va, <vscale x 4 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vmadd_vx_nxv4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v10
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 4 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %head, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i32> %va, %splat
+  %y = add <vscale x 4 x i32> %x, %vb
+  ret <vscale x 4 x i32> %y
+}
+
+define <vscale x 8 x i32> @vmadd_vv_nxv8i32(<vscale x 8 x i32> %va, <vscale x 8 x i32> %vb, <vscale x 8 x i32> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vmacc.vv v8, v16, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i32> %vb, %vc
+  %y = add <vscale x 8 x i32> %x, %va
+  ret <vscale x 8 x i32> %y
+}
+
+define <vscale x 8 x i32> @vmadd_vx_nxv8i32(<vscale x 8 x i32> %va, <vscale x 8 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vmadd_vx_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vmacc.vx v8, a0, v12
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 8 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i32> %head, <vscale x 8 x i32> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i32> %vb, %splat
+  %y = add <vscale x 8 x i32> %x, %va
+  ret <vscale x 8 x i32> %y
+}
+
+define <vscale x 16 x i32> @vmadd_vv_nxv16i32(<vscale x 16 x i32> %va, <vscale x 16 x i32> %vb, <vscale x 16 x i32> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv16i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8re32.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v24, v16
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 16 x i32> %vc, %va
+  %y = add <vscale x 16 x i32> %x, %vb
+  ret <vscale x 16 x i32> %y
+}
+
+define <vscale x 16 x i32> @vmadd_vx_nxv16i32(<vscale x 16 x i32> %va, <vscale x 16 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vmadd_vx_nxv16i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vmadd.vx v8, a0, v16
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 16 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 16 x i32> %head, <vscale x 16 x i32> undef, <vscale x 16 x i32> zeroinitializer
+  %x = mul <vscale x 16 x i32> %va, %splat
+  %y = add <vscale x 16 x i32> %x, %vb
+  ret <vscale x 16 x i32> %y
+}
+
+define <vscale x 1 x i64> @vmadd_vv_nxv1i64(<vscale x 1 x i64> %va, <vscale x 1 x i64> %vb, <vscale x 1 x i64> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv1i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i64> %va, %vb
+  %y = add <vscale x 1 x i64> %x, %vc
+  ret <vscale x 1 x i64> %y
+}
+
+define <vscale x 1 x i64> @vmadd_vx_nxv1i64(<vscale x 1 x i64> %va, <vscale x 1 x i64> %vb, i64 %c) {
+; RV32-LABEL: vmadd_vx_nxv1i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v25, (a0), zero
+; RV32-NEXT:    vmadd.vv v8, v25, v9
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vmadd_vx_nxv1i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m1, ta, mu
+; RV64-NEXT:    vmadd.vx v8, a0, v9
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 1 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i64> %head, <vscale x 1 x i64> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i64> %va, %splat
+  %y = add <vscale x 1 x i64> %x, %vb
+  ret <vscale x 1 x i64> %y
+}
+
+define <vscale x 2 x i64> @vmadd_vv_nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i64> %vb, <vscale x 2 x i64> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v12, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i64> %va, %vc
+  %y = add <vscale x 2 x i64> %x, %vb
+  ret <vscale x 2 x i64> %y
+}
+
+define <vscale x 2 x i64> @vmadd_vx_nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i64> %vb, i64 %c) {
+; RV32-LABEL: vmadd_vx_nxv2i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v26, (a0), zero
+; RV32-NEXT:    vmacc.vv v8, v10, v26
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vmadd_vx_nxv2i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m2, ta, mu
+; RV64-NEXT:    vmacc.vx v8, a0, v10
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 2 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %head, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i64> %vb, %splat
+  %y = add <vscale x 2 x i64> %x, %va
+  ret <vscale x 2 x i64> %y
+}
+
+define <vscale x 4 x i64> @vmadd_vv_nxv4i64(<vscale x 4 x i64> %va, <vscale x 4 x i64> %vb, <vscale x 4 x i64> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vmadd.vv v8, v12, v16
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i64> %vb, %va
+  %y = add <vscale x 4 x i64> %x, %vc
+  ret <vscale x 4 x i64> %y
+}
+
+define <vscale x 4 x i64> @vmadd_vx_nxv4i64(<vscale x 4 x i64> %va, <vscale x 4 x i64> %vb, i64 %c) {
+; RV32-LABEL: vmadd_vx_nxv4i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v28, (a0), zero
+; RV32-NEXT:    vmadd.vv v8, v28, v12
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vmadd_vx_nxv4i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m4, ta, mu
+; RV64-NEXT:    vmadd.vx v8, a0, v12
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 4 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i64> %head, <vscale x 4 x i64> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i64> %va, %splat
+  %y = add <vscale x 4 x i64> %x, %vb
+  ret <vscale x 4 x i64> %y
+}
+
+define <vscale x 8 x i64> @vmadd_vv_nxv8i64(<vscale x 8 x i64> %va, <vscale x 8 x i64> %vb, <vscale x 8 x i64> %vc) {
+; CHECK-LABEL: vmadd_vv_nxv8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8re64.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vmacc.vv v8, v16, v24
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i64> %vb, %vc
+  %y = add <vscale x 8 x i64> %x, %va
+  ret <vscale x 8 x i64> %y
+}
+
+define <vscale x 8 x i64> @vmadd_vx_nxv8i64(<vscale x 8 x i64> %va, <vscale x 8 x i64> %vb, i64 %c) {
+; RV32-LABEL: vmadd_vx_nxv8i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v24, (a0), zero
+; RV32-NEXT:    vmacc.vv v8, v16, v24
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vmadd_vx_nxv8i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, mu
+; RV64-NEXT:    vmacc.vx v8, a0, v16
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 8 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i64> %head, <vscale x 8 x i64> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i64> %vb, %splat
+  %y = add <vscale x 8 x i64> %x, %va
+  ret <vscale x 8 x i64> %y
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vmulh-sdnode-rv32.ll b/llvm/test/CodeGen/RISCV/rvv/vmulh-sdnode-rv32.ll
--- a/llvm/test/CodeGen/RISCV/rvv/vmulh-sdnode-rv32.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vmulh-sdnode-rv32.ll
@@ -6,11 +6,11 @@
 define <vscale x 4 x i1> @srem_eq_fold_nxv4i8(<vscale x 4 x i8> %va) {
 ; CHECK-LABEL: srem_eq_fold_nxv4i8:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    addi a0, zero, -85
-; CHECK-NEXT:    vsetvli a1, zero, e8, mf2, ta, mu
-; CHECK-NEXT:    vmul.vx v25, v8, a0
 ; CHECK-NEXT:    addi a0, zero, 42
-; CHECK-NEXT:    vadd.vx v25, v25, a0
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf2, ta, mu
+; CHECK-NEXT:    vmv.v.x v25, a0
+; CHECK-NEXT:    addi a1, zero, -85
+; CHECK-NEXT:    vmacc.vx v25, a1, v8
 ; CHECK-NEXT:    vsll.vi v26, v25, 7
 ; CHECK-NEXT:    vsrl.vi v25, v25, 1
 ; CHECK-NEXT:    vor.vv v25, v25, v26
diff --git a/llvm/test/CodeGen/RISCV/rvv/vnmsub-sdnode.ll b/llvm/test/CodeGen/RISCV/rvv/vnmsub-sdnode.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/rvv/vnmsub-sdnode.ll
@@ -0,0 +1,592 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=riscv32 -mattr=+experimental-v -target-abi=ilp32 \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV32
+; RUN: llc -mtriple=riscv64 -mattr=+experimental-v -target-abi=lp64 \
+; RUN:     -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV64
+
+; This tests a mix of vmacc and vmsub by using different operand orders to
+; trigger commuting in TwosubressInstructionPass.
+
+define <vscale x 1 x i8> @vnmsub_vv_nxv1i8(<vscale x 1 x i8> %va, <vscale x 1 x i8> %vb, <vscale x 1 x i8> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv1i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf8, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i8> %va, %vb
+  %y = sub <vscale x 1 x i8> %vc, %x
+  ret <vscale x 1 x i8> %y
+}
+
+define <vscale x 1 x i8> @vnmsub_vx_nxv1i8(<vscale x 1 x i8> %va, <vscale x 1 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv1i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf8, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 1 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i8> %head, <vscale x 1 x i8> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i8> %va, %splat
+  %y = sub <vscale x 1 x i8> %vb, %x
+  ret <vscale x 1 x i8> %y
+}
+
+define <vscale x 2 x i8> @vnmsub_vv_nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i8> %vb, <vscale x 2 x i8> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i8> %va, %vc
+  %y = sub <vscale x 2 x i8> %vb, %x
+  ret <vscale x 2 x i8> %y
+}
+
+define <vscale x 2 x i8> @vnmsub_vx_nxv2i8(<vscale x 2 x i8> %va, <vscale x 2 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv2i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf4, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 2 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i8> %head, <vscale x 2 x i8> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i8> %vb, %splat
+  %y = sub <vscale x 2 x i8> %va, %x
+  ret <vscale x 2 x i8> %y
+}
+
+define <vscale x 4 x i8> @vnmsub_vv_nxv4i8(<vscale x 4 x i8> %va, <vscale x 4 x i8> %vb, <vscale x 4 x i8> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, mf2, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i8> %vb, %va
+  %y = sub <vscale x 4 x i8> %vc, %x
+  ret <vscale x 4 x i8> %y
+}
+
+define <vscale x 4 x i8> @vnmsub_vx_nxv4i8(<vscale x 4 x i8> %va, <vscale x 4 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv4i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf2, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 4 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i8> %head, <vscale x 4 x i8> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i8> %va, %splat
+  %y = sub <vscale x 4 x i8> %vb, %x
+  ret <vscale x 4 x i8> %y
+}
+
+define <vscale x 8 x i8> @vnmsub_vv_nxv8i8(<vscale x 8 x i8> %va, <vscale x 8 x i8> %vb, <vscale x 8 x i8> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv8i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, m1, ta, mu
+; CHECK-NEXT:    vnmsac.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i8> %vb, %vc
+  %y = sub <vscale x 8 x i8> %va, %x
+  ret <vscale x 8 x i8> %y
+}
+
+define <vscale x 8 x i8> @vnmsub_vx_nxv8i8(<vscale x 8 x i8> %va, <vscale x 8 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv8i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m1, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 8 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i8> %head, <vscale x 8 x i8> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i8> %vb, %splat
+  %y = sub <vscale x 8 x i8> %va, %x
+  ret <vscale x 8 x i8> %y
+}
+
+define <vscale x 16 x i8> @vnmsub_vv_nxv16i8(<vscale x 16 x i8> %va, <vscale x 16 x i8> %vb, <vscale x 16 x i8> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, m2, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v12, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 16 x i8> %vc, %va
+  %y = sub <vscale x 16 x i8> %vb, %x
+  ret <vscale x 16 x i8> %y
+}
+
+define <vscale x 16 x i8> @vnmsub_vx_nxv16i8(<vscale x 16 x i8> %va, <vscale x 16 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m2, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v10
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 16 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %head, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+  %x = mul <vscale x 16 x i8> %va, %splat
+  %y = sub <vscale x 16 x i8> %vb, %x
+  ret <vscale x 16 x i8> %y
+}
+
+define <vscale x 32 x i8> @vnmsub_vv_nxv32i8(<vscale x 32 x i8> %va, <vscale x 32 x i8> %vb, <vscale x 32 x i8> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv32i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e8, m4, ta, mu
+; CHECK-NEXT:    vnmsac.vv v8, v16, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 32 x i8> %vc, %vb
+  %y = sub <vscale x 32 x i8> %va, %x
+  ret <vscale x 32 x i8> %y
+}
+
+define <vscale x 32 x i8> @vnmsub_vx_nxv32i8(<vscale x 32 x i8> %va, <vscale x 32 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv32i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m4, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v12
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 32 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 32 x i8> %head, <vscale x 32 x i8> undef, <vscale x 32 x i32> zeroinitializer
+  %x = mul <vscale x 32 x i8> %vb, %splat
+  %y = sub <vscale x 32 x i8> %va, %x
+  ret <vscale x 32 x i8> %y
+}
+
+define <vscale x 64 x i8> @vnmsub_vv_nxv64i8(<vscale x 64 x i8> %va, <vscale x 64 x i8> %vb, <vscale x 64 x i8> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv64i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8r.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e8, m8, ta, mu
+; CHECK-NEXT:    vnmsac.vv v8, v16, v24
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 64 x i8> %vc, %vb
+  %y = sub <vscale x 64 x i8> %va, %x
+  ret <vscale x 64 x i8> %y
+}
+
+define <vscale x 64 x i8> @vnmsub_vx_nxv64i8(<vscale x 64 x i8> %va, <vscale x 64 x i8> %vb, i8 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv64i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, m8, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 64 x i8> undef, i8 %c, i32 0
+  %splat = shufflevector <vscale x 64 x i8> %head, <vscale x 64 x i8> undef, <vscale x 64 x i32> zeroinitializer
+  %x = mul <vscale x 64 x i8> %vb, %splat
+  %y = sub <vscale x 64 x i8> %va, %x
+  ret <vscale x 64 x i8> %y
+}
+
+define <vscale x 1 x i16> @vnmsub_vv_nxv1i16(<vscale x 1 x i16> %va, <vscale x 1 x i16> %vb, <vscale x 1 x i16> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv1i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i16> %va, %vb
+  %y = sub <vscale x 1 x i16> %vc, %x
+  ret <vscale x 1 x i16> %y
+}
+
+define <vscale x 1 x i16> @vnmsub_vx_nxv1i16(<vscale x 1 x i16> %va, <vscale x 1 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv1i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf4, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 1 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i16> %head, <vscale x 1 x i16> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i16> %va, %splat
+  %y = sub <vscale x 1 x i16> %vb, %x
+  ret <vscale x 1 x i16> %y
+}
+
+define <vscale x 2 x i16> @vnmsub_vv_nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i16> %vb, <vscale x 2 x i16> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i16> %va, %vc
+  %y = sub <vscale x 2 x i16> %vb, %x
+  ret <vscale x 2 x i16> %y
+}
+
+define <vscale x 2 x i16> @vnmsub_vx_nxv2i16(<vscale x 2 x i16> %va, <vscale x 2 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv2i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, mf2, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 2 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i16> %head, <vscale x 2 x i16> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i16> %vb, %splat
+  %y = sub <vscale x 2 x i16> %va, %x
+  ret <vscale x 2 x i16> %y
+}
+
+define <vscale x 4 x i16> @vnmsub_vv_nxv4i16(<vscale x 4 x i16> %va, <vscale x 4 x i16> %vb, <vscale x 4 x i16> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i16> %vb, %va
+  %y = sub <vscale x 4 x i16> %vc, %x
+  ret <vscale x 4 x i16> %y
+}
+
+define <vscale x 4 x i16> @vnmsub_vx_nxv4i16(<vscale x 4 x i16> %va, <vscale x 4 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv4i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m1, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 4 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i16> %head, <vscale x 4 x i16> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i16> %va, %splat
+  %y = sub <vscale x 4 x i16> %vb, %x
+  ret <vscale x 4 x i16> %y
+}
+
+define <vscale x 8 x i16> @vnmsub_vv_nxv8i16(<vscale x 8 x i16> %va, <vscale x 8 x i16> %vb, <vscale x 8 x i16> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vnmsac.vv v8, v12, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i16> %vb, %vc
+  %y = sub <vscale x 8 x i16> %va, %x
+  ret <vscale x 8 x i16> %y
+}
+
+define <vscale x 8 x i16> @vnmsub_vx_nxv8i16(<vscale x 8 x i16> %va, <vscale x 8 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv8i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m2, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v10
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 8 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i16> %head, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i16> %vb, %splat
+  %y = sub <vscale x 8 x i16> %va, %x
+  ret <vscale x 8 x i16> %y
+}
+
+define <vscale x 16 x i16> @vnmsub_vv_nxv16i16(<vscale x 16 x i16> %va, <vscale x 16 x i16> %vb, <vscale x 16 x i16> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v16, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 16 x i16> %vc, %va
+  %y = sub <vscale x 16 x i16> %vb, %x
+  ret <vscale x 16 x i16> %y
+}
+
+define <vscale x 16 x i16> @vnmsub_vx_nxv16i16(<vscale x 16 x i16> %va, <vscale x 16 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv16i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m4, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v12
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 16 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 16 x i16> %head, <vscale x 16 x i16> undef, <vscale x 16 x i32> zeroinitializer
+  %x = mul <vscale x 16 x i16> %va, %splat
+  %y = sub <vscale x 16 x i16> %vb, %x
+  ret <vscale x 16 x i16> %y
+}
+
+define <vscale x 32 x i16> @vnmsub_vv_nxv32i16(<vscale x 32 x i16> %va, <vscale x 32 x i16> %vb, <vscale x 32 x i16> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv32i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8re16.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vnmsac.vv v8, v16, v24
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 32 x i16> %vc, %vb
+  %y = sub <vscale x 32 x i16> %va, %x
+  ret <vscale x 32 x i16> %y
+}
+
+define <vscale x 32 x i16> @vnmsub_vx_nxv32i16(<vscale x 32 x i16> %va, <vscale x 32 x i16> %vb, i16 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv32i16:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e16, m8, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 32 x i16> undef, i16 %c, i32 0
+  %splat = shufflevector <vscale x 32 x i16> %head, <vscale x 32 x i16> undef, <vscale x 32 x i32> zeroinitializer
+  %x = mul <vscale x 32 x i16> %vb, %splat
+  %y = sub <vscale x 32 x i16> %va, %x
+  ret <vscale x 32 x i16> %y
+}
+
+define <vscale x 1 x i32> @vnmsub_vv_nxv1i32(<vscale x 1 x i32> %va, <vscale x 1 x i32> %vb, <vscale x 1 x i32> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv1i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i32> %va, %vb
+  %y = sub <vscale x 1 x i32> %vc, %x
+  ret <vscale x 1 x i32> %y
+}
+
+define <vscale x 1 x i32> @vnmsub_vx_nxv1i32(<vscale x 1 x i32> %va, <vscale x 1 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv1i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, mf2, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 1 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i32> %head, <vscale x 1 x i32> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i32> %va, %splat
+  %y = sub <vscale x 1 x i32> %vb, %x
+  ret <vscale x 1 x i32> %y
+}
+
+define <vscale x 2 x i32> @vnmsub_vv_nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i32> %vb, <vscale x 2 x i32> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v10, v9
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i32> %va, %vc
+  %y = sub <vscale x 2 x i32> %vb, %x
+  ret <vscale x 2 x i32> %y
+}
+
+define <vscale x 2 x i32> @vnmsub_vx_nxv2i32(<vscale x 2 x i32> %va, <vscale x 2 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv2i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m1, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v9
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 2 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i32> %head, <vscale x 2 x i32> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i32> %vb, %splat
+  %y = sub <vscale x 2 x i32> %va, %x
+  ret <vscale x 2 x i32> %y
+}
+
+define <vscale x 4 x i32> @vnmsub_vv_nxv4i32(<vscale x 4 x i32> %va, <vscale x 4 x i32> %vb, <vscale x 4 x i32> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v10, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i32> %vb, %va
+  %y = sub <vscale x 4 x i32> %vc, %x
+  ret <vscale x 4 x i32> %y
+}
+
+define <vscale x 4 x i32> @vnmsub_vx_nxv4i32(<vscale x 4 x i32> %va, <vscale x 4 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv4i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m2, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v10
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 4 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i32> %head, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i32> %va, %splat
+  %y = sub <vscale x 4 x i32> %vb, %x
+  ret <vscale x 4 x i32> %y
+}
+
+define <vscale x 8 x i32> @vnmsub_vv_nxv8i32(<vscale x 8 x i32> %va, <vscale x 8 x i32> %vb, <vscale x 8 x i32> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vnmsac.vv v8, v16, v12
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i32> %vb, %vc
+  %y = sub <vscale x 8 x i32> %va, %x
+  ret <vscale x 8 x i32> %y
+}
+
+define <vscale x 8 x i32> @vnmsub_vx_nxv8i32(<vscale x 8 x i32> %va, <vscale x 8 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m4, ta, mu
+; CHECK-NEXT:    vnmsac.vx v8, a0, v12
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 8 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i32> %head, <vscale x 8 x i32> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i32> %vb, %splat
+  %y = sub <vscale x 8 x i32> %va, %x
+  ret <vscale x 8 x i32> %y
+}
+
+define <vscale x 16 x i32> @vnmsub_vv_nxv16i32(<vscale x 16 x i32> %va, <vscale x 16 x i32> %vb, <vscale x 16 x i32> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv16i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8re32.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v24, v16
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 16 x i32> %vc, %va
+  %y = sub <vscale x 16 x i32> %vb, %x
+  ret <vscale x 16 x i32> %y
+}
+
+define <vscale x 16 x i32> @vnmsub_vx_nxv16i32(<vscale x 16 x i32> %va, <vscale x 16 x i32> %vb, i32 %c) {
+; CHECK-LABEL: vnmsub_vx_nxv16i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e32, m8, ta, mu
+; CHECK-NEXT:    vnmsub.vx v8, a0, v16
+; CHECK-NEXT:    ret
+  %head = insertelement <vscale x 16 x i32> undef, i32 %c, i32 0
+  %splat = shufflevector <vscale x 16 x i32> %head, <vscale x 16 x i32> undef, <vscale x 16 x i32> zeroinitializer
+  %x = mul <vscale x 16 x i32> %va, %splat
+  %y = sub <vscale x 16 x i32> %vb, %x
+  ret <vscale x 16 x i32> %y
+}
+
+define <vscale x 1 x i64> @vnmsub_vv_nxv1i64(<vscale x 1 x i64> %va, <vscale x 1 x i64> %vb, <vscale x 1 x i64> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv1i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v9, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 1 x i64> %va, %vb
+  %y = sub <vscale x 1 x i64> %vc, %x
+  ret <vscale x 1 x i64> %y
+}
+
+define <vscale x 1 x i64> @vnmsub_vx_nxv1i64(<vscale x 1 x i64> %va, <vscale x 1 x i64> %vb, i64 %c) {
+; RV32-LABEL: vnmsub_vx_nxv1i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m1, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v25, (a0), zero
+; RV32-NEXT:    vnmsub.vv v8, v25, v9
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vnmsub_vx_nxv1i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m1, ta, mu
+; RV64-NEXT:    vnmsub.vx v8, a0, v9
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 1 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 1 x i64> %head, <vscale x 1 x i64> undef, <vscale x 1 x i32> zeroinitializer
+  %x = mul <vscale x 1 x i64> %va, %splat
+  %y = sub <vscale x 1 x i64> %vb, %x
+  ret <vscale x 1 x i64> %y
+}
+
+define <vscale x 2 x i64> @vnmsub_vv_nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i64> %vb, <vscale x 2 x i64> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv2i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v12, v10
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 2 x i64> %va, %vc
+  %y = sub <vscale x 2 x i64> %vb, %x
+  ret <vscale x 2 x i64> %y
+}
+
+define <vscale x 2 x i64> @vnmsub_vx_nxv2i64(<vscale x 2 x i64> %va, <vscale x 2 x i64> %vb, i64 %c) {
+; RV32-LABEL: vnmsub_vx_nxv2i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m2, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v26, (a0), zero
+; RV32-NEXT:    vnmsac.vv v8, v10, v26
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vnmsub_vx_nxv2i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m2, ta, mu
+; RV64-NEXT:    vnmsac.vx v8, a0, v10
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 2 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 2 x i64> %head, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
+  %x = mul <vscale x 2 x i64> %vb, %splat
+  %y = sub <vscale x 2 x i64> %va, %x
+  ret <vscale x 2 x i64> %y
+}
+
+define <vscale x 4 x i64> @vnmsub_vv_nxv4i64(<vscale x 4 x i64> %va, <vscale x 4 x i64> %vb, <vscale x 4 x i64> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; CHECK-NEXT:    vnmsub.vv v8, v12, v16
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 4 x i64> %vb, %va
+  %y = sub <vscale x 4 x i64> %vc, %x
+  ret <vscale x 4 x i64> %y
+}
+
+define <vscale x 4 x i64> @vnmsub_vx_nxv4i64(<vscale x 4 x i64> %va, <vscale x 4 x i64> %vb, i64 %c) {
+; RV32-LABEL: vnmsub_vx_nxv4i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m4, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v28, (a0), zero
+; RV32-NEXT:    vnmsub.vv v8, v28, v12
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vnmsub_vx_nxv4i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m4, ta, mu
+; RV64-NEXT:    vnmsub.vx v8, a0, v12
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 4 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 4 x i64> %head, <vscale x 4 x i64> undef, <vscale x 4 x i32> zeroinitializer
+  %x = mul <vscale x 4 x i64> %va, %splat
+  %y = sub <vscale x 4 x i64> %vb, %x
+  ret <vscale x 4 x i64> %y
+}
+
+define <vscale x 8 x i64> @vnmsub_vv_nxv8i64(<vscale x 8 x i64> %va, <vscale x 8 x i64> %vb, <vscale x 8 x i64> %vc) {
+; CHECK-LABEL: vnmsub_vv_nxv8i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vl8re64.v v24, (a0)
+; CHECK-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; CHECK-NEXT:    vnmsac.vv v8, v16, v24
+; CHECK-NEXT:    ret
+  %x = mul <vscale x 8 x i64> %vb, %vc
+  %y = sub <vscale x 8 x i64> %va, %x
+  ret <vscale x 8 x i64> %y
+}
+
+define <vscale x 8 x i64> @vnmsub_vx_nxv8i64(<vscale x 8 x i64> %va, <vscale x 8 x i64> %vb, i64 %c) {
+; RV32-LABEL: vnmsub_vx_nxv8i64:
+; RV32:       # %bb.0:
+; RV32-NEXT:    addi sp, sp, -16
+; RV32-NEXT:    .cfi_def_cfa_offset 16
+; RV32-NEXT:    sw a1, 12(sp)
+; RV32-NEXT:    sw a0, 8(sp)
+; RV32-NEXT:    vsetvli a0, zero, e64, m8, ta, mu
+; RV32-NEXT:    addi a0, sp, 8
+; RV32-NEXT:    vlse64.v v24, (a0), zero
+; RV32-NEXT:    vnmsac.vv v8, v16, v24
+; RV32-NEXT:    addi sp, sp, 16
+; RV32-NEXT:    ret
+;
+; RV64-LABEL: vnmsub_vx_nxv8i64:
+; RV64:       # %bb.0:
+; RV64-NEXT:    vsetvli a1, zero, e64, m8, ta, mu
+; RV64-NEXT:    vnmsac.vx v8, a0, v16
+; RV64-NEXT:    ret
+  %head = insertelement <vscale x 8 x i64> undef, i64 %c, i32 0
+  %splat = shufflevector <vscale x 8 x i64> %head, <vscale x 8 x i64> undef, <vscale x 8 x i32> zeroinitializer
+  %x = mul <vscale x 8 x i64> %vb, %splat
+  %y = sub <vscale x 8 x i64> %va, %x
+  ret <vscale x 8 x i64> %y
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv32.ll b/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv32.ll
--- a/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv32.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv32.ll
@@ -34,8 +34,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 1 x i8> %head, <vscale x 1 x i8> undef, <vscale x 1 x i32> zeroinitializer
@@ -76,8 +75,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 2 x i8> %head, <vscale x 2 x i8> undef, <vscale x 2 x i32> zeroinitializer
@@ -118,8 +116,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 4 x i8> %head, <vscale x 4 x i8> undef, <vscale x 4 x i32> zeroinitializer
@@ -160,8 +157,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 8 x i8> %head, <vscale x 8 x i8> undef, <vscale x 8 x i32> zeroinitializer
@@ -202,8 +198,7 @@
 ; CHECK-NEXT:    vsrl.vi v28, v26, 7
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 16 x i8> %head, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
@@ -244,8 +239,7 @@
 ; CHECK-NEXT:    vsrl.vi v12, v28, 7
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 32 x i8> %head, <vscale x 32 x i8> undef, <vscale x 32 x i32> zeroinitializer
@@ -286,8 +280,7 @@
 ; CHECK-NEXT:    vsrl.vi v24, v16, 7
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 64 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 64 x i8> %head, <vscale x 64 x i8> undef, <vscale x 64 x i32> zeroinitializer
@@ -328,8 +321,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 15
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 1 x i16> %head, <vscale x 1 x i16> undef, <vscale x 1 x i32> zeroinitializer
@@ -370,8 +362,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 15
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 2 x i16> %head, <vscale x 2 x i16> undef, <vscale x 2 x i32> zeroinitializer
@@ -412,8 +403,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 15
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 4 x i16> %head, <vscale x 4 x i16> undef, <vscale x 4 x i32> zeroinitializer
@@ -454,8 +444,7 @@
 ; CHECK-NEXT:    vsrl.vi v28, v26, 15
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 8 x i16> %head, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
@@ -496,8 +485,7 @@
 ; CHECK-NEXT:    vsrl.vi v12, v28, 15
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 16 x i16> %head, <vscale x 16 x i16> undef, <vscale x 16 x i32> zeroinitializer
@@ -538,8 +526,7 @@
 ; CHECK-NEXT:    vsrl.vi v24, v16, 15
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 32 x i16> %head, <vscale x 32 x i16> undef, <vscale x 32 x i32> zeroinitializer
@@ -581,8 +568,7 @@
 ; CHECK-NEXT:    vsra.vi v25, v25, 2
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 1 x i32> %head, <vscale x 1 x i32> undef, <vscale x 1 x i32> zeroinitializer
@@ -624,8 +610,7 @@
 ; CHECK-NEXT:    vsra.vi v25, v25, 2
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 2 x i32> %head, <vscale x 2 x i32> undef, <vscale x 2 x i32> zeroinitializer
@@ -667,8 +652,7 @@
 ; CHECK-NEXT:    vsra.vi v26, v26, 2
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 4 x i32> %head, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
@@ -710,8 +694,7 @@
 ; CHECK-NEXT:    vsra.vi v28, v28, 2
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 8 x i32> %head, <vscale x 8 x i32> undef, <vscale x 8 x i32> zeroinitializer
@@ -753,8 +736,7 @@
 ; CHECK-NEXT:    vsra.vi v16, v16, 2
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 16 x i32> %head, <vscale x 16 x i32> undef, <vscale x 16 x i32> zeroinitializer
@@ -811,8 +793,7 @@
 ; CHECK-NEXT:    vsra.vi v25, v25, 1
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i64> undef, i64 -7, i32 0
@@ -870,8 +851,7 @@
 ; CHECK-NEXT:    vsra.vi v26, v26, 1
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i64> undef, i64 -7, i32 0
@@ -929,8 +909,7 @@
 ; CHECK-NEXT:    vsra.vi v28, v28, 1
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i64> undef, i64 -7, i32 0
@@ -988,8 +967,7 @@
 ; CHECK-NEXT:    vsra.vi v16, v16, 1
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i64> undef, i64 -7, i32 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv64.ll b/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv64.ll
--- a/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv64.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vrem-sdnode-rv64.ll
@@ -34,8 +34,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 1 x i8> %head, <vscale x 1 x i8> undef, <vscale x 1 x i32> zeroinitializer
@@ -76,8 +75,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 2 x i8> %head, <vscale x 2 x i8> undef, <vscale x 2 x i32> zeroinitializer
@@ -118,8 +116,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 4 x i8> %head, <vscale x 4 x i8> undef, <vscale x 4 x i32> zeroinitializer
@@ -160,8 +157,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 7
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 8 x i8> %head, <vscale x 8 x i8> undef, <vscale x 8 x i32> zeroinitializer
@@ -202,8 +198,7 @@
 ; CHECK-NEXT:    vsrl.vi v28, v26, 7
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 16 x i8> %head, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
@@ -244,8 +239,7 @@
 ; CHECK-NEXT:    vsrl.vi v12, v28, 7
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 32 x i8> %head, <vscale x 32 x i8> undef, <vscale x 32 x i32> zeroinitializer
@@ -286,8 +280,7 @@
 ; CHECK-NEXT:    vsrl.vi v24, v16, 7
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 64 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 64 x i8> %head, <vscale x 64 x i8> undef, <vscale x 64 x i32> zeroinitializer
@@ -328,8 +321,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 15
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 1 x i16> %head, <vscale x 1 x i16> undef, <vscale x 1 x i32> zeroinitializer
@@ -370,8 +362,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 15
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 2 x i16> %head, <vscale x 2 x i16> undef, <vscale x 2 x i32> zeroinitializer
@@ -412,8 +403,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 15
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 4 x i16> %head, <vscale x 4 x i16> undef, <vscale x 4 x i32> zeroinitializer
@@ -454,8 +444,7 @@
 ; CHECK-NEXT:    vsrl.vi v28, v26, 15
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 8 x i16> %head, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
@@ -496,8 +485,7 @@
 ; CHECK-NEXT:    vsrl.vi v12, v28, 15
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 16 x i16> %head, <vscale x 16 x i16> undef, <vscale x 16 x i32> zeroinitializer
@@ -538,8 +526,7 @@
 ; CHECK-NEXT:    vsrl.vi v24, v16, 15
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 32 x i16> %head, <vscale x 32 x i16> undef, <vscale x 32 x i32> zeroinitializer
@@ -581,8 +568,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 31
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 1 x i32> %head, <vscale x 1 x i32> undef, <vscale x 1 x i32> zeroinitializer
@@ -624,8 +610,7 @@
 ; CHECK-NEXT:    vsrl.vi v26, v25, 31
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 2 x i32> %head, <vscale x 2 x i32> undef, <vscale x 2 x i32> zeroinitializer
@@ -667,8 +652,7 @@
 ; CHECK-NEXT:    vsrl.vi v28, v26, 31
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 4 x i32> %head, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
@@ -710,8 +694,7 @@
 ; CHECK-NEXT:    vsrl.vi v12, v28, 31
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 8 x i32> %head, <vscale x 8 x i32> undef, <vscale x 8 x i32> zeroinitializer
@@ -753,8 +736,7 @@
 ; CHECK-NEXT:    vsrl.vi v24, v16, 31
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 16 x i32> %head, <vscale x 16 x i32> undef, <vscale x 16 x i32> zeroinitializer
@@ -802,8 +784,7 @@
 ; CHECK-NEXT:    vsra.vi v25, v25, 1
 ; CHECK-NEXT:    vadd.vv v25, v25, v26
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 1 x i64> %head, <vscale x 1 x i64> undef, <vscale x 1 x i32> zeroinitializer
@@ -851,8 +832,7 @@
 ; CHECK-NEXT:    vsra.vi v26, v26, 1
 ; CHECK-NEXT:    vadd.vv v26, v26, v28
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 2 x i64> %head, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
@@ -900,8 +880,7 @@
 ; CHECK-NEXT:    vsra.vi v28, v28, 1
 ; CHECK-NEXT:    vadd.vv v28, v28, v12
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 4 x i64> %head, <vscale x 4 x i64> undef, <vscale x 4 x i32> zeroinitializer
@@ -949,8 +928,7 @@
 ; CHECK-NEXT:    vsra.vi v16, v16, 1
 ; CHECK-NEXT:    vadd.vv v16, v16, v24
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 8 x i64> %head, <vscale x 8 x i64> undef, <vscale x 8 x i32> zeroinitializer
diff --git a/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv32.ll b/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv32.ll
--- a/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv32.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv32.ll
@@ -31,8 +31,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 1 x i8> %head, <vscale x 1 x i8> undef, <vscale x 1 x i32> zeroinitializer
@@ -70,8 +69,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 2 x i8> %head, <vscale x 2 x i8> undef, <vscale x 2 x i32> zeroinitializer
@@ -109,8 +107,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 4 x i8> %head, <vscale x 4 x i8> undef, <vscale x 4 x i32> zeroinitializer
@@ -148,8 +145,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 8 x i8> %head, <vscale x 8 x i8> undef, <vscale x 8 x i32> zeroinitializer
@@ -187,8 +183,7 @@
 ; CHECK-NEXT:    vmulhu.vx v26, v8, a0
 ; CHECK-NEXT:    vsrl.vi v26, v26, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 16 x i8> %head, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
@@ -226,8 +221,7 @@
 ; CHECK-NEXT:    vmulhu.vx v28, v8, a0
 ; CHECK-NEXT:    vsrl.vi v28, v28, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 32 x i8> %head, <vscale x 32 x i8> undef, <vscale x 32 x i32> zeroinitializer
@@ -265,8 +259,7 @@
 ; CHECK-NEXT:    vmulhu.vx v16, v8, a0
 ; CHECK-NEXT:    vsrl.vi v16, v16, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 64 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 64 x i8> %head, <vscale x 64 x i8> undef, <vscale x 64 x i32> zeroinitializer
@@ -305,8 +298,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 1 x i16> %head, <vscale x 1 x i16> undef, <vscale x 1 x i32> zeroinitializer
@@ -345,8 +337,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 2 x i16> %head, <vscale x 2 x i16> undef, <vscale x 2 x i32> zeroinitializer
@@ -385,8 +376,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 4 x i16> %head, <vscale x 4 x i16> undef, <vscale x 4 x i32> zeroinitializer
@@ -425,8 +415,7 @@
 ; CHECK-NEXT:    vmulhu.vx v26, v8, a0
 ; CHECK-NEXT:    vsrl.vi v26, v26, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 8 x i16> %head, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
@@ -465,8 +454,7 @@
 ; CHECK-NEXT:    vmulhu.vx v28, v8, a0
 ; CHECK-NEXT:    vsrl.vi v28, v28, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 16 x i16> %head, <vscale x 16 x i16> undef, <vscale x 16 x i32> zeroinitializer
@@ -505,8 +493,7 @@
 ; CHECK-NEXT:    vmulhu.vx v16, v8, a0
 ; CHECK-NEXT:    vsrl.vi v16, v16, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 32 x i16> %head, <vscale x 32 x i16> undef, <vscale x 32 x i32> zeroinitializer
@@ -545,8 +532,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 1 x i32> %head, <vscale x 1 x i32> undef, <vscale x 1 x i32> zeroinitializer
@@ -585,8 +571,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 2 x i32> %head, <vscale x 2 x i32> undef, <vscale x 2 x i32> zeroinitializer
@@ -625,8 +610,7 @@
 ; CHECK-NEXT:    vmulhu.vx v26, v8, a0
 ; CHECK-NEXT:    vsrl.vi v26, v26, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 4 x i32> %head, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
@@ -665,8 +649,7 @@
 ; CHECK-NEXT:    vmulhu.vx v28, v8, a0
 ; CHECK-NEXT:    vsrl.vi v28, v28, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 8 x i32> %head, <vscale x 8 x i32> undef, <vscale x 8 x i32> zeroinitializer
@@ -705,8 +688,7 @@
 ; CHECK-NEXT:    vmulhu.vx v16, v8, a0
 ; CHECK-NEXT:    vsrl.vi v16, v16, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 16 x i32> %head, <vscale x 16 x i32> undef, <vscale x 16 x i32> zeroinitializer
@@ -759,8 +741,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v25, v25, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i64> undef, i64 -7, i32 0
@@ -814,8 +795,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v26, v26, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i64> undef, i64 -7, i32 0
@@ -869,8 +849,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v28, v28, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i64> undef, i64 -7, i32 0
@@ -924,8 +903,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v16, v16, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    addi sp, sp, 16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i64> undef, i64 -7, i32 0
diff --git a/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv64.ll b/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv64.ll
--- a/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv64.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vremu-sdnode-rv64.ll
@@ -31,8 +31,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 1 x i8> %head, <vscale x 1 x i8> undef, <vscale x 1 x i32> zeroinitializer
@@ -70,8 +69,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 2 x i8> %head, <vscale x 2 x i8> undef, <vscale x 2 x i32> zeroinitializer
@@ -109,8 +107,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 4 x i8> %head, <vscale x 4 x i8> undef, <vscale x 4 x i32> zeroinitializer
@@ -148,8 +145,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 8 x i8> %head, <vscale x 8 x i8> undef, <vscale x 8 x i32> zeroinitializer
@@ -187,8 +183,7 @@
 ; CHECK-NEXT:    vmulhu.vx v26, v8, a0
 ; CHECK-NEXT:    vsrl.vi v26, v26, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 16 x i8> %head, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
@@ -226,8 +221,7 @@
 ; CHECK-NEXT:    vmulhu.vx v28, v8, a0
 ; CHECK-NEXT:    vsrl.vi v28, v28, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 32 x i8> %head, <vscale x 32 x i8> undef, <vscale x 32 x i32> zeroinitializer
@@ -265,8 +259,7 @@
 ; CHECK-NEXT:    vmulhu.vx v16, v8, a0
 ; CHECK-NEXT:    vsrl.vi v16, v16, 5
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 64 x i8> undef, i8 -7, i32 0
   %splat = shufflevector <vscale x 64 x i8> %head, <vscale x 64 x i8> undef, <vscale x 64 x i32> zeroinitializer
@@ -305,8 +298,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 1 x i16> %head, <vscale x 1 x i16> undef, <vscale x 1 x i32> zeroinitializer
@@ -345,8 +337,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 2 x i16> %head, <vscale x 2 x i16> undef, <vscale x 2 x i32> zeroinitializer
@@ -385,8 +376,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 4 x i16> %head, <vscale x 4 x i16> undef, <vscale x 4 x i32> zeroinitializer
@@ -425,8 +415,7 @@
 ; CHECK-NEXT:    vmulhu.vx v26, v8, a0
 ; CHECK-NEXT:    vsrl.vi v26, v26, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 8 x i16> %head, <vscale x 8 x i16> undef, <vscale x 8 x i32> zeroinitializer
@@ -465,8 +454,7 @@
 ; CHECK-NEXT:    vmulhu.vx v28, v8, a0
 ; CHECK-NEXT:    vsrl.vi v28, v28, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 16 x i16> %head, <vscale x 16 x i16> undef, <vscale x 16 x i32> zeroinitializer
@@ -505,8 +493,7 @@
 ; CHECK-NEXT:    vmulhu.vx v16, v8, a0
 ; CHECK-NEXT:    vsrl.vi v16, v16, 13
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 32 x i16> undef, i16 -7, i32 0
   %splat = shufflevector <vscale x 32 x i16> %head, <vscale x 32 x i16> undef, <vscale x 32 x i32> zeroinitializer
@@ -545,8 +532,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 1 x i32> %head, <vscale x 1 x i32> undef, <vscale x 1 x i32> zeroinitializer
@@ -585,8 +571,7 @@
 ; CHECK-NEXT:    vmulhu.vx v25, v8, a0
 ; CHECK-NEXT:    vsrl.vi v25, v25, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 2 x i32> %head, <vscale x 2 x i32> undef, <vscale x 2 x i32> zeroinitializer
@@ -625,8 +610,7 @@
 ; CHECK-NEXT:    vmulhu.vx v26, v8, a0
 ; CHECK-NEXT:    vsrl.vi v26, v26, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 4 x i32> %head, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
@@ -665,8 +649,7 @@
 ; CHECK-NEXT:    vmulhu.vx v28, v8, a0
 ; CHECK-NEXT:    vsrl.vi v28, v28, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 8 x i32> %head, <vscale x 8 x i32> undef, <vscale x 8 x i32> zeroinitializer
@@ -705,8 +688,7 @@
 ; CHECK-NEXT:    vmulhu.vx v16, v8, a0
 ; CHECK-NEXT:    vsrl.vi v16, v16, 29
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 16 x i32> undef, i32 -7, i32 0
   %splat = shufflevector <vscale x 16 x i32> %head, <vscale x 16 x i32> undef, <vscale x 16 x i32> zeroinitializer
@@ -747,8 +729,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v25, v25, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v25, v25, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v25
+; CHECK-NEXT:    vnmsac.vx v8, a0, v25
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 1 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 1 x i64> %head, <vscale x 1 x i64> undef, <vscale x 1 x i32> zeroinitializer
@@ -789,8 +770,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v26, v26, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v26, v26, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v26
+; CHECK-NEXT:    vnmsac.vx v8, a0, v26
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 2 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 2 x i64> %head, <vscale x 2 x i64> undef, <vscale x 2 x i32> zeroinitializer
@@ -831,8 +811,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v28, v28, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v28, v28, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v28
+; CHECK-NEXT:    vnmsac.vx v8, a0, v28
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 4 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 4 x i64> %head, <vscale x 4 x i64> undef, <vscale x 4 x i32> zeroinitializer
@@ -873,8 +852,7 @@
 ; CHECK-NEXT:    addi a0, zero, 61
 ; CHECK-NEXT:    vsrl.vx v16, v16, a0
 ; CHECK-NEXT:    addi a0, zero, -7
-; CHECK-NEXT:    vmul.vx v16, v16, a0
-; CHECK-NEXT:    vsub.vv v8, v8, v16
+; CHECK-NEXT:    vnmsac.vx v8, a0, v16
 ; CHECK-NEXT:    ret
   %head = insertelement <vscale x 8 x i64> undef, i64 -7, i32 0
   %splat = shufflevector <vscale x 8 x i64> %head, <vscale x 8 x i64> undef, <vscale x 8 x i32> zeroinitializer