Index: llvm/lib/Target/ARM/ARMInstrInfo.td
===================================================================
--- llvm/lib/Target/ARM/ARMInstrInfo.td
+++ llvm/lib/Target/ARM/ARMInstrInfo.td
@@ -498,6 +498,18 @@
 }]>;
 
 
+def ARMimmAllZerosV: PatLeaf<(bitconvert (v4i32 (ARMvmovImm (i32 0))))>;
+def ARMimmAllZerosD: PatLeaf<(bitconvert (v2i32 (ARMvmovImm (i32 0))))>;
+def ARMimmAllOnesV: PatLeaf<(bitconvert (v16i8 (ARMvmovImm (i32 0xEFF))))>;
+def ARMimmAllOnesD: PatLeaf<(bitconvert (v8i8 (ARMvmovImm (i32 0xEFF))))>;
+
+def ARMimmOneV: PatLeaf<(ARMvmovImm (i32 timm)), [{
+  ConstantSDNode *ConstVal = cast<ConstantSDNode>(N->getOperand(0));
+  unsigned EltBits = 0;
+  uint64_t EltVal = ARM_AM::decodeVMOVModImm(ConstVal->getZExtValue(), EltBits);
+  return (EltBits == N->getValueType(0).getScalarSizeInBits() && EltVal == 0x01);
+}]>;
+
 
 //===----------------------------------------------------------------------===//
 // Operand Definitions.
Index: llvm/lib/Target/ARM/ARMInstrMVE.td
===================================================================
--- llvm/lib/Target/ARM/ARMInstrMVE.td
+++ llvm/lib/Target/ARM/ARMInstrMVE.td
@@ -318,9 +318,9 @@
 def MVE_v16p8 : MVEVectorVTInfo<v16i8, v8i16, v16i1, v8i1, 0b11, "p", 0b0>;
 def MVE_v8p16 : MVEVectorVTInfo<v8i16, v4i32, v8i1,  v4i1, 0b11, "p", 0b1>;
 
-
 multiclass MVE_TwoOpPattern<MVEVectorVTInfo VTI, PatFrag Op, Intrinsic PredInt,
-                            dag PredOperands, Instruction Inst> {
+                            dag PredOperands, Instruction Inst,
+                            SDPatternOperator IdentityVec = null_frag> {
   // Unpredicated
   def : Pat<(VTI.Vec (Op (VTI.Vec MQPR:$Qm), (VTI.Vec MQPR:$Qn))),
             (VTI.Vec (Inst (VTI.Vec MQPR:$Qm), (VTI.Vec MQPR:$Qn)))>;
@@ -334,6 +334,15 @@
               (VTI.Vec (Inst (VTI.Vec MQPR:$Qm), (VTI.Vec MQPR:$Qn),
                               ARMVCCThen, (VTI.Pred VCCR:$mask),
                               (VTI.Vec MQPR:$inactive)))>;
+
+    // Optionally with the select folded through the op
+    def : Pat<(VTI.Vec (Op (VTI.Vec MQPR:$Qm),
+                           (VTI.Vec (vselect (VTI.Pred VCCR:$mask),
+                                             (VTI.Vec MQPR:$Qn),
+                                             (VTI.Vec IdentityVec))))),
+              (VTI.Vec (Inst (VTI.Vec MQPR:$Qm), (VTI.Vec MQPR:$Qn),
+                              ARMVCCThen, (VTI.Pred VCCR:$mask),
+                              (VTI.Vec MQPR:$Qm)))>;
   }
 
   // Predicated with intrinsic
@@ -346,7 +355,8 @@
 }
 
 multiclass MVE_TwoOpPatternDup<MVEVectorVTInfo VTI, PatFrag Op, Intrinsic PredInt,
-                               dag PredOperands, Instruction Inst> {
+                               dag PredOperands, Instruction Inst,
+                               SDPatternOperator IdentityVec = null_frag> {
   // Unpredicated
   def : Pat<(VTI.Vec (Op (VTI.Vec MQPR:$Qm), (VTI.Vec (ARMvdup rGPR:$Rn)))),
             (VTI.Vec (Inst (VTI.Vec MQPR:$Qm), rGPR:$Rn))>;
@@ -360,6 +370,15 @@
               (VTI.Vec (Inst (VTI.Vec MQPR:$Qm), rGPR:$Rn,
                               ARMVCCThen, (VTI.Pred VCCR:$mask),
                               (VTI.Vec MQPR:$inactive)))>;
+
+    // Optionally with the select folded through the op
+    def : Pat<(VTI.Vec (Op (VTI.Vec MQPR:$Qm),
+                           (VTI.Vec (vselect (VTI.Pred VCCR:$mask),
+                                             (ARMvdup rGPR:$Rn),
+                                             (VTI.Vec IdentityVec))))),
+              (VTI.Vec (Inst (VTI.Vec MQPR:$Qm), rGPR:$Rn,
+                              ARMVCCThen, (VTI.Pred VCCR:$mask),
+                              (VTI.Vec MQPR:$Qm)))>;
   }
 
   // Predicated with intrinsic
@@ -1492,20 +1511,20 @@
 }
 
 let Predicates = [HasMVEInt] in {
-  defm : MVE_TwoOpPattern<MVE_v16i8, and, int_arm_mve_and_predicated, (? ), MVE_VAND>;
-  defm : MVE_TwoOpPattern<MVE_v8i16, and, int_arm_mve_and_predicated, (? ), MVE_VAND>;
-  defm : MVE_TwoOpPattern<MVE_v4i32, and, int_arm_mve_and_predicated, (? ), MVE_VAND>;
-  defm : MVE_TwoOpPattern<MVE_v2i64, and, int_arm_mve_and_predicated, (? ), MVE_VAND>;
+  defm : MVE_TwoOpPattern<MVE_v16i8, and, int_arm_mve_and_predicated, (? ), MVE_VAND, ARMimmAllOnesV>;
+  defm : MVE_TwoOpPattern<MVE_v8i16, and, int_arm_mve_and_predicated, (? ), MVE_VAND, ARMimmAllOnesV>;
+  defm : MVE_TwoOpPattern<MVE_v4i32, and, int_arm_mve_and_predicated, (? ), MVE_VAND, ARMimmAllOnesV>;
+  defm : MVE_TwoOpPattern<MVE_v2i64, and, int_arm_mve_and_predicated, (? ), MVE_VAND, ARMimmAllOnesV>;
 
-  defm : MVE_TwoOpPattern<MVE_v16i8, or, int_arm_mve_orr_predicated, (? ), MVE_VORR>;
-  defm : MVE_TwoOpPattern<MVE_v8i16, or, int_arm_mve_orr_predicated, (? ), MVE_VORR>;
-  defm : MVE_TwoOpPattern<MVE_v4i32, or, int_arm_mve_orr_predicated, (? ), MVE_VORR>;
-  defm : MVE_TwoOpPattern<MVE_v2i64, or, int_arm_mve_orr_predicated, (? ), MVE_VORR>;
+  defm : MVE_TwoOpPattern<MVE_v16i8, or, int_arm_mve_orr_predicated, (? ), MVE_VORR, ARMimmAllZerosV>;
+  defm : MVE_TwoOpPattern<MVE_v8i16, or, int_arm_mve_orr_predicated, (? ), MVE_VORR, ARMimmAllZerosV>;
+  defm : MVE_TwoOpPattern<MVE_v4i32, or, int_arm_mve_orr_predicated, (? ), MVE_VORR, ARMimmAllZerosV>;
+  defm : MVE_TwoOpPattern<MVE_v2i64, or, int_arm_mve_orr_predicated, (? ), MVE_VORR, ARMimmAllZerosV>;
 
-  defm : MVE_TwoOpPattern<MVE_v16i8, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR>;
-  defm : MVE_TwoOpPattern<MVE_v8i16, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR>;
-  defm : MVE_TwoOpPattern<MVE_v4i32, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR>;
-  defm : MVE_TwoOpPattern<MVE_v2i64, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR>;
+  defm : MVE_TwoOpPattern<MVE_v16i8, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR, ARMimmAllZerosV>;
+  defm : MVE_TwoOpPattern<MVE_v8i16, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR, ARMimmAllZerosV>;
+  defm : MVE_TwoOpPattern<MVE_v4i32, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR, ARMimmAllZerosV>;
+  defm : MVE_TwoOpPattern<MVE_v2i64, xor, int_arm_mve_eor_predicated, (? ), MVE_VEOR, ARMimmAllZerosV>;
 
   defm : MVE_TwoOpPattern<MVE_v16i8, BinOpFrag<(and node:$LHS, (vnotq node:$RHS))>,
                           int_arm_mve_bic_predicated, (? ), MVE_VBIC>;
@@ -1775,7 +1794,7 @@
 
   let Predicates = [HasMVEInt] in {
     defm : MVE_TwoOpPattern<VTI, mul, int_arm_mve_mul_predicated, (? ),
-                            !cast<Instruction>(NAME)>;
+                            !cast<Instruction>(NAME), ARMimmOneV>;
   }
 }
 
@@ -1849,7 +1868,7 @@
   defvar Inst = !cast<Instruction>(NAME);
 
   let Predicates = [HasMVEInt] in {
-    defm : MVE_TwoOpPattern<VTI, Op, PredInt, (? ), !cast<Instruction>(NAME)>;
+    defm : MVE_TwoOpPattern<VTI, Op, PredInt, (? ), !cast<Instruction>(NAME), ARMimmAllZerosV>;
   }
 }
 
@@ -4984,7 +5003,7 @@
                             SDNode Op, Intrinsic PredInt> {
   def "" : MVE_VADDSUB_qr<iname, VTI.Suffix, VTI.Size, 0b0, subtract, 0b1, 0b0>;
   let Predicates = [HasMVEInt] in {
-    defm : MVE_TwoOpPatternDup<VTI, Op, PredInt, (? ), !cast<Instruction>(NAME)>;
+    defm : MVE_TwoOpPatternDup<VTI, Op, PredInt, (? ), !cast<Instruction>(NAME), ARMimmAllZerosV>;
   }
 }
 
@@ -5270,7 +5289,7 @@
 multiclass MVE_VMUL_qr_int_m<MVEVectorVTInfo VTI> {
   def "" : MVE_VMUL_qr_int<"vmul", VTI.Suffix, VTI.Size>;
   defm : MVE_TwoOpPatternDup<VTI, mul, int_arm_mve_mul_predicated, (? ),
-                             !cast<Instruction>(NAME)>;
+                             !cast<Instruction>(NAME), ARMimmOneV>;
 }
 
 defm MVE_VMUL_qr_i8  : MVE_VMUL_qr_int_m<MVE_v16i8>;
@@ -6864,7 +6883,7 @@
 
 class MVE_vector_maskedload_typed<ValueType Ty, Instruction RegImmInst,
                                   PatFrag LoadKind, int shift>
-  : Pat<(Ty (LoadKind t2addrmode_imm7<shift>:$addr, VCCR:$pred, (Ty NEONimmAllZerosV))),
+  : Pat<(Ty (LoadKind t2addrmode_imm7<shift>:$addr, VCCR:$pred, (Ty (ARMvmovImm (i32 0))))),
         (Ty (RegImmInst t2addrmode_imm7<shift>:$addr, ARMVCCThen, VCCR:$pred))>;
 
 multiclass MVE_vector_load<Instruction RegImmInst, PatFrag LoadKind,
@@ -7031,11 +7050,11 @@
             (VT (LoadUInst taddrmode_imm7<Shift>:$addr))>;
 
   // Masked ext loads
-  def : Pat<(VT (!cast<PatFrag>("aligned_extmaskedload"#Amble) taddrmode_imm7<Shift>:$addr, VCCR:$pred, (VT NEONimmAllZerosV))),
+  def : Pat<(VT (!cast<PatFrag>("aligned_extmaskedload"#Amble) taddrmode_imm7<Shift>:$addr, VCCR:$pred, (VT (ARMvmovImm (i32 0))))),
             (VT (LoadUInst taddrmode_imm7<Shift>:$addr, ARMVCCThen, VCCR:$pred))>;
-  def : Pat<(VT (!cast<PatFrag>("aligned_sextmaskedload"#Amble) taddrmode_imm7<Shift>:$addr, VCCR:$pred, (VT NEONimmAllZerosV))),
+  def : Pat<(VT (!cast<PatFrag>("aligned_sextmaskedload"#Amble) taddrmode_imm7<Shift>:$addr, VCCR:$pred, (VT (ARMvmovImm (i32 0))))),
             (VT (LoadSInst taddrmode_imm7<Shift>:$addr, ARMVCCThen, VCCR:$pred))>;
-  def : Pat<(VT (!cast<PatFrag>("aligned_zextmaskedload"#Amble) taddrmode_imm7<Shift>:$addr, VCCR:$pred, (VT NEONimmAllZerosV))),
+  def : Pat<(VT (!cast<PatFrag>("aligned_zextmaskedload"#Amble) taddrmode_imm7<Shift>:$addr, VCCR:$pred, (VT (ARMvmovImm (i32 0))))),
             (VT (LoadUInst taddrmode_imm7<Shift>:$addr, ARMVCCThen, VCCR:$pred))>;
 }
 
Index: llvm/lib/Target/ARM/ARMInstrNEON.td
===================================================================
--- llvm/lib/Target/ARM/ARMInstrNEON.td
+++ llvm/lib/Target/ARM/ARMInstrNEON.td
@@ -534,20 +534,6 @@
 def NEONvtbl2     : SDNode<"ARMISD::VTBL2", SDTARMVTBL2>;
 
 
-def NEONimmAllZerosV: PatLeaf<(ARMvmovImm (i32 timm)), [{
-  ConstantSDNode *ConstVal = cast<ConstantSDNode>(N->getOperand(0));
-  unsigned EltBits = 0;
-  uint64_t EltVal = ARM_AM::decodeVMOVModImm(ConstVal->getZExtValue(), EltBits);
-  return (EltBits == 32 && EltVal == 0);
-}]>;
-
-def NEONimmAllOnesV: PatLeaf<(ARMvmovImm (i32 timm)), [{
-  ConstantSDNode *ConstVal = cast<ConstantSDNode>(N->getOperand(0));
-  unsigned EltBits = 0;
-  uint64_t EltVal = ARM_AM::decodeVMOVModImm(ConstVal->getZExtValue(), EltBits);
-  return (EltBits == 8 && EltVal == 0xff);
-}]>;
-
 //===----------------------------------------------------------------------===//
 // NEON load / store instructions
 //===----------------------------------------------------------------------===//
@@ -5273,9 +5259,9 @@
 // Vector Bitwise Operations.
 
 def vnotd : PatFrag<(ops node:$in),
-                    (xor node:$in, (bitconvert (v8i8 NEONimmAllOnesV)))>;
+                    (xor node:$in, ARMimmAllOnesD)>;
 def vnotq : PatFrag<(ops node:$in),
-                    (xor node:$in, (bitconvert (v16i8 NEONimmAllOnesV)))>;
+                    (xor node:$in, ARMimmAllOnesV)>;
 
 
 //   VAND     : Vector Bitwise AND
@@ -6054,9 +6040,9 @@
 // Vector Negate.
 
 def vnegd  : PatFrag<(ops node:$in),
-                     (sub (bitconvert (v2i32 NEONimmAllZerosV)), node:$in)>;
+                     (sub ARMimmAllZerosD, node:$in)>;
 def vnegq  : PatFrag<(ops node:$in),
-                     (sub (bitconvert (v4i32 NEONimmAllZerosV)), node:$in)>;
+                     (sub ARMimmAllZerosV, node:$in)>;
 
 class VNEGD<bits<2> size, string OpcodeStr, string Dt, ValueType Ty>
   : N2V<0b11, 0b11, size, 0b01, 0b00111, 0, 0, (outs DPR:$Vd), (ins DPR:$Vm),
@@ -6270,11 +6256,11 @@
 
 let AddedComplexity = 50, isAsCheapAsAMove = 1, isReMaterializable = 1 in {
   def VMOVD0 : ARMPseudoExpand<(outs DPR:$Vd), (ins), 4, IIC_VMOVImm,
-                               [(set DPR:$Vd, (v2i32 NEONimmAllZerosV))],
+                               [(set DPR:$Vd, (v2i32 ARMimmAllZerosD))],
                                (VMOVv2i32 DPR:$Vd, 0, (ops 14, zero_reg))>,
                Requires<[HasZCZ]>;
   def VMOVQ0 : ARMPseudoExpand<(outs QPR:$Vd), (ins), 4, IIC_VMOVImm,
-                               [(set QPR:$Vd, (v4i32 NEONimmAllZerosV))],
+                               [(set QPR:$Vd, (v4i32 ARMimmAllZerosV))],
                                (VMOVv4i32 QPR:$Vd, 0, (ops 14, zero_reg))>,
                Requires<[HasZCZ]>;
 }
Index: llvm/test/CodeGen/Thumb2/mve-pred-selectop3.ll
===================================================================
--- llvm/test/CodeGen/Thumb2/mve-pred-selectop3.ll
+++ llvm/test/CodeGen/Thumb2/mve-pred-selectop3.ll
@@ -4,10 +4,9 @@
 define arm_aapcs_vfpcc <4 x i32> @add_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: add_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vadd.i32 q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i32 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -19,10 +18,9 @@
 define arm_aapcs_vfpcc <8 x i16> @add_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: add_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vadd.i16 q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i16 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -34,10 +32,9 @@
 define arm_aapcs_vfpcc <16 x i8> @add_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: add_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vadd.i8 q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i8 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -49,10 +46,9 @@
 define arm_aapcs_vfpcc <4 x i32> @sub_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: sub_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vsub.i32 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vsubt.i32 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -64,10 +60,9 @@
 define arm_aapcs_vfpcc <8 x i16> @sub_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: sub_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vsub.i16 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vsubt.i16 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -79,10 +74,9 @@
 define arm_aapcs_vfpcc <16 x i8> @sub_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: sub_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vsub.i8 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vsubt.i8 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -94,10 +88,9 @@
 define arm_aapcs_vfpcc <4 x i32> @mul_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: mul_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x1
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vmul.i32 q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i32 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -109,10 +102,9 @@
 define arm_aapcs_vfpcc <8 x i16> @mul_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: mul_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i16 q2, #0x1
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vmul.i16 q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i16 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -124,10 +116,9 @@
 define arm_aapcs_vfpcc <16 x i8> @mul_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: mul_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0x1
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vmul.i8 q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i8 q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -139,10 +130,9 @@
 define arm_aapcs_vfpcc <4 x i32> @and_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: and_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vand q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vandt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -154,10 +144,9 @@
 define arm_aapcs_vfpcc <8 x i16> @and_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: and_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vand q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vandt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -169,10 +158,9 @@
 define arm_aapcs_vfpcc <16 x i8> @and_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: and_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vand q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vandt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -184,10 +172,9 @@
 define arm_aapcs_vfpcc <4 x i32> @or_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: or_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vorr q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vorrt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -199,10 +186,9 @@
 define arm_aapcs_vfpcc <8 x i16> @or_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: or_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vorr q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vorrt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -214,10 +200,9 @@
 define arm_aapcs_vfpcc <16 x i8> @or_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: or_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    vorr q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vorrt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -229,10 +214,9 @@
 define arm_aapcs_vfpcc <4 x i32> @xor_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: xor_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    veor q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    veort q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -244,10 +228,9 @@
 define arm_aapcs_vfpcc <8 x i16> @xor_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: xor_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    veor q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    veort q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -259,10 +242,9 @@
 define arm_aapcs_vfpcc <16 x i8> @xor_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: xor_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q1, q1, q2
-; CHECK-NEXT:    veor q0, q1, q0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    veort q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -274,11 +256,10 @@
 define arm_aapcs_vfpcc <4 x i32> @andnot_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: andnot_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
+; CHECK-NEXT:    vmvn q1, q1
 ; CHECK-NEXT:    vctp.32 r0
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    veort q2, q1, q2
-; CHECK-NEXT:    vand q0, q2, q0
+; CHECK-NEXT:    vandt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -291,11 +272,10 @@
 define arm_aapcs_vfpcc <8 x i16> @andnot_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: andnot_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
+; CHECK-NEXT:    vmvn q1, q1
 ; CHECK-NEXT:    vctp.16 r0
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    veort q2, q1, q2
-; CHECK-NEXT:    vand q0, q2, q0
+; CHECK-NEXT:    vandt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -308,11 +288,10 @@
 define arm_aapcs_vfpcc <16 x i8> @andnot_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: andnot_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
+; CHECK-NEXT:    vmvn q1, q1
 ; CHECK-NEXT:    vctp.8 r0
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    veort q2, q1, q2
-; CHECK-NEXT:    vand q0, q2, q0
+; CHECK-NEXT:    vandt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -325,12 +304,10 @@
 define arm_aapcs_vfpcc <4 x i32> @ornot_v4i32_x(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: ornot_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
-; CHECK-NEXT:    vmov.i8 q3, #0xff
+; CHECK-NEXT:    vmvn q1, q1
 ; CHECK-NEXT:    vctp.32 r0
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    veort q2, q1, q3
-; CHECK-NEXT:    vorr q0, q2, q0
+; CHECK-NEXT:    vorrt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -343,12 +320,10 @@
 define arm_aapcs_vfpcc <8 x i16> @ornot_v8i16_x(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: ornot_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
-; CHECK-NEXT:    vmov.i8 q3, #0xff
+; CHECK-NEXT:    vmvn q1, q1
 ; CHECK-NEXT:    vctp.16 r0
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    veort q2, q1, q3
-; CHECK-NEXT:    vorr q0, q2, q0
+; CHECK-NEXT:    vorrt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -361,12 +336,10 @@
 define arm_aapcs_vfpcc <16 x i8> @ornot_v16i8_x(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: ornot_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
-; CHECK-NEXT:    vmov.i8 q3, #0xff
+; CHECK-NEXT:    vmvn q1, q1
 ; CHECK-NEXT:    vctp.8 r0
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    veort q2, q1, q3
-; CHECK-NEXT:    vorr q0, q2, q0
+; CHECK-NEXT:    vorrt q0, q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -871,11 +844,9 @@
 define arm_aapcs_vfpcc <4 x i32> @addqr_v4i32_x(<4 x i32> %x, i32 %y, i32 %n) {
 ; CHECK-LABEL: addqr_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-NEXT:    vctp.32 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.32 q1, r0
-; CHECK-NEXT:    vadd.i32 q0, q1, q0
+; CHECK-NEXT:    vaddt.i32 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -889,11 +860,9 @@
 define arm_aapcs_vfpcc <8 x i16> @addqr_v8i16_x(<8 x i16> %x, i16 %y, i32 %n) {
 ; CHECK-LABEL: addqr_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-NEXT:    vctp.16 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.16 q1, r0
-; CHECK-NEXT:    vadd.i16 q0, q1, q0
+; CHECK-NEXT:    vaddt.i16 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -907,11 +876,9 @@
 define arm_aapcs_vfpcc <16 x i8> @addqr_v16i8_x(<16 x i8> %x, i8 %y, i32 %n) {
 ; CHECK-LABEL: addqr_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-NEXT:    vctp.8 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.8 q1, r0
-; CHECK-NEXT:    vadd.i8 q0, q1, q0
+; CHECK-NEXT:    vaddt.i8 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -925,11 +892,9 @@
 define arm_aapcs_vfpcc <4 x i32> @subqr_v4i32_x(<4 x i32> %x, i32 %y, i32 %n) {
 ; CHECK-LABEL: subqr_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-NEXT:    vctp.32 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.32 q1, r0
-; CHECK-NEXT:    vsub.i32 q0, q0, q1
+; CHECK-NEXT:    vsubt.i32 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -943,11 +908,9 @@
 define arm_aapcs_vfpcc <8 x i16> @subqr_v8i16_x(<8 x i16> %x, i16 %y, i32 %n) {
 ; CHECK-LABEL: subqr_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-NEXT:    vctp.16 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.16 q1, r0
-; CHECK-NEXT:    vsub.i16 q0, q0, q1
+; CHECK-NEXT:    vsubt.i16 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -961,11 +924,9 @@
 define arm_aapcs_vfpcc <16 x i8> @subqr_v16i8_x(<16 x i8> %x, i8 %y, i32 %n) {
 ; CHECK-LABEL: subqr_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-NEXT:    vctp.8 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.8 q1, r0
-; CHECK-NEXT:    vsub.i8 q0, q0, q1
+; CHECK-NEXT:    vsubt.i8 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -979,11 +940,9 @@
 define arm_aapcs_vfpcc <4 x i32> @mulqr_v4i32_x(<4 x i32> %x, i32 %y, i32 %n) {
 ; CHECK-LABEL: mulqr_v4i32_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x1
 ; CHECK-NEXT:    vctp.32 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.32 q1, r0
-; CHECK-NEXT:    vmul.i32 q0, q1, q0
+; CHECK-NEXT:    vmult.i32 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -997,11 +956,9 @@
 define arm_aapcs_vfpcc <8 x i16> @mulqr_v8i16_x(<8 x i16> %x, i16 %y, i32 %n) {
 ; CHECK-LABEL: mulqr_v8i16_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i16 q1, #0x1
 ; CHECK-NEXT:    vctp.16 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.16 q1, r0
-; CHECK-NEXT:    vmul.i16 q0, q1, q0
+; CHECK-NEXT:    vmult.i16 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -1015,11 +972,9 @@
 define arm_aapcs_vfpcc <16 x i8> @mulqr_v16i8_x(<16 x i8> %x, i8 %y, i32 %n) {
 ; CHECK-LABEL: mulqr_v16i8_x:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q1, #0x1
 ; CHECK-NEXT:    vctp.8 r1
 ; CHECK-NEXT:    vpst
-; CHECK-NEXT:    vdupt.8 q1, r0
-; CHECK-NEXT:    vmul.i8 q0, q1, q0
+; CHECK-NEXT:    vmult.i8 q0, q0, r0
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -1327,10 +1282,10 @@
 define arm_aapcs_vfpcc <4 x i32> @add_v4i32_y(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: add_v4i32_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vadd.i32 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i32 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -1342,10 +1297,10 @@
 define arm_aapcs_vfpcc <8 x i16> @add_v8i16_y(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: add_v8i16_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vadd.i16 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i16 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -1357,10 +1312,10 @@
 define arm_aapcs_vfpcc <16 x i8> @add_v16i8_y(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: add_v16i8_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vadd.i8 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i8 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -1417,10 +1372,10 @@
 define arm_aapcs_vfpcc <4 x i32> @mul_v4i32_y(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: mul_v4i32_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x1
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vmul.i32 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i32 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -1432,10 +1387,10 @@
 define arm_aapcs_vfpcc <8 x i16> @mul_v8i16_y(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: mul_v8i16_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i16 q2, #0x1
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vmul.i16 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i16 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -1447,10 +1402,10 @@
 define arm_aapcs_vfpcc <16 x i8> @mul_v16i8_y(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: mul_v16i8_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0x1
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vmul.i8 q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i8 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -1462,10 +1417,10 @@
 define arm_aapcs_vfpcc <4 x i32> @and_v4i32_y(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: and_v4i32_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vand q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vandt q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -1477,10 +1432,10 @@
 define arm_aapcs_vfpcc <8 x i16> @and_v8i16_y(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: and_v8i16_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vand q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vandt q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -1492,10 +1447,10 @@
 define arm_aapcs_vfpcc <16 x i8> @and_v16i8_y(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: and_v16i8_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q2, #0xff
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vand q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vandt q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -1507,10 +1462,10 @@
 define arm_aapcs_vfpcc <4 x i32> @or_v4i32_y(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: or_v4i32_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vorrt q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -1522,10 +1477,10 @@
 define arm_aapcs_vfpcc <8 x i16> @or_v8i16_y(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: or_v8i16_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vorrt q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -1537,10 +1492,10 @@
 define arm_aapcs_vfpcc <16 x i8> @or_v16i8_y(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: or_v16i8_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vorrt q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -1552,10 +1507,10 @@
 define arm_aapcs_vfpcc <4 x i32> @xor_v4i32_y(<4 x i32> %x, <4 x i32> %y, i32 %n) {
 ; CHECK-LABEL: xor_v4i32_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.32 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    veor q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    veort q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -1567,10 +1522,10 @@
 define arm_aapcs_vfpcc <8 x i16> @xor_v8i16_y(<8 x i16> %x, <8 x i16> %y, i32 %n) {
 ; CHECK-LABEL: xor_v8i16_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.16 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    veor q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    veort q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -1582,10 +1537,10 @@
 define arm_aapcs_vfpcc <16 x i8> @xor_v16i8_y(<16 x i8> %x, <16 x i8> %y, i32 %n) {
 ; CHECK-LABEL: xor_v16i8_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q2, #0x0
 ; CHECK-NEXT:    vctp.8 r0
-; CHECK-NEXT:    vpsel q0, q0, q2
-; CHECK-NEXT:    veor q0, q0, q1
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    veort q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -2219,10 +2174,11 @@
 define arm_aapcs_vfpcc <4 x i32> @addqr_v4i32_y(<4 x i32> %x, i32 %y, i32 %n) {
 ; CHECK-LABEL: addqr_v4i32_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
+; CHECK-NEXT:    vdup.32 q1, r0
 ; CHECK-NEXT:    vctp.32 r1
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    vadd.i32 q0, q0, r0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i32 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -2236,10 +2192,11 @@
 define arm_aapcs_vfpcc <8 x i16> @addqr_v8i16_y(<8 x i16> %x, i16 %y, i32 %n) {
 ; CHECK-LABEL: addqr_v8i16_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
+; CHECK-NEXT:    vdup.16 q1, r0
 ; CHECK-NEXT:    vctp.16 r1
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    vadd.i16 q0, q0, r0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i16 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -2253,10 +2210,11 @@
 define arm_aapcs_vfpcc <16 x i8> @addqr_v16i8_y(<16 x i8> %x, i8 %y, i32 %n) {
 ; CHECK-LABEL: addqr_v16i8_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x0
+; CHECK-NEXT:    vdup.8 q1, r0
 ; CHECK-NEXT:    vctp.8 r1
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    vadd.i8 q0, q0, r0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vaddt.i8 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
@@ -2324,10 +2282,11 @@
 define arm_aapcs_vfpcc <4 x i32> @mulqr_v4i32_y(<4 x i32> %x, i32 %y, i32 %n) {
 ; CHECK-LABEL: mulqr_v4i32_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i32 q1, #0x1
+; CHECK-NEXT:    vdup.32 q1, r0
 ; CHECK-NEXT:    vctp.32 r1
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    vmul.i32 q0, q0, r0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i32 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <4 x i1> @llvm.arm.mve.vctp32(i32 %n)
@@ -2341,10 +2300,11 @@
 define arm_aapcs_vfpcc <8 x i16> @mulqr_v8i16_y(<8 x i16> %x, i16 %y, i32 %n) {
 ; CHECK-LABEL: mulqr_v8i16_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i16 q1, #0x1
+; CHECK-NEXT:    vdup.16 q1, r0
 ; CHECK-NEXT:    vctp.16 r1
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    vmul.i16 q0, q0, r0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i16 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <8 x i1> @llvm.arm.mve.vctp16(i32 %n)
@@ -2358,10 +2318,11 @@
 define arm_aapcs_vfpcc <16 x i8> @mulqr_v16i8_y(<16 x i8> %x, i8 %y, i32 %n) {
 ; CHECK-LABEL: mulqr_v16i8_y:
 ; CHECK:       @ %bb.0: @ %entry
-; CHECK-NEXT:    vmov.i8 q1, #0x1
+; CHECK-NEXT:    vdup.8 q1, r0
 ; CHECK-NEXT:    vctp.8 r1
-; CHECK-NEXT:    vpsel q0, q0, q1
-; CHECK-NEXT:    vmul.i8 q0, q0, r0
+; CHECK-NEXT:    vpst
+; CHECK-NEXT:    vmult.i8 q1, q1, q0
+; CHECK-NEXT:    vmov q0, q1
 ; CHECK-NEXT:    bx lr
 entry:
   %c = call <16 x i1> @llvm.arm.mve.vctp8(i32 %n)
Index: llvm/utils/TableGen/CodeGenDAGPatterns.cpp
===================================================================
--- llvm/utils/TableGen/CodeGenDAGPatterns.cpp
+++ llvm/utils/TableGen/CodeGenDAGPatterns.cpp
@@ -3589,6 +3589,9 @@
   if (Operator->getName() == "null_frag") return true;
   // If any of the arguments reference the null fragment, return true.
   for (unsigned i = 0, e = DI->getNumArgs(); i != e; ++i) {
+    if (auto Arg = dyn_cast<DefInit>(DI->getArg(i)))
+      if (Arg->getDef()->getName() == "null_frag")
+        return true;
     DagInit *Arg = dyn_cast<DagInit>(DI->getArg(i));
     if (Arg && hasNullFragReference(Arg))
       return true;