diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.h b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.h
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.h
@@ -1091,6 +1091,7 @@
   SDValue LowerWindowsDYNAMIC_STACKALLOC(SDValue Op, SDValue Chain,
                                          SDValue &Size,
                                          SelectionDAG &DAG) const;
+  SDValue LowerAVGFloor_AVGCeil(SDValue Node, SelectionDAG &DAG) const;
 
   SDValue LowerFixedLengthVectorIntDivideToSVE(SDValue Op,
                                                SelectionDAG &DAG) const;
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -1288,12 +1288,10 @@
       setOperationAction(ISD::SDIVREM, VT, Expand);
       setOperationAction(ISD::UDIVREM, VT, Expand);
 
-      if (Subtarget->hasSVE2()) {
-        setOperationAction(ISD::AVGFLOORS, VT, Custom);
-        setOperationAction(ISD::AVGFLOORU, VT, Custom);
-        setOperationAction(ISD::AVGCEILS, VT, Custom);
-        setOperationAction(ISD::AVGCEILU, VT, Custom);
-      }
+      setOperationAction(ISD::AVGFLOORS, VT, Custom);
+      setOperationAction(ISD::AVGFLOORU, VT, Custom);
+      setOperationAction(ISD::AVGCEILS, VT, Custom);
+      setOperationAction(ISD::AVGCEILU, VT, Custom);
     }
 
     // Illegal unpacked integer vector types.
@@ -6076,13 +6074,21 @@
   case ISD::ABDU:
     return LowerToPredicatedOp(Op, DAG, AArch64ISD::ABDU_PRED);
   case ISD::AVGFLOORS:
-    return LowerToPredicatedOp(Op, DAG, AArch64ISD::HADDS_PRED);
+    if(Subtarget->hasSVE2())
+      return LowerToPredicatedOp(Op, DAG, AArch64ISD::HADDS_PRED);
+    return LowerAVGFloor_AVGCeil(Op, DAG);
   case ISD::AVGFLOORU:
-    return LowerToPredicatedOp(Op, DAG, AArch64ISD::HADDU_PRED);
+    if(Subtarget->hasSVE2())
+      return LowerToPredicatedOp(Op, DAG, AArch64ISD::HADDU_PRED);
+    return LowerAVGFloor_AVGCeil(Op, DAG);
   case ISD::AVGCEILS:
-    return LowerToPredicatedOp(Op, DAG, AArch64ISD::RHADDS_PRED);
+    if(Subtarget->hasSVE2())
+      return LowerToPredicatedOp(Op, DAG, AArch64ISD::RHADDS_PRED);
+    return LowerAVGFloor_AVGCeil(Op, DAG);
   case ISD::AVGCEILU:
-    return LowerToPredicatedOp(Op, DAG, AArch64ISD::RHADDU_PRED);
+    if(Subtarget->hasSVE2())
+      return LowerToPredicatedOp(Op, DAG, AArch64ISD::RHADDU_PRED);
+    return LowerAVGFloor_AVGCeil(Op, DAG);
   case ISD::BITREVERSE:
     return LowerBitreverse(Op, DAG);
   case ISD::BSWAP:
@@ -13362,6 +13368,34 @@
   return Chain;
 }
 
+SDValue AArch64TargetLowering::LowerAVGFloor_AVGCeil(SDValue Node,
+                                                   SelectionDAG &DAG) const {
+  SDLoc dl(Node);
+  SDValue OpA = Node->getOperand(0);
+  SDValue OpB = Node->getOperand(1);
+  EVT VT = Node->getValueType(0);
+  SDValue ConstantOne = DAG.getConstant(1, dl, VT);
+
+  assert(VT.isScalableVector() && "Only expect to lower scalable vector op!");
+
+  SDValue srlA = DAG.getNode(ISD::SRL, dl, VT, OpA, ConstantOne);
+  SDValue srlB = DAG.getNode(ISD::SRL, dl, VT, OpB, ConstantOne);
+
+  SDValue tmp;
+  if(Node->getOpcode() == ISD::AVGFLOORU ||
+      Node->getOpcode() == ISD::AVGFLOORS) {
+      tmp = DAG.getNode(ISD::AND, dl, VT, OpA, OpB);
+  }
+  else {
+    tmp = DAG.getNode(ISD::OR, dl, VT, OpA, OpB);
+  }
+
+  tmp = DAG.getNode(ISD::AND, dl, VT, tmp, ConstantOne);
+  SDValue Add = DAG.getNode(ISD::ADD, dl, VT, srlA, srlB);
+  return DAG.getNode(ISD::ADD, dl, VT, Add, tmp);
+}
+
+
 SDValue
 AArch64TargetLowering::LowerDYNAMIC_STACKALLOC(SDValue Op,
                                                SelectionDAG &DAG) const {
@@ -17593,7 +17627,8 @@
 }
 
 static SDValue performTruncateCombine(SDNode *N,
-                                      SelectionDAG &DAG) {
+                                      SelectionDAG &DAG,
+                                      const AArch64Subtarget *Subtarget) {
   EVT VT = N->getValueType(0);
   SDValue N0 = N->getOperand(0);
   if (VT.isFixedLengthVector() && VT.is64BitVector() && N0.hasOneUse() &&
@@ -17605,6 +17640,77 @@
     return DAG.getNode(N0.getOpcode(), SDLoc(N), VT, Op);
   }
 
+  // If we see something like trunc( lshr( add(add(a, b), c) , 1) ), where one of (a,b,c) should be 0 or 1 then
+  // we can convert that into lshr(a, 1) + lshr(b, 1) + (a|b)&1
+  if(!Subtarget->hasSVE2() && N->getOperand(0).getOpcode() == ISD::SRL) {
+    SDValue Op0 = N->getOperand(0).getOperand(0); //add(add(a, b), c)
+    SDValue Op1 = N->getOperand(0).getOperand(1); //shiftAmount
+
+    if(Op0.getOpcode() != ISD::ADD)
+      return SDValue();
+
+    if(!isOneConstant(Op1))
+      return SDValue();
+
+    if(Op0.getOperand(0).getOpcode() != ISD::ADD) // add(a, b)
+      return SDValue();
+
+    SDValue A, B, C;
+    C = Op0.getOperand(1);
+    A = Op0.getOperand(0).getOperand(0); // a
+    B = Op0.getOperand(0).getOperand(1); // b
+
+    // we have add(add(a, b), c),
+    // one of a or b or c should be a constant 0 or 1:
+    // usually C is the constant, but in some cases, A or B may be the constant.
+    // check if any A or B is the constants, and then swap them with C
+    if(isNullConstant(A) || isOneConstant(A)) {
+      SDValue tmp = A;
+      A = C;
+      C = tmp;
+    }
+    else if(isNullConstant(B) || isOneConstant(B)) {
+      SDValue tmp = B;
+      B = C;
+      C = tmp;
+    }
+    else if(!isNullConstant(C) && !isOneConstant(C))
+      // no constant found
+      return SDValue();
+
+    if(A.getOpcode() == ISD::ZERO_EXTEND)
+      A = A.getOperand(0);
+
+    if(B.getOpcode() == ISD::ZERO_EXTEND)
+      B = B.getOperand(0);
+
+    // (A >> 1)
+    SDValue AShifted = DAG.getNode(ISD::SRL, SDLoc(N), A.getValueType(), A, Op1);
+    // (B >> 1)
+    SDValue BShifted = DAG.getNode(ISD::SRL, SDLoc(N), B.getValueType(), B, Op1);
+
+    SDValue AB;
+    if(isNullConstant(C)) {
+    // (A & B)
+      AB = DAG.getNode(ISD::AND, SDLoc(N), A.getValueType(), A, B);
+    }
+    else {
+    // (A | B)
+      AB = DAG.getNode(ISD::OR, SDLoc(N), A.getValueType(), A, B);
+    }
+
+    // (A & B)&1
+    // (A | B)&1
+    SDValue ABndOne = DAG.getNode(ISD::AND, SDLoc(N), A.getValueType(), AB, DAG.getConstant(1, SDLoc(N), AB.getValueType()));
+
+    // AShifted + BShifted
+    SDValue Add1 = DAG.getNode(ISD::ADD, SDLoc(N), AShifted.getValueType(), AShifted, BShifted);
+    // (AShifted + BShifted) + ABndOne
+    SDValue newInstr = DAG.getNode(ISD::ADD, SDLoc(N), AShifted.getValueType(), Add1, ABndOne);
+
+    return newInstr;
+  }
+
   return SDValue();
 }
 
@@ -18489,7 +18595,8 @@
 
 static SDValue performExtendCombine(SDNode *N,
                                     TargetLowering::DAGCombinerInfo &DCI,
-                                    SelectionDAG &DAG) {
+                                    SelectionDAG &DAG,
+                                    const AArch64Subtarget *Subtarget) {
   // If we see something like (zext (sabd (extract_high ...), (DUP ...))) then
   // we can convert that DUP into another extract_high (of a bigger DUP), which
   // helps the backend to decide that an sabdl2 would be useful, saving a real
@@ -18511,6 +18618,66 @@
       N->getOperand(0)->getOpcode() == ISD::SETCC)
     return performSignExtendSetCCCombine(N, DCI, DAG);
 
+  // If we see something like zext( lshr( add(add(a, b), 1) , 1) ) then
+  // we can convert that into lshr(a, 1) + lshr(b, 1) + (a|b)&1
+  if(!Subtarget->hasSVE2() && N->getOperand(0).getOpcode() == ISD::SRL) {
+    SDValue Op0 = N->getOperand(0).getOperand(0); //add(add(a, b), 1)
+    SDValue Op1 = N->getOperand(0).getOperand(1); //shiftAmount
+    if(Op0.getOpcode() != ISD::ADD || !isOneConstant(Op1))
+      return SDValue();
+
+    if(Op0.getOperand(0).getOpcode() != ISD::ADD) // add(a, b)
+      return SDValue();
+
+    SDValue AddedValue = Op0.getOperand(1);
+    // AddedValue must be 0 or 1
+    if(!isNullConstant(AddedValue) && !isOneConstant(AddedValue))
+      return SDValue();
+
+    SDValue A, B;
+    if(Op0.getOperand(0).getOperand(0).getOpcode() == ISD::TRUNCATE)
+      A = Op0.getOperand(0).getOperand(0).getOperand(0); // a
+    else
+      A = Op0.getOperand(0).getOperand(0); // a
+
+    if(Op0.getOperand(0).getOperand(1).getOpcode() == ISD::TRUNCATE)
+      B = Op0.getOperand(0).getOperand(1).getOperand(0); // b
+    else
+      B = Op0.getOperand(0).getOperand(1); // b
+
+    // (A >> 1)
+    SDValue AShifted = DAG.getNode(ISD::SRL, SDLoc(N), A.getValueType(), A, Op1);
+    // (B >> 1)
+    SDValue BShifted = DAG.getNode(ISD::SRL, SDLoc(N), B.getValueType(), B, Op1);
+
+    SDValue AB;
+    if(isNullConstant(AddedValue)) {
+    // (A & B)
+      AB = DAG.getNode(ISD::AND, SDLoc(N), A.getValueType(), A, B);
+    }
+    else {
+    // (A | B)
+      AB = DAG.getNode(ISD::OR, SDLoc(N), A.getValueType(), A, B);
+    }
+
+    // (A & B)&1
+    // (A | B)&1
+    SDValue ABndOne = DAG.getNode(ISD::AND, SDLoc(N), A.getValueType(), AB, DAG.getConstant(1, SDLoc(N), AB.getValueType()));
+
+    // AShifted + BShifted
+    SDValue Add1 = DAG.getNode(ISD::ADD, SDLoc(N), AShifted.getValueType(), AShifted, BShifted);
+    // (AShifted + BShifted) + ABndOne
+    SDValue newInstr = DAG.getNode(ISD::ADD, SDLoc(N), AShifted.getValueType(), Add1, ABndOne);
+
+    if(N->getValueType(0).getSizeInBits() == newInstr.getValueSizeInBits())
+      return newInstr;
+
+    else if(N->getValueType(0).getSizeInBits() > newInstr.getValueSizeInBits())
+      return DAG.getNode(ISD::ZERO_EXTEND, SDLoc(N), N->getValueType(0), newInstr);
+
+    return DAG.getNode(ISD::TRUNCATE, SDLoc(N), N->getValueType(0), newInstr);
+  }
+
   return SDValue();
 }
 
@@ -21497,7 +21664,7 @@
   case ISD::BUILD_VECTOR:
     return performBuildVectorCombine(N, DCI, DAG);
   case ISD::TRUNCATE:
-    return performTruncateCombine(N, DAG);
+    return performTruncateCombine(N, DAG, Subtarget);
   case AArch64ISD::ANDS:
     return performFlagSettingCombine(N, DCI, ISD::AND);
   case AArch64ISD::ADC:
@@ -21537,7 +21704,7 @@
   case ISD::ANY_EXTEND:
   case ISD::ZERO_EXTEND:
   case ISD::SIGN_EXTEND:
-    return performExtendCombine(N, DCI, DAG);
+    return performExtendCombine(N, DCI, DAG, Subtarget);
   case ISD::SIGN_EXTEND_INREG:
     return performSignExtendInRegCombine(N, DCI, DAG);
   case ISD::CONCAT_VECTORS:
diff --git a/llvm/test/CodeGen/AArch64/neon-lshr.ll b/llvm/test/CodeGen/AArch64/neon-lshr.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/neon-lshr.ll
@@ -0,0 +1,23 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=aarch64-none-linux-gnu -mattr=+neon | FileCheck %s
+
+define i8 @lshr_trunc(i8 %a, i8%b)
+; CHECK-LABEL: lshr_trunc:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    and w8, w1, #0xfe
+; CHECK-NEXT:    ubfx w9, w0, #1, #7
+; CHECK-NEXT:    orr w10, w0, w1
+; CHECK-NEXT:    add w8, w9, w8, lsr #1
+; CHECK-NEXT:    and w9, w10, #0x1
+; CHECK-NEXT:    add w0, w8, w9
+; CHECK-NEXT:    ret
+{
+  %zexta = zext i8 %a to i16
+  %zextb = zext i8 %b to i16
+  %zextab = add i16 %zexta, %zextb
+  %add = add i16 %zextab, 1
+  %shift = lshr i16 %add, 1
+  %trunca = trunc i16 %shift to i8
+  ret i8 %trunca
+}
+
diff --git a/llvm/test/CodeGen/AArch64/sve-avg_floor_ceil.ll b/llvm/test/CodeGen/AArch64/sve-avg_floor_ceil.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sve-avg_floor_ceil.ll
@@ -0,0 +1,723 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=aarch64-none-linux-gnu  -mcpu=neoverse-v1 | FileCheck %s
+
+define <vscale x 2 x i64> @hadds_v2i64(<vscale x 2 x i64> %s0, <vscale x 2 x i64> %s1) {
+; CHECK-LABEL: hadds_v2i64:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.d, z1.d, #1
+; CHECK-NEXT:    lsr z3.d, z0.d, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.d, z3.d, z2.d
+; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    add z0.d, z2.d, z0.d
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 2 x i64> %s0 to <vscale x 2 x i128>
+  %s1s = sext <vscale x 2 x i64> %s1 to <vscale x 2 x i128>
+  %m = add nsw <vscale x 2 x i128> %s0s, %s1s
+  %s = lshr <vscale x 2 x i128> %m, shufflevector (<vscale x 2 x i128> insertelement (<vscale x 2 x i128> poison, i128 1, i32 0), <vscale x 2 x i128> poison, <vscale x 2 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 2 x i128> %s to <vscale x 2 x i64>
+  ret <vscale x 2 x i64> %s2
+}
+
+define <vscale x 2 x i64> @haddu_v2i64(<vscale x 2 x i64> %s0, <vscale x 2 x i64> %s1) {
+; CHECK-LABEL: haddu_v2i64:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.d, z1.d, #1
+; CHECK-NEXT:    lsr z3.d, z0.d, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.d, z3.d, z2.d
+; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    add z0.d, z2.d, z0.d
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 2 x i64> %s0 to <vscale x 2 x i128>
+  %s1s = zext <vscale x 2 x i64> %s1 to <vscale x 2 x i128>
+  %m = add nuw nsw <vscale x 2 x i128> %s0s, %s1s
+  %s = lshr <vscale x 2 x i128> %m, shufflevector (<vscale x 2 x i128> insertelement (<vscale x 2 x i128> poison, i128 1, i32 0), <vscale x 2 x i128> poison, <vscale x 2 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 2 x i128> %s to <vscale x 2 x i64>
+  ret <vscale x 2 x i64> %s2
+}
+
+define <vscale x 2 x i32> @hadds_v2i32(<vscale x 2 x i32> %s0, <vscale x 2 x i32> %s1) {
+; CHECK-LABEL: hadds_v2i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    sxtw z0.d, p0/m, z0.d
+; CHECK-NEXT:    adr z0.d, [z0.d, z1.d, sxtw]
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 2 x i32> %s0 to <vscale x 2 x i64>
+  %s1s = sext <vscale x 2 x i32> %s1 to <vscale x 2 x i64>
+  %m = add nsw <vscale x 2 x i64> %s0s, %s1s
+  %s = lshr <vscale x 2 x i64> %m, shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 2 x i64> %s to <vscale x 2 x i32>
+  ret <vscale x 2 x i32> %s2
+}
+
+define <vscale x 2 x i32> @haddu_v2i32(<vscale x 2 x i32> %s0, <vscale x 2 x i32> %s1) {
+; CHECK-LABEL: haddu_v2i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    and z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    and z1.d, z1.d, #0xffffffff
+; CHECK-NEXT:    lsr z1.d, z1.d, #1
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    and z2.d, z2.d, #0x1
+; CHECK-NEXT:    add z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z0.d, z0.d, z2.d
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 2 x i32> %s0 to <vscale x 2 x i64>
+  %s1s = zext <vscale x 2 x i32> %s1 to <vscale x 2 x i64>
+  %m = add nuw nsw <vscale x 2 x i64> %s0s, %s1s
+  %s = lshr <vscale x 2 x i64> %m, shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 2 x i64> %s to <vscale x 2 x i32>
+  ret <vscale x 2 x i32> %s2
+}
+
+define <vscale x 4 x i32> @hadds_v4i32(<vscale x 4 x i32> %s0, <vscale x 4 x i32> %s1) {
+; CHECK-LABEL: hadds_v4i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.s, z1.s, #1
+; CHECK-NEXT:    lsr z3.s, z0.s, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.s, z3.s, z2.s
+; CHECK-NEXT:    and z0.s, z0.s, #0x1
+; CHECK-NEXT:    add z0.s, z2.s, z0.s
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 4 x i32> %s0 to <vscale x 4 x i64>
+  %s1s = sext <vscale x 4 x i32> %s1 to <vscale x 4 x i64>
+  %m = add nsw <vscale x 4 x i64> %s0s, %s1s
+  %s = lshr <vscale x 4 x i64> %m, shufflevector (<vscale x 4 x i64> insertelement (<vscale x 4 x i64> poison, i64 1, i32 0), <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 4 x i64> %s to <vscale x 4 x i32>
+  ret <vscale x 4 x i32> %s2
+}
+
+define <vscale x 4 x i32> @haddu_v4i32(<vscale x 4 x i32> %s0, <vscale x 4 x i32> %s1) {
+; CHECK-LABEL: haddu_v4i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.s, z1.s, #1
+; CHECK-NEXT:    lsr z3.s, z0.s, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.s, z3.s, z2.s
+; CHECK-NEXT:    and z0.s, z0.s, #0x1
+; CHECK-NEXT:    add z0.s, z2.s, z0.s
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 4 x i32> %s0 to <vscale x 4 x i64>
+  %s1s = zext <vscale x 4 x i32> %s1 to <vscale x 4 x i64>
+  %m = add nuw nsw <vscale x 4 x i64> %s0s, %s1s
+  %s = lshr <vscale x 4 x i64> %m, shufflevector (<vscale x 4 x i64> insertelement (<vscale x 4 x i64> poison, i64 1, i32 0), <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 4 x i64> %s to <vscale x 4 x i32>
+  ret <vscale x 4 x i32> %s2
+}
+
+define <vscale x 2 x i16> @hadds_v2i16(<vscale x 2 x i16> %s0, <vscale x 2 x i16> %s1) {
+; CHECK-LABEL: hadds_v2i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    sxth z0.d, p0/m, z0.d
+; CHECK-NEXT:    sxth z1.d, p0/m, z1.d
+; CHECK-NEXT:    add z0.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 2 x i16> %s0 to <vscale x 2 x i32>
+  %s1s = sext <vscale x 2 x i16> %s1 to <vscale x 2 x i32>
+  %m = add nsw <vscale x 2 x i32> %s0s, %s1s
+  %s = lshr <vscale x 2 x i32> %m, shufflevector (<vscale x 2 x i32> insertelement (<vscale x 2 x i32> poison, i32 1, i32 0), <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 2 x i32> %s to <vscale x 2 x i16>
+  ret <vscale x 2 x i16> %s2
+}
+
+define <vscale x 2 x i16> @haddu_v2i16(<vscale x 2 x i16> %s0, <vscale x 2 x i16> %s1) {
+; CHECK-LABEL: haddu_v2i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    and z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.d, z0.d, #0xffff
+; CHECK-NEXT:    and z1.d, z1.d, #0xffff
+; CHECK-NEXT:    lsr z1.d, z1.d, #1
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    and z2.d, z2.d, #0x1
+; CHECK-NEXT:    add z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z0.d, z0.d, z2.d
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 2 x i16> %s0 to <vscale x 2 x i32>
+  %s1s = zext <vscale x 2 x i16> %s1 to <vscale x 2 x i32>
+  %m = add nuw nsw <vscale x 2 x i32> %s0s, %s1s
+  %s = lshr <vscale x 2 x i32> %m, shufflevector (<vscale x 2 x i32> insertelement (<vscale x 2 x i32> poison, i32 1, i32 0), <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 2 x i32> %s to <vscale x 2 x i16>
+  ret <vscale x 2 x i16> %s2
+}
+
+define <vscale x 4 x i16> @hadds_v4i16(<vscale x 4 x i16> %s0, <vscale x 4 x i16> %s1) {
+; CHECK-LABEL: hadds_v4i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    sxth z0.s, p0/m, z0.s
+; CHECK-NEXT:    sxth z1.s, p0/m, z1.s
+; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 4 x i16> %s0 to <vscale x 4 x i32>
+  %s1s = sext <vscale x 4 x i16> %s1 to <vscale x 4 x i32>
+  %m = add nsw <vscale x 4 x i32> %s0s, %s1s
+  %s = lshr <vscale x 4 x i32> %m, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i32 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 4 x i32> %s to <vscale x 4 x i16>
+  ret <vscale x 4 x i16> %s2
+}
+
+define <vscale x 4 x i16> @haddu_v4i16(<vscale x 4 x i16> %s0, <vscale x 4 x i16> %s1) {
+; CHECK-LABEL: haddu_v4i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    and z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.s, z0.s, #0xffff
+; CHECK-NEXT:    and z1.s, z1.s, #0xffff
+; CHECK-NEXT:    lsr z1.s, z1.s, #1
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    and z2.s, z2.s, #0x1
+; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    add z0.s, z0.s, z2.s
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 4 x i16> %s0 to <vscale x 4 x i32>
+  %s1s = zext <vscale x 4 x i16> %s1 to <vscale x 4 x i32>
+  %m = add nuw nsw <vscale x 4 x i32> %s0s, %s1s
+  %s = lshr <vscale x 4 x i32> %m, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i32 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 4 x i32> %s to <vscale x 4 x i16>
+  ret <vscale x 4 x i16> %s2
+}
+
+define <vscale x 8 x i16> @hadds_v8i16(<vscale x 8 x i16> %s0, <vscale x 8 x i16> %s1) {
+; CHECK-LABEL: hadds_v8i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.h, z1.h, #1
+; CHECK-NEXT:    lsr z3.h, z0.h, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.h, z3.h, z2.h
+; CHECK-NEXT:    and z0.h, z0.h, #0x1
+; CHECK-NEXT:    add z0.h, z2.h, z0.h
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 8 x i16> %s0 to <vscale x 8 x i32>
+  %s1s = sext <vscale x 8 x i16> %s1 to <vscale x 8 x i32>
+  %m = add nsw <vscale x 8 x i32> %s0s, %s1s
+  %s = lshr <vscale x 8 x i32> %m, shufflevector (<vscale x 8 x i32> insertelement (<vscale x 8 x i32> poison, i32 1, i32 0), <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 8 x i32> %s to <vscale x 8 x i16>
+  ret <vscale x 8 x i16> %s2
+}
+
+define <vscale x 8 x i16> @haddu_v8i16(<vscale x 8 x i16> %s0, <vscale x 8 x i16> %s1) {
+; CHECK-LABEL: haddu_v8i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.h, z1.h, #1
+; CHECK-NEXT:    lsr z3.h, z0.h, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.h, z3.h, z2.h
+; CHECK-NEXT:    and z0.h, z0.h, #0x1
+; CHECK-NEXT:    add z0.h, z2.h, z0.h
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 8 x i16> %s0 to <vscale x 8 x i32>
+  %s1s = zext <vscale x 8 x i16> %s1 to <vscale x 8 x i32>
+  %m = add nuw nsw <vscale x 8 x i32> %s0s, %s1s
+  %s = lshr <vscale x 8 x i32> %m, shufflevector (<vscale x 8 x i32> insertelement (<vscale x 8 x i32> poison, i32 1, i32 0), <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 8 x i32> %s to <vscale x 8 x i16>
+  ret <vscale x 8 x i16> %s2
+}
+
+define <vscale x 4 x i8> @hadds_v4i8(<vscale x 4 x i8> %s0, <vscale x 4 x i8> %s1) {
+; CHECK-LABEL: hadds_v4i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    sxtb z0.s, p0/m, z0.s
+; CHECK-NEXT:    sxtb z1.s, p0/m, z1.s
+; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    and z0.s, z0.s, #0xffff
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 4 x i8> %s0 to <vscale x 4 x i16>
+  %s1s = sext <vscale x 4 x i8> %s1 to <vscale x 4 x i16>
+  %m = add nsw <vscale x 4 x i16> %s0s, %s1s
+  %s = lshr <vscale x 4 x i16> %m, shufflevector (<vscale x 4 x i16> insertelement (<vscale x 4 x i16> poison, i16 1, i32 0), <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 4 x i16> %s to <vscale x 4 x i8>
+  ret <vscale x 4 x i8> %s2
+}
+
+define <vscale x 4 x i8> @haddu_v4i8(<vscale x 4 x i8> %s0, <vscale x 4 x i8> %s1) {
+; CHECK-LABEL: haddu_v4i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    and z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.s, z0.s, #0xff
+; CHECK-NEXT:    and z1.s, z1.s, #0xff
+; CHECK-NEXT:    lsr z1.s, z1.s, #1
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    and z2.s, z2.s, #0x1
+; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    add z0.s, z0.s, z2.s
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 4 x i8> %s0 to <vscale x 4 x i16>
+  %s1s = zext <vscale x 4 x i8> %s1 to <vscale x 4 x i16>
+  %m = add nuw nsw <vscale x 4 x i16> %s0s, %s1s
+  %s = lshr <vscale x 4 x i16> %m, shufflevector (<vscale x 4 x i16> insertelement (<vscale x 4 x i16> poison, i16 1, i32 0), <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 4 x i16> %s to <vscale x 4 x i8>
+  ret <vscale x 4 x i8> %s2
+}
+
+define <vscale x 8 x i8> @hadds_v8i8(<vscale x 8 x i8> %s0, <vscale x 8 x i8> %s1) {
+; CHECK-LABEL: hadds_v8i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    sxtb z0.h, p0/m, z0.h
+; CHECK-NEXT:    sxtb z1.h, p0/m, z1.h
+; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    lsr z0.h, z0.h, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 8 x i8> %s0 to <vscale x 8 x i16>
+  %s1s = sext <vscale x 8 x i8> %s1 to <vscale x 8 x i16>
+  %m = add nsw <vscale x 8 x i16> %s0s, %s1s
+  %s = lshr <vscale x 8 x i16> %m, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 1, i32 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 8 x i16> %s to <vscale x 8 x i8>
+  ret <vscale x 8 x i8> %s2
+}
+
+define <vscale x 8 x i8> @haddu_v8i8(<vscale x 8 x i8> %s0, <vscale x 8 x i8> %s1) {
+; CHECK-LABEL: haddu_v8i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    and z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.h, z0.h, #0xff
+; CHECK-NEXT:    and z1.h, z1.h, #0xff
+; CHECK-NEXT:    lsr z1.h, z1.h, #1
+; CHECK-NEXT:    lsr z0.h, z0.h, #1
+; CHECK-NEXT:    and z2.h, z2.h, #0x1
+; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    add z0.h, z0.h, z2.h
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 8 x i8> %s0 to <vscale x 8 x i16>
+  %s1s = zext <vscale x 8 x i8> %s1 to <vscale x 8 x i16>
+  %m = add nuw nsw <vscale x 8 x i16> %s0s, %s1s
+  %s = lshr <vscale x 8 x i16> %m, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 1, i32 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 8 x i16> %s to <vscale x 8 x i8>
+  ret <vscale x 8 x i8> %s2
+}
+
+define <vscale x 16 x i8> @hadds_v16i8(<vscale x 16 x i8> %s0, <vscale x 16 x i8> %s1) {
+; CHECK-LABEL: hadds_v16i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.b, z1.b, #1
+; CHECK-NEXT:    lsr z3.b, z0.b, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.b, z3.b, z2.b
+; CHECK-NEXT:    and z0.b, z0.b, #0x1
+; CHECK-NEXT:    add z0.b, z2.b, z0.b
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 16 x i8> %s0 to <vscale x 16 x i16>
+  %s1s = sext <vscale x 16 x i8> %s1 to <vscale x 16 x i16>
+  %m = add nsw <vscale x 16 x i16> %s0s, %s1s
+  %s = lshr <vscale x 16 x i16> %m, shufflevector (<vscale x 16 x i16> insertelement (<vscale x 16 x i16> poison, i16 1, i32 0), <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 16 x i16> %s to <vscale x 16 x i8>
+  ret <vscale x 16 x i8> %s2
+}
+
+define <vscale x 16 x i8> @haddu_v16i8(<vscale x 16 x i8> %s0, <vscale x 16 x i8> %s1) {
+; CHECK-LABEL: haddu_v16i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.b, z1.b, #1
+; CHECK-NEXT:    lsr z3.b, z0.b, #1
+; CHECK-NEXT:    and z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.b, z3.b, z2.b
+; CHECK-NEXT:    and z0.b, z0.b, #0x1
+; CHECK-NEXT:    add z0.b, z2.b, z0.b
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 16 x i8> %s0 to <vscale x 16 x i16>
+  %s1s = zext <vscale x 16 x i8> %s1 to <vscale x 16 x i16>
+  %m = add nuw nsw <vscale x 16 x i16> %s0s, %s1s
+  %s = lshr <vscale x 16 x i16> %m, shufflevector (<vscale x 16 x i16> insertelement (<vscale x 16 x i16> poison, i16 1, i32 0), <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer)
+  %s2 = trunc <vscale x 16 x i16> %s to <vscale x 16 x i8>
+  ret <vscale x 16 x i8> %s2
+}
+
+define <vscale x 2 x i64> @rhadds_v2i64(<vscale x 2 x i64> %s0, <vscale x 2 x i64> %s1) {
+; CHECK-LABEL: rhadds_v2i64:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.d, z1.d, #1
+; CHECK-NEXT:    lsr z3.d, z0.d, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.d, z3.d, z2.d
+; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    add z0.d, z2.d, z0.d
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 2 x i64> %s0 to <vscale x 2 x i128>
+  %s1s = sext <vscale x 2 x i64> %s1 to <vscale x 2 x i128>
+  %add = add <vscale x 2 x i128> %s0s, shufflevector (<vscale x 2 x i128> insertelement (<vscale x 2 x i128> poison, i128 1, i32 0), <vscale x 2 x i128> poison, <vscale x 2 x i32> zeroinitializer)
+  %add2 = add <vscale x 2 x i128> %add, %s1s
+  %s = lshr <vscale x 2 x i128> %add2, shufflevector (<vscale x 2 x i128> insertelement (<vscale x 2 x i128> poison, i128 1, i32 0), <vscale x 2 x i128> poison, <vscale x 2 x i32> zeroinitializer)
+  %result = trunc <vscale x 2 x i128> %s to <vscale x 2 x i64>
+  ret <vscale x 2 x i64> %result
+}
+
+define <vscale x 2 x i64> @rhaddu_v2i64(<vscale x 2 x i64> %s0, <vscale x 2 x i64> %s1) {
+; CHECK-LABEL: rhaddu_v2i64:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.d, z1.d, #1
+; CHECK-NEXT:    lsr z3.d, z0.d, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.d, z3.d, z2.d
+; CHECK-NEXT:    and z0.d, z0.d, #0x1
+; CHECK-NEXT:    add z0.d, z2.d, z0.d
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 2 x i64> %s0 to <vscale x 2 x i128>
+  %s1s = zext <vscale x 2 x i64> %s1 to <vscale x 2 x i128>
+  %add = add nuw nsw <vscale x 2 x i128> %s0s, shufflevector (<vscale x 2 x i128> insertelement (<vscale x 2 x i128> poison, i128 1, i32 0), <vscale x 2 x i128> poison, <vscale x 2 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 2 x i128> %add, %s1s
+  %s = lshr <vscale x 2 x i128> %add2, shufflevector (<vscale x 2 x i128> insertelement (<vscale x 2 x i128> poison, i128 1, i32 0), <vscale x 2 x i128> poison, <vscale x 2 x i32> zeroinitializer)
+  %result = trunc <vscale x 2 x i128> %s to <vscale x 2 x i64>
+  ret <vscale x 2 x i64> %result
+}
+
+define <vscale x 2 x i32> @rhadds_v2i32(<vscale x 2 x i32> %s0, <vscale x 2 x i32> %s1) {
+; CHECK-LABEL: rhadds_v2i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z2.d, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    sxtw z0.d, p0/m, z0.d
+; CHECK-NEXT:    sxtw z1.d, p0/m, z1.d
+; CHECK-NEXT:    eor z0.d, z0.d, z2.d
+; CHECK-NEXT:    sub z0.d, z1.d, z0.d
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 2 x i32> %s0 to <vscale x 2 x i64>
+  %s1s = sext <vscale x 2 x i32> %s1 to <vscale x 2 x i64>
+  %add = add <vscale x 2 x i64> %s0s, shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+  %add2 = add <vscale x 2 x i64> %add, %s1s
+  %s = lshr <vscale x 2 x i64> %add2, shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+  %result = trunc <vscale x 2 x i64> %s to <vscale x 2 x i32>
+  ret <vscale x 2 x i32> %result
+}
+
+define <vscale x 2 x i32> @rhaddu_v2i32(<vscale x 2 x i32> %s0, <vscale x 2 x i32> %s1) {
+; CHECK-LABEL: rhaddu_v2i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    orr z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    and z1.d, z1.d, #0xffffffff
+; CHECK-NEXT:    lsr z1.d, z1.d, #1
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    and z2.d, z2.d, #0x1
+; CHECK-NEXT:    add z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z0.d, z0.d, z2.d
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 2 x i32> %s0 to <vscale x 2 x i64>
+  %s1s = zext <vscale x 2 x i32> %s1 to <vscale x 2 x i64>
+  %add = add nuw nsw <vscale x 2 x i64> %s0s, shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 2 x i64> %add, %s1s
+  %s = lshr <vscale x 2 x i64> %add2, shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i32 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+  %result = trunc <vscale x 2 x i64> %s to <vscale x 2 x i32>
+  ret <vscale x 2 x i32> %result
+}
+
+define <vscale x 4 x i32> @rhadds_v4i32(<vscale x 4 x i32> %s0, <vscale x 4 x i32> %s1) {
+; CHECK-LABEL: rhadds_v4i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.s, z1.s, #1
+; CHECK-NEXT:    lsr z3.s, z0.s, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.s, z3.s, z2.s
+; CHECK-NEXT:    and z0.s, z0.s, #0x1
+; CHECK-NEXT:    add z0.s, z2.s, z0.s
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 4 x i32> %s0 to <vscale x 4 x i64>
+  %s1s = sext <vscale x 4 x i32> %s1 to <vscale x 4 x i64>
+  %add = add <vscale x 4 x i64> %s0s, shufflevector (<vscale x 4 x i64> insertelement (<vscale x 4 x i64> poison, i64 1, i32 0), <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer)
+  %add2 = add <vscale x 4 x i64> %add, %s1s
+  %s = lshr <vscale x 4 x i64> %add2, shufflevector (<vscale x 4 x i64> insertelement (<vscale x 4 x i64> poison, i64 1, i32 0), <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer)
+  %result = trunc <vscale x 4 x i64> %s to <vscale x 4 x i32>
+  ret <vscale x 4 x i32> %result
+}
+
+define <vscale x 4 x i32> @rhaddu_v4i32(<vscale x 4 x i32> %s0, <vscale x 4 x i32> %s1) {
+; CHECK-LABEL: rhaddu_v4i32:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.s, z1.s, #1
+; CHECK-NEXT:    lsr z3.s, z0.s, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.s, z3.s, z2.s
+; CHECK-NEXT:    and z0.s, z0.s, #0x1
+; CHECK-NEXT:    add z0.s, z2.s, z0.s
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 4 x i32> %s0 to <vscale x 4 x i64>
+  %s1s = zext <vscale x 4 x i32> %s1 to <vscale x 4 x i64>
+  %add = add nuw nsw <vscale x 4 x i64> %s0s, shufflevector (<vscale x 4 x i64> insertelement (<vscale x 4 x i64> poison, i64 1, i32 0), <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 4 x i64> %add, %s1s
+  %s = lshr <vscale x 4 x i64> %add2, shufflevector (<vscale x 4 x i64> insertelement (<vscale x 4 x i64> poison, i64 1, i32 0), <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer)
+  %result = trunc <vscale x 4 x i64> %s to <vscale x 4 x i32>
+  ret <vscale x 4 x i32> %result
+}
+
+define <vscale x 2 x i16> @rhadds_v2i16(<vscale x 2 x i16> %s0, <vscale x 2 x i16> %s1) {
+; CHECK-LABEL: rhadds_v2i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.d
+; CHECK-NEXT:    mov z2.d, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    sxth z0.d, p0/m, z0.d
+; CHECK-NEXT:    sxth z1.d, p0/m, z1.d
+; CHECK-NEXT:    eor z0.d, z0.d, z2.d
+; CHECK-NEXT:    sub z0.d, z1.d, z0.d
+; CHECK-NEXT:    and z0.d, z0.d, #0xffffffff
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 2 x i16> %s0 to <vscale x 2 x i32>
+  %s1s = sext <vscale x 2 x i16> %s1 to <vscale x 2 x i32>
+  %add = add <vscale x 2 x i32> %s0s, shufflevector (<vscale x 2 x i32> insertelement (<vscale x 2 x i32> poison, i32 1, i32 0), <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer)
+  %add2 = add <vscale x 2 x i32> %add, %s1s
+  %s = lshr <vscale x 2 x i32> %add2, shufflevector (<vscale x 2 x i32> insertelement (<vscale x 2 x i32> poison, i32 1, i32 0), <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer)
+  %result = trunc <vscale x 2 x i32> %s to <vscale x 2 x i16>
+  ret <vscale x 2 x i16> %result
+}
+
+define <vscale x 2 x i16> @rhaddu_v2i16(<vscale x 2 x i16> %s0, <vscale x 2 x i16> %s1) {
+; CHECK-LABEL: rhaddu_v2i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    and z0.d, z0.d, #0xffff
+; CHECK-NEXT:    mov z2.d, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    and z1.d, z1.d, #0xffff
+; CHECK-NEXT:    eor z0.d, z0.d, z2.d
+; CHECK-NEXT:    sub z0.d, z1.d, z0.d
+; CHECK-NEXT:    lsr z0.d, z0.d, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 2 x i16> %s0 to <vscale x 2 x i32>
+  %s1s = zext <vscale x 2 x i16> %s1 to <vscale x 2 x i32>
+  %add = add nuw nsw <vscale x 2 x i32> %s0s, shufflevector (<vscale x 2 x i32> insertelement (<vscale x 2 x i32> poison, i32 1, i32 0), <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 2 x i32> %add, %s1s
+  %s = lshr <vscale x 2 x i32> %add2, shufflevector (<vscale x 2 x i32> insertelement (<vscale x 2 x i32> poison, i32 1, i32 0), <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer)
+  %result = trunc <vscale x 2 x i32> %s to <vscale x 2 x i16>
+  ret <vscale x 2 x i16> %result
+}
+
+define <vscale x 4 x i16> @rhadds_v4i16(<vscale x 4 x i16> %s0, <vscale x 4 x i16> %s1) {
+; CHECK-LABEL: rhadds_v4i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    mov z2.s, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    sxth z0.s, p0/m, z0.s
+; CHECK-NEXT:    sxth z1.s, p0/m, z1.s
+; CHECK-NEXT:    eor z0.d, z0.d, z2.d
+; CHECK-NEXT:    sub z0.s, z1.s, z0.s
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 4 x i16> %s0 to <vscale x 4 x i32>
+  %s1s = sext <vscale x 4 x i16> %s1 to <vscale x 4 x i32>
+  %add = add <vscale x 4 x i32> %s0s, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i32 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+  %add2 = add <vscale x 4 x i32> %add, %s1s
+  %s = lshr <vscale x 4 x i32> %add2, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i32 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+  %result = trunc <vscale x 4 x i32> %s to <vscale x 4 x i16>
+  ret <vscale x 4 x i16> %result
+}
+
+define <vscale x 4 x i16> @rhaddu_v4i16(<vscale x 4 x i16> %s0, <vscale x 4 x i16> %s1) {
+; CHECK-LABEL: rhaddu_v4i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    orr z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.s, z0.s, #0xffff
+; CHECK-NEXT:    and z1.s, z1.s, #0xffff
+; CHECK-NEXT:    lsr z1.s, z1.s, #1
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    and z2.s, z2.s, #0x1
+; CHECK-NEXT:    add z0.s, z0.s, z1.s
+; CHECK-NEXT:    add z0.s, z0.s, z2.s
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 4 x i16> %s0 to <vscale x 4 x i32>
+  %s1s = zext <vscale x 4 x i16> %s1 to <vscale x 4 x i32>
+  %add = add nuw nsw <vscale x 4 x i32> %s0s, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i32 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 4 x i32> %add, %s1s
+  %s = lshr <vscale x 4 x i32> %add2, shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i32 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
+  %result = trunc <vscale x 4 x i32> %s to <vscale x 4 x i16>
+  ret <vscale x 4 x i16> %result
+}
+
+define <vscale x 8 x i16> @rhadds_v8i16(<vscale x 8 x i16> %s0, <vscale x 8 x i16> %s1) {
+; CHECK-LABEL: rhadds_v8i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.h, z1.h, #1
+; CHECK-NEXT:    lsr z3.h, z0.h, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.h, z3.h, z2.h
+; CHECK-NEXT:    and z0.h, z0.h, #0x1
+; CHECK-NEXT:    add z0.h, z2.h, z0.h
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 8 x i16> %s0 to <vscale x 8 x i32>
+  %s1s = sext <vscale x 8 x i16> %s1 to <vscale x 8 x i32>
+  %add = add <vscale x 8 x i32> %s0s, shufflevector (<vscale x 8 x i32> insertelement (<vscale x 8 x i32> poison, i32 1, i32 0), <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer)
+  %add2 = add <vscale x 8 x i32> %add, %s1s
+  %s = lshr <vscale x 8 x i32> %add2, shufflevector (<vscale x 8 x i32> insertelement (<vscale x 8 x i32> poison, i32 1, i32 0), <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer)
+  %result = trunc <vscale x 8 x i32> %s to <vscale x 8 x i16>
+  ret <vscale x 8 x i16> %result
+}
+
+define <vscale x 8 x i16> @rhaddu_v8i16(<vscale x 8 x i16> %s0, <vscale x 8 x i16> %s1) {
+; CHECK-LABEL: rhaddu_v8i16:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.h, z1.h, #1
+; CHECK-NEXT:    lsr z3.h, z0.h, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.h, z3.h, z2.h
+; CHECK-NEXT:    and z0.h, z0.h, #0x1
+; CHECK-NEXT:    add z0.h, z2.h, z0.h
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 8 x i16> %s0 to <vscale x 8 x i32>
+  %s1s = zext <vscale x 8 x i16> %s1 to <vscale x 8 x i32>
+  %add = add nuw nsw <vscale x 8 x i32> %s0s, shufflevector (<vscale x 8 x i32> insertelement (<vscale x 8 x i32> poison, i32 1, i32 0), <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 8 x i32> %add, %s1s
+  %s = lshr <vscale x 8 x i32> %add2, shufflevector (<vscale x 8 x i32> insertelement (<vscale x 8 x i32> poison, i32 1, i32 0), <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer)
+  %result = trunc <vscale x 8 x i32> %s to <vscale x 8 x i16>
+  ret <vscale x 8 x i16> %result
+}
+
+define <vscale x 4 x i8> @rhadds_v4i8(<vscale x 4 x i8> %s0, <vscale x 4 x i8> %s1) {
+; CHECK-LABEL: rhadds_v4i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.s
+; CHECK-NEXT:    mov z2.s, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    sxtb z0.s, p0/m, z0.s
+; CHECK-NEXT:    sxtb z1.s, p0/m, z1.s
+; CHECK-NEXT:    eor z0.d, z0.d, z2.d
+; CHECK-NEXT:    sub z0.s, z1.s, z0.s
+; CHECK-NEXT:    and z0.s, z0.s, #0xffff
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 4 x i8> %s0 to <vscale x 4 x i16>
+  %s1s = sext <vscale x 4 x i8> %s1 to <vscale x 4 x i16>
+  %add = add <vscale x 4 x i16> %s0s, shufflevector (<vscale x 4 x i16> insertelement (<vscale x 4 x i16> poison, i16 1, i32 0), <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer)
+  %add2 = add <vscale x 4 x i16> %add, %s1s
+  %s = lshr <vscale x 4 x i16> %add2, shufflevector (<vscale x 4 x i16> insertelement (<vscale x 4 x i16> poison, i16 1, i32 0), <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer)
+  %result = trunc <vscale x 4 x i16> %s to <vscale x 4 x i8>
+  ret <vscale x 4 x i8> %result
+}
+
+define <vscale x 4 x i8> @rhaddu_v4i8(<vscale x 4 x i8> %s0, <vscale x 4 x i8> %s1) {
+; CHECK-LABEL: rhaddu_v4i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    and z0.s, z0.s, #0xff
+; CHECK-NEXT:    mov z2.s, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    and z1.s, z1.s, #0xff
+; CHECK-NEXT:    eor z0.d, z0.d, z2.d
+; CHECK-NEXT:    sub z0.s, z1.s, z0.s
+; CHECK-NEXT:    lsr z0.s, z0.s, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 4 x i8> %s0 to <vscale x 4 x i16>
+  %s1s = zext <vscale x 4 x i8> %s1 to <vscale x 4 x i16>
+  %add = add nuw nsw <vscale x 4 x i16> %s0s, shufflevector (<vscale x 4 x i16> insertelement (<vscale x 4 x i16> poison, i16 1, i32 0), <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 4 x i16> %add, %s1s
+  %s = lshr <vscale x 4 x i16> %add2, shufflevector (<vscale x 4 x i16> insertelement (<vscale x 4 x i16> poison, i16 1, i32 0), <vscale x 4 x i16> poison, <vscale x 4 x i32> zeroinitializer)
+  %result = trunc <vscale x 4 x i16> %s to <vscale x 4 x i8>
+  ret <vscale x 4 x i8> %result
+}
+
+define <vscale x 8 x i8> @rhadds_v8i8(<vscale x 8 x i8> %s0, <vscale x 8 x i8> %s1) {
+; CHECK-LABEL: rhadds_v8i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ptrue p0.h
+; CHECK-NEXT:    mov z2.h, #-1 // =0xffffffffffffffff
+; CHECK-NEXT:    sxtb z0.h, p0/m, z0.h
+; CHECK-NEXT:    sxtb z1.h, p0/m, z1.h
+; CHECK-NEXT:    eor z0.d, z0.d, z2.d
+; CHECK-NEXT:    sub z0.h, z1.h, z0.h
+; CHECK-NEXT:    lsr z0.h, z0.h, #1
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 8 x i8> %s0 to <vscale x 8 x i16>
+  %s1s = sext <vscale x 8 x i8> %s1 to <vscale x 8 x i16>
+  %add = add <vscale x 8 x i16> %s0s, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 1, i32 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %add2 = add <vscale x 8 x i16> %add, %s1s
+  %s = lshr <vscale x 8 x i16> %add2, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 1, i32 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %result = trunc <vscale x 8 x i16> %s to <vscale x 8 x i8>
+  ret <vscale x 8 x i8> %result
+}
+
+define <vscale x 8 x i8> @rhaddu_v8i8(<vscale x 8 x i8> %s0, <vscale x 8 x i8> %s1) {
+; CHECK-LABEL: rhaddu_v8i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    orr z2.d, z0.d, z1.d
+; CHECK-NEXT:    and z0.h, z0.h, #0xff
+; CHECK-NEXT:    and z1.h, z1.h, #0xff
+; CHECK-NEXT:    lsr z1.h, z1.h, #1
+; CHECK-NEXT:    lsr z0.h, z0.h, #1
+; CHECK-NEXT:    and z2.h, z2.h, #0x1
+; CHECK-NEXT:    add z0.h, z0.h, z1.h
+; CHECK-NEXT:    add z0.h, z0.h, z2.h
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 8 x i8> %s0 to <vscale x 8 x i16>
+  %s1s = zext <vscale x 8 x i8> %s1 to <vscale x 8 x i16>
+  %add = add nuw nsw <vscale x 8 x i16> %s0s, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 1, i32 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 8 x i16> %add, %s1s
+  %s = lshr <vscale x 8 x i16> %add2, shufflevector (<vscale x 8 x i16> insertelement (<vscale x 8 x i16> poison, i16 1, i32 0), <vscale x 8 x i16> poison, <vscale x 8 x i32> zeroinitializer)
+  %result = trunc <vscale x 8 x i16> %s to <vscale x 8 x i8>
+  ret <vscale x 8 x i8> %result
+}
+
+define <vscale x 16 x i8> @rhadds_v16i8(<vscale x 16 x i8> %s0, <vscale x 16 x i8> %s1) {
+; CHECK-LABEL: rhadds_v16i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.b, z1.b, #1
+; CHECK-NEXT:    lsr z3.b, z0.b, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.b, z3.b, z2.b
+; CHECK-NEXT:    and z0.b, z0.b, #0x1
+; CHECK-NEXT:    add z0.b, z2.b, z0.b
+; CHECK-NEXT:    ret
+entry:
+  %s0s = sext <vscale x 16 x i8> %s0 to <vscale x 16 x i16>
+  %s1s = sext <vscale x 16 x i8> %s1 to <vscale x 16 x i16>
+  %add = add <vscale x 16 x i16> %s0s, shufflevector (<vscale x 16 x i16> insertelement (<vscale x 16 x i16> poison, i16 1, i32 0), <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer)
+  %add2 = add <vscale x 16 x i16> %add, %s1s
+  %s = lshr <vscale x 16 x i16> %add2, shufflevector (<vscale x 16 x i16> insertelement (<vscale x 16 x i16> poison, i16 1, i32 0), <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer)
+  %result = trunc <vscale x 16 x i16> %s to <vscale x 16 x i8>
+  ret <vscale x 16 x i8> %result
+}
+
+define <vscale x 16 x i8> @rhaddu_v16i8(<vscale x 16 x i8> %s0, <vscale x 16 x i8> %s1) {
+; CHECK-LABEL: rhaddu_v16i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    lsr z2.b, z1.b, #1
+; CHECK-NEXT:    lsr z3.b, z0.b, #1
+; CHECK-NEXT:    orr z0.d, z0.d, z1.d
+; CHECK-NEXT:    add z2.b, z3.b, z2.b
+; CHECK-NEXT:    and z0.b, z0.b, #0x1
+; CHECK-NEXT:    add z0.b, z2.b, z0.b
+; CHECK-NEXT:    ret
+entry:
+  %s0s = zext <vscale x 16 x i8> %s0 to <vscale x 16 x i16>
+  %s1s = zext <vscale x 16 x i8> %s1 to <vscale x 16 x i16>
+  %add = add nuw nsw <vscale x 16 x i16> %s0s, shufflevector (<vscale x 16 x i16> insertelement (<vscale x 16 x i16> poison, i16 1, i32 0), <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer)
+  %add2 = add nuw nsw <vscale x 16 x i16> %add, %s1s
+  %s = lshr <vscale x 16 x i16> %add2, shufflevector (<vscale x 16 x i16> insertelement (<vscale x 16 x i16> poison, i16 1, i32 0), <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer)
+  %result = trunc <vscale x 16 x i16> %s to <vscale x 16 x i8>
+  ret <vscale x 16 x i8> %result
+}