Teach the AArch64 backend about v4f16 and v8f16

[olista01]

Summary

This is the vector counterpart to D4879, and teaches the AArch64 backend to deal with the operations required to deal with the operations on v4f16 and v8f16 which are exposed by NEON intrinsics, plus the add, sub, mul and div operations.

[t.p.northover]

Hi Oliver,

It looks like there are various components to this patch, some without tests. And some of the tests that are here are a bit generic (it'd be nice to track that the correct operands get used, rather than just that an "fdiv" is emitted, for example).

• There appear to be no tests for the ABI changes.
• Hardly any coverage of load/store patterns or the C++ ISel
• bitcast also seems to get short shrift, given how substantial the code changes are (mostly just used when it has to be as part of other shuffle testing).

We don't necessarily need blanket coverage (v4f16 is fundamentally very similar to v4i16) but something in between would be good.

Cheers.

Tim.

lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp

387–394

This renders the comment just above wrong. Could you fix that?

[olista01]

• Added ABI tests
• Added tests for NEON loads/stores (and added a few missing patterns)
• Added tests for bitcasts
• Tightened up tests which do not specify registers (ignoring some of the larger v8f16 outputs, as they use many vector inserts/extracts and re-use registers which I could not work out a good way to test)
• Fixed out of date comment

[olista01]

Ping?

[rengolin]

Hi Oliver,

Tim is on holidays...

Given that I didn't find anything wrong with the patch and that Tim's only complaints were the lack of testing (and that you have added a lot more and fixed the ones he complained about), I'm going to approve this patch.

I'll also send an email to Tim, so if he has any additional concerns, he can do a post-commit review.

Thanks,
--renato

[olista01]

Thanks, committed revision 216555.

include/llvm/CodeGen/MachineValueType.h

public:
     bool is16BitVector() const {
       return (SimpleTy == MVT::v2i8  || SimpleTy == MVT::v1i16 ||
               SimpleTy == MVT::v16i1);
     }
 
     /// is32BitVector - Return true if this is a 32-bit vector type.
     bool is32BitVector() const {
       return (SimpleTy == MVT::v4i8  || SimpleTy == MVT::v2i16 ||
-              SimpleTy == MVT::v1i32);
+              SimpleTy == MVT::v1i32 || SimpleTy == MVT::v2f16 ||
+              SimpleTy == MVT::v1f32);
     }
 
     /// is64BitVector - Return true if this is a 64-bit vector type.
     bool is64BitVector() const {
       return (SimpleTy == MVT::v8i8  || SimpleTy == MVT::v4i16 ||
               SimpleTy == MVT::v2i32 || SimpleTy == MVT::v1i64 ||
-              SimpleTy == MVT::v1f64 || SimpleTy == MVT::v2f32);
+              SimpleTy == MVT::v4f16 || SimpleTy == MVT::v2f32 ||
+              SimpleTy == MVT::v1f64);
     }
 
     /// is128BitVector - Return true if this is a 128-bit vector type.
     bool is128BitVector() const {
       return (SimpleTy == MVT::v16i8 || SimpleTy == MVT::v8i16 ||
               SimpleTy == MVT::v4i32 || SimpleTy == MVT::v2i64 ||
-              SimpleTy == MVT::v4f32 || SimpleTy == MVT::v2f64);
+              SimpleTy == MVT::v8f16 || SimpleTy == MVT::v4f32 ||
+              SimpleTy == MVT::v2f64);
     }
 
     /// is256BitVector - Return true if this is a 256-bit vector type.
     bool is256BitVector() const {
       return (SimpleTy == MVT::v8f32 || SimpleTy == MVT::v4f64  ||
               SimpleTy == MVT::v32i8 || SimpleTy == MVT::v16i16 ||
               SimpleTy == MVT::v8i32 || SimpleTy == MVT::v4i64);
     }

lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp

case ISD::UINT_TO_FP:
     // "Promote" the operation by extending the operand.
     return PromoteINT_TO_FP(Op);
   case ISD::FP_TO_UINT:
   case ISD::FP_TO_SINT:
     // Promote the operation by extending the operand.
     return PromoteFP_TO_INT(Op, Op->getOpcode() == ISD::FP_TO_SINT);
   }
 
-  // The rest of the time, vector "promotion" is basically just bitcasting and
-  // doing the operation in a different type.  For example, x86 promotes
-  // ISD::AND on v2i32 to v1i64.
+  // There are currently two cases of vector promotion:
+  // 1) Bitcasting a vector of integers to a different type to a vector of the
+  //    same overall length. For example, x86 promotes ISD::AND on v2i32 to v1i64.
+  // 2) Extending a vector of floats to a vector of the same number oflarger
+  //    floats. For example, AArch64 promotes ISD::FADD on v4f16 to v4f32.
   MVT VT = Op.getSimpleValueType();
   assert(Op.getNode()->getNumValues() == 1 &&
          "Can't promote a vector with multiple results!");
   MVT NVT = TLI.getTypeToPromoteTo(Op.getOpcode(), VT);
   SDLoc dl(Op);
   SmallVector<SDValue, 4> Operands(Op.getNumOperands());
 
   for (unsigned j = 0; j != Op.getNumOperands(); ++j) {
     if (Op.getOperand(j).getValueType().isVector())
+      if (Op.getOperand(j)
+              .getValueType()
+              .getVectorElementType()
+              .isFloatingPoint())
+        Operands[j] = DAG.getNode(ISD::FP_EXTEND, dl, NVT, Op.getOperand(j));
+      else
         Operands[j] = DAG.getNode(ISD::BITCAST, dl, NVT, Op.getOperand(j));
     else
       Operands[j] = Op.getOperand(j);
   }
 
   Op = DAG.getNode(Op.getOpcode(), dl, NVT, Operands);
-
+  if (VT.isFloatingPoint() ||
+      (VT.isVector() && VT.getVectorElementType().isFloatingPoint()))
+    return DAG.getNode(ISD::FP_ROUND, dl, VT, Op, DAG.getIntPtrConstant(0));
+  else
     return DAG.getNode(ISD::BITCAST, dl, VT, Op);
 }
 
 SDValue VectorLegalizer::PromoteINT_TO_FP(SDValue Op) {
   // INT_TO_FP operations may require the input operand be promoted even
   // when the type is otherwise legal.
   EVT VT = Op.getOperand(0).getValueType();
   assert(Op.getNode()->getNumValues() == 1 &&
          "Can't promote a vector with multiple results!");

lib/Target/AArch64/AArch64CallingConvention.td

def CC_AArch64_AAPCS : CallingConv<[
   CCIfType<[i64], CCAssignToRegWithShadow<[X0, X1, X2, X3, X4, X5, X6, X7],
                                           [W0, W1, W2, W3, W4, W5, W6, W7]>>,
   CCIfType<[f16], CCAssignToRegWithShadow<[H0, H1, H2, H3, H4, H5, H6, H7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
   CCIfType<[f32], CCAssignToRegWithShadow<[S0, S1, S2, S3, S4, S5, S6, S7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
   CCIfType<[f64], CCAssignToRegWithShadow<[D0, D1, D2, D3, D4, D5, D6, D7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
-  CCIfType<[v1i64, v2i32, v4i16, v8i8, v1f64, v2f32],
+  CCIfType<[v1i64, v2i32, v4i16, v8i8, v1f64, v2f32, v4f16],
            CCAssignToRegWithShadow<[D0, D1, D2, D3, D4, D5, D6, D7],
                                    [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
-  CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64],
+  CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16],
            CCAssignToReg<[Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
 
   // If more than will fit in registers, pass them on the stack instead.
   CCIfType<[i1, i8, i16, f16], CCAssignToStack<8, 8>>,
   CCIfType<[i32, f32], CCAssignToStack<8, 8>>,
-  CCIfType<[i64, f64, v1f64, v2f32, v1i64, v2i32, v4i16, v8i8],
+  CCIfType<[i64, f64, v1f64, v2f32, v1i64, v2i32, v4i16, v8i8, v4f16],
            CCAssignToStack<8, 8>>,
-  CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64],
+  CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16],
            CCAssignToStack<16, 16>>
 ]>;
 
 def RetCC_AArch64_AAPCS : CallingConv<[
   CCIfType<[v2f32], CCBitConvertToType<v2i32>>,
   CCIfType<[v2f64, v4f32], CCBitConvertToType<v2i64>>,
 
   // Big endian vectors must be passed as if they were 1-element vectors so that
   // their lanes are in a consistent order.
   CCIfBigEndian<CCIfType<[v2i32, v2f32, v4i16, v4f16, v8i8],
                          CCBitConvertToType<f64>>>,
   CCIfBigEndian<CCIfType<[v2i64, v2f64, v4i32, v4f32, v8i16, v8f16, v16i8],
                          CCBitConvertToType<f128>>>,
 
   CCIfType<[i32], CCAssignToRegWithShadow<[W0, W1, W2, W3, W4, W5, W6, W7],
                                           [X0, X1, X2, X3, X4, X5, X6, X7]>>,
   CCIfType<[i64], CCAssignToRegWithShadow<[X0, X1, X2, X3, X4, X5, X6, X7],
                                           [W0, W1, W2, W3, W4, W5, W6, W7]>>,
   CCIfType<[f16], CCAssignToRegWithShadow<[H0, H1, H2, H3, H4, H5, H6, H7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
   CCIfType<[f32], CCAssignToRegWithShadow<[S0, S1, S2, S3, S4, S5, S6, S7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
   CCIfType<[f64], CCAssignToRegWithShadow<[D0, D1, D2, D3, D4, D5, D6, D7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
-  CCIfType<[v1i64, v2i32, v4i16, v8i8, v1f64, v2f32],
+  CCIfType<[v1i64, v2i32, v4i16, v8i8, v1f64, v2f32, v4f16],
       CCAssignToRegWithShadow<[D0, D1, D2, D3, D4, D5, D6, D7],
                               [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
-  CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64],
+  CCIfType<[f128, v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16],
       CCAssignToReg<[Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>
 ]>;
 
 
 // Darwin uses a calling convention which differs in only two ways
 // from the standard one at this level:
 //     + i128s (i.e. split i64s) don't need even registers.
 //     + Stack slots are sized as needed rather than being at least 64-bit.
def CC_AArch64_DarwinPCS : CallingConv<[
   CCIfType<[i64], CCAssignToRegWithShadow<[X0, X1, X2, X3, X4, X5, X6, X7],
                                           [W0, W1, W2, W3, W4, W5, W6, W7]>>,
   CCIfType<[f16], CCAssignToRegWithShadow<[H0, H1, H2, H3, H4, H5, H6, H7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
   CCIfType<[f32], CCAssignToRegWithShadow<[S0, S1, S2, S3, S4, S5, S6, S7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
   CCIfType<[f64], CCAssignToRegWithShadow<[D0, D1, D2, D3, D4, D5, D6, D7],
                                           [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
-  CCIfType<[v1i64, v2i32, v4i16, v8i8, v1f64, v2f32],
+  CCIfType<[v1i64, v2i32, v4i16, v8i8, v1f64, v2f32, v4f16],
            CCAssignToRegWithShadow<[D0, D1, D2, D3, D4, D5, D6, D7],
                                    [Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
-  CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64],
+  CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16],
            CCAssignToReg<[Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>,
 
   // If more than will fit in registers, pass them on the stack instead.
   CCIf<"ValVT == MVT::i1 || ValVT == MVT::i8", CCAssignToStack<1, 1>>,
   CCIf<"ValVT == MVT::i16 || ValVT == MVT::f16", CCAssignToStack<2, 2>>,
   CCIfType<[i32, f32], CCAssignToStack<4, 4>>,
-  CCIfType<[i64, f64, v1f64, v2f32, v1i64, v2i32, v4i16, v8i8],
+  CCIfType<[i64, f64, v1f64, v2f32, v1i64, v2i32, v4i16, v8i8, v4f16],
            CCAssignToStack<8, 8>>,
-  CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64], CCAssignToStack<16, 16>>
+  CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16],
+           CCAssignToStack<16, 16>>
 ]>;
 
 def CC_AArch64_DarwinPCS_VarArg : CallingConv<[
   CCIfType<[v2f32], CCBitConvertToType<v2i32>>,
   CCIfType<[v2f64, v4f32, f128], CCBitConvertToType<v2i64>>,
 
   // Handle all scalar types as either i64 or f64.
   CCIfType<[i8, i16, i32], CCPromoteToType<i64>>,
   CCIfType<[f16, f32],     CCPromoteToType<f64>>,
 
   // Everything is on the stack.
   // i128 is split to two i64s, and its stack alignment is 16 bytes.
   CCIfType<[i64], CCIfSplit<CCAssignToStack<8, 16>>>,
-  CCIfType<[i64, f64, v1i64, v2i32, v4i16, v8i8, v1f64, v2f32], CCAssignToStack<8, 8>>,
-  CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64],   CCAssignToStack<16, 16>>
+  CCIfType<[i64, f64, v1i64, v2i32, v4i16, v8i8, v1f64, v2f32, v4f16],
+           CCAssignToStack<8, 8>>,
+  CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, v2f64, v8f16],
+           CCAssignToStack<16, 16>>
 ]>;
 
 // The WebKit_JS calling convention only passes the first argument (the callee)
 // in register and the remaining arguments on stack. We allow 32bit stack slots,
 // so that WebKit can write partial values in the stack and define the other
 // 32bit quantity as undef.
 def CC_AArch64_WebKit_JS : CallingConv<[
   // Handle i1, i8, i16, i32, and i64 passing in register X0 (W0).

lib/Target/AArch64/AArch64ISelDAGToDAG.cpp

case ISD::EXTRACT_VECTOR_ELT: {
     default:
       llvm_unreachable("Unexpected vector element type!");
     case 64:
       SubReg = AArch64::dsub;
       break;
     case 32:
       SubReg = AArch64::ssub;
       break;
-    case 16: // FALLTHROUGH
+    case 16:
+      SubReg = AArch64::hsub;
+      break;
     case 8:
       llvm_unreachable("unexpected zext-requiring extract element!");
     }
     SDValue Extract = CurDAG->getTargetExtractSubreg(SubReg, SDLoc(Node), VT,
                                                      Node->getOperand(0));
     DEBUG(dbgs() << "ISEL: Custom selection!\n=> ");
     DEBUG(Extract->dumpr(CurDAG));
     DEBUG(dbgs() << "\n");
case Intrinsic::aarch64_stxp: {
 
       return St;
     }
     case Intrinsic::aarch64_neon_ld1x2:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 2, AArch64::LD1Twov8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 2, AArch64::LD1Twov16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 2, AArch64::LD1Twov4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 2, AArch64::LD1Twov8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 2, AArch64::LD1Twov2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 2, AArch64::LD1Twov4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 2, AArch64::LD1Twov1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 2, AArch64::LD1Twov2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld1x3:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 3, AArch64::LD1Threev8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 3, AArch64::LD1Threev16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 3, AArch64::LD1Threev4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 3, AArch64::LD1Threev8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 3, AArch64::LD1Threev2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 3, AArch64::LD1Threev4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 3, AArch64::LD1Threev1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 3, AArch64::LD1Threev2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld1x4:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 4, AArch64::LD1Fourv8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 4, AArch64::LD1Fourv16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 4, AArch64::LD1Fourv4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 4, AArch64::LD1Fourv8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 4, AArch64::LD1Fourv2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 4, AArch64::LD1Fourv4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 4, AArch64::LD1Fourv1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 4, AArch64::LD1Fourv2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld2:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 2, AArch64::LD2Twov8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 2, AArch64::LD2Twov16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 2, AArch64::LD2Twov4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 2, AArch64::LD2Twov8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 2, AArch64::LD2Twov2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 2, AArch64::LD2Twov4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 2, AArch64::LD1Twov1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 2, AArch64::LD2Twov2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld3:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 3, AArch64::LD3Threev8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 3, AArch64::LD3Threev16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 3, AArch64::LD3Threev4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 3, AArch64::LD3Threev8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 3, AArch64::LD3Threev2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 3, AArch64::LD3Threev4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 3, AArch64::LD1Threev1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 3, AArch64::LD3Threev2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld4:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 4, AArch64::LD4Fourv8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 4, AArch64::LD4Fourv16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 4, AArch64::LD4Fourv4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16  || VT == MVT::v8f16)
         return SelectLoad(Node, 4, AArch64::LD4Fourv8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 4, AArch64::LD4Fourv2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 4, AArch64::LD4Fourv4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 4, AArch64::LD1Fourv1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 4, AArch64::LD4Fourv2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld2r:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 2, AArch64::LD2Rv8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 2, AArch64::LD2Rv16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 2, AArch64::LD2Rv4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 2, AArch64::LD2Rv8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 2, AArch64::LD2Rv2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 2, AArch64::LD2Rv4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 2, AArch64::LD2Rv1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 2, AArch64::LD2Rv2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld3r:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 3, AArch64::LD3Rv8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 3, AArch64::LD3Rv16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 3, AArch64::LD3Rv4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 3, AArch64::LD3Rv8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 3, AArch64::LD3Rv2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 3, AArch64::LD3Rv4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 3, AArch64::LD3Rv1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 3, AArch64::LD3Rv2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld4r:
       if (VT == MVT::v8i8)
         return SelectLoad(Node, 4, AArch64::LD4Rv8b, AArch64::dsub0);
       else if (VT == MVT::v16i8)
         return SelectLoad(Node, 4, AArch64::LD4Rv16b, AArch64::qsub0);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectLoad(Node, 4, AArch64::LD4Rv4h, AArch64::dsub0);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectLoad(Node, 4, AArch64::LD4Rv8h, AArch64::qsub0);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectLoad(Node, 4, AArch64::LD4Rv2s, AArch64::dsub0);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectLoad(Node, 4, AArch64::LD4Rv4s, AArch64::qsub0);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectLoad(Node, 4, AArch64::LD4Rv1d, AArch64::dsub0);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectLoad(Node, 4, AArch64::LD4Rv2d, AArch64::qsub0);
       break;
     case Intrinsic::aarch64_neon_ld2lane:
       if (VT == MVT::v16i8 || VT == MVT::v8i8)
         return SelectLoadLane(Node, 2, AArch64::LD2i8);
-      else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+               VT == MVT::v8f16)
         return SelectLoadLane(Node, 2, AArch64::LD2i16);
       else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
                VT == MVT::v2f32)
         return SelectLoadLane(Node, 2, AArch64::LD2i32);
       else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
                VT == MVT::v1f64)
         return SelectLoadLane(Node, 2, AArch64::LD2i64);
       break;
     case Intrinsic::aarch64_neon_ld3lane:
       if (VT == MVT::v16i8 || VT == MVT::v8i8)
         return SelectLoadLane(Node, 3, AArch64::LD3i8);
-      else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+               VT == MVT::v8f16)
         return SelectLoadLane(Node, 3, AArch64::LD3i16);
       else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
                VT == MVT::v2f32)
         return SelectLoadLane(Node, 3, AArch64::LD3i32);
       else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
                VT == MVT::v1f64)
         return SelectLoadLane(Node, 3, AArch64::LD3i64);
       break;
     case Intrinsic::aarch64_neon_ld4lane:
       if (VT == MVT::v16i8 || VT == MVT::v8i8)
         return SelectLoadLane(Node, 4, AArch64::LD4i8);
-      else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+               VT == MVT::v8f16)
         return SelectLoadLane(Node, 4, AArch64::LD4i16);
       else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
                VT == MVT::v2f32)
         return SelectLoadLane(Node, 4, AArch64::LD4i32);
       else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
                VT == MVT::v1f64)
         return SelectLoadLane(Node, 4, AArch64::LD4i64);
       break;
case ISD::INTRINSIC_VOID: {
     switch (IntNo) {
     default:
       break;
     case Intrinsic::aarch64_neon_st1x2: {
       if (VT == MVT::v8i8)
         return SelectStore(Node, 2, AArch64::ST1Twov8b);
       else if (VT == MVT::v16i8)
         return SelectStore(Node, 2, AArch64::ST1Twov16b);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectStore(Node, 2, AArch64::ST1Twov4h);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectStore(Node, 2, AArch64::ST1Twov8h);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectStore(Node, 2, AArch64::ST1Twov2s);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectStore(Node, 2, AArch64::ST1Twov4s);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectStore(Node, 2, AArch64::ST1Twov2d);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectStore(Node, 2, AArch64::ST1Twov1d);
       break;
     }
     case Intrinsic::aarch64_neon_st1x3: {
       if (VT == MVT::v8i8)
         return SelectStore(Node, 3, AArch64::ST1Threev8b);
       else if (VT == MVT::v16i8)
         return SelectStore(Node, 3, AArch64::ST1Threev16b);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectStore(Node, 3, AArch64::ST1Threev4h);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectStore(Node, 3, AArch64::ST1Threev8h);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectStore(Node, 3, AArch64::ST1Threev2s);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectStore(Node, 3, AArch64::ST1Threev4s);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectStore(Node, 3, AArch64::ST1Threev2d);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectStore(Node, 3, AArch64::ST1Threev1d);
       break;
     }
     case Intrinsic::aarch64_neon_st1x4: {
       if (VT == MVT::v8i8)
         return SelectStore(Node, 4, AArch64::ST1Fourv8b);
       else if (VT == MVT::v16i8)
         return SelectStore(Node, 4, AArch64::ST1Fourv16b);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectStore(Node, 4, AArch64::ST1Fourv4h);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectStore(Node, 4, AArch64::ST1Fourv8h);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectStore(Node, 4, AArch64::ST1Fourv2s);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectStore(Node, 4, AArch64::ST1Fourv4s);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectStore(Node, 4, AArch64::ST1Fourv2d);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectStore(Node, 4, AArch64::ST1Fourv1d);
       break;
     }
     case Intrinsic::aarch64_neon_st2: {
       if (VT == MVT::v8i8)
         return SelectStore(Node, 2, AArch64::ST2Twov8b);
       else if (VT == MVT::v16i8)
         return SelectStore(Node, 2, AArch64::ST2Twov16b);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectStore(Node, 2, AArch64::ST2Twov4h);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectStore(Node, 2, AArch64::ST2Twov8h);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectStore(Node, 2, AArch64::ST2Twov2s);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectStore(Node, 2, AArch64::ST2Twov4s);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectStore(Node, 2, AArch64::ST2Twov2d);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectStore(Node, 2, AArch64::ST1Twov1d);
       break;
     }
     case Intrinsic::aarch64_neon_st3: {
       if (VT == MVT::v8i8)
         return SelectStore(Node, 3, AArch64::ST3Threev8b);
       else if (VT == MVT::v16i8)
         return SelectStore(Node, 3, AArch64::ST3Threev16b);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectStore(Node, 3, AArch64::ST3Threev4h);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectStore(Node, 3, AArch64::ST3Threev8h);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectStore(Node, 3, AArch64::ST3Threev2s);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectStore(Node, 3, AArch64::ST3Threev4s);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectStore(Node, 3, AArch64::ST3Threev2d);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectStore(Node, 3, AArch64::ST1Threev1d);
       break;
     }
     case Intrinsic::aarch64_neon_st4: {
       if (VT == MVT::v8i8)
         return SelectStore(Node, 4, AArch64::ST4Fourv8b);
       else if (VT == MVT::v16i8)
         return SelectStore(Node, 4, AArch64::ST4Fourv16b);
-      else if (VT == MVT::v4i16)
+      else if (VT == MVT::v4i16 || VT == MVT::v4f16)
         return SelectStore(Node, 4, AArch64::ST4Fourv4h);
-      else if (VT == MVT::v8i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v8f16)
         return SelectStore(Node, 4, AArch64::ST4Fourv8h);
       else if (VT == MVT::v2i32 || VT == MVT::v2f32)
         return SelectStore(Node, 4, AArch64::ST4Fourv2s);
       else if (VT == MVT::v4i32 || VT == MVT::v4f32)
         return SelectStore(Node, 4, AArch64::ST4Fourv4s);
       else if (VT == MVT::v2i64 || VT == MVT::v2f64)
         return SelectStore(Node, 4, AArch64::ST4Fourv2d);
       else if (VT == MVT::v1i64 || VT == MVT::v1f64)
         return SelectStore(Node, 4, AArch64::ST1Fourv1d);
       break;
     }
     case Intrinsic::aarch64_neon_st2lane: {
       if (VT == MVT::v16i8 || VT == MVT::v8i8)
         return SelectStoreLane(Node, 2, AArch64::ST2i8);
-      else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+               VT == MVT::v8f16)
         return SelectStoreLane(Node, 2, AArch64::ST2i16);
       else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
                VT == MVT::v2f32)
         return SelectStoreLane(Node, 2, AArch64::ST2i32);
       else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
                VT == MVT::v1f64)
         return SelectStoreLane(Node, 2, AArch64::ST2i64);
       break;
     }
     case Intrinsic::aarch64_neon_st3lane: {
       if (VT == MVT::v16i8 || VT == MVT::v8i8)
         return SelectStoreLane(Node, 3, AArch64::ST3i8);
-      else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+               VT == MVT::v8f16)
         return SelectStoreLane(Node, 3, AArch64::ST3i16);
       else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
                VT == MVT::v2f32)
         return SelectStoreLane(Node, 3, AArch64::ST3i32);
       else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
                VT == MVT::v1f64)
         return SelectStoreLane(Node, 3, AArch64::ST3i64);
       break;
     }
     case Intrinsic::aarch64_neon_st4lane: {
       if (VT == MVT::v16i8 || VT == MVT::v8i8)
         return SelectStoreLane(Node, 4, AArch64::ST4i8);
-      else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+      else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+               VT == MVT::v8f16)
         return SelectStoreLane(Node, 4, AArch64::ST4i16);
       else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
                VT == MVT::v2f32)
         return SelectStoreLane(Node, 4, AArch64::ST4i32);
       else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
                VT == MVT::v1f64)
         return SelectStoreLane(Node, 4, AArch64::ST4i64);
       break;
     }
     }
   }
   case AArch64ISD::LD2post: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 2, AArch64::LD2Twov8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 2, AArch64::LD2Twov16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 2, AArch64::LD2Twov4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 2, AArch64::LD2Twov8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 2, AArch64::LD2Twov2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 2, AArch64::LD2Twov4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 2, AArch64::LD2Twov2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD3post: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 3, AArch64::LD3Threev8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 3, AArch64::LD3Threev16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 3, AArch64::LD3Threev4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 3, AArch64::LD3Threev8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 3, AArch64::LD3Threev2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 3, AArch64::LD3Threev4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 3, AArch64::LD3Threev2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD4post: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 4, AArch64::LD4Fourv8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 4, AArch64::LD4Fourv16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 4, AArch64::LD4Fourv4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 4, AArch64::LD4Fourv8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 4, AArch64::LD4Fourv2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 4, AArch64::LD4Fourv4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 4, AArch64::LD4Fourv2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD1x2post: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 2, AArch64::LD1Twov2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD1x3post: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 3, AArch64::LD1Threev2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD1x4post: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 4, AArch64::LD1Fourv2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD1DUPpost: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 1, AArch64::LD1Rv2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD2DUPpost: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 2, AArch64::LD2Rv2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD3DUPpost: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 3, AArch64::LD3Rv2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD4DUPpost: {
     if (VT == MVT::v8i8)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv8b_POST, AArch64::dsub0);
     else if (VT == MVT::v16i8)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv16b_POST, AArch64::qsub0);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv4h_POST, AArch64::dsub0);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv8h_POST, AArch64::qsub0);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv2s_POST, AArch64::dsub0);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv4s_POST, AArch64::qsub0);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv1d_POST, AArch64::dsub0);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostLoad(Node, 4, AArch64::LD4Rv2d_POST, AArch64::qsub0);
     break;
   }
   case AArch64ISD::LD1LANEpost: {
     if (VT == MVT::v16i8 || VT == MVT::v8i8)
       return SelectPostLoadLane(Node, 1, AArch64::LD1i8_POST);
-    else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+             VT == MVT::v8f16)
       return SelectPostLoadLane(Node, 1, AArch64::LD1i16_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
              VT == MVT::v2f32)
       return SelectPostLoadLane(Node, 1, AArch64::LD1i32_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
              VT == MVT::v1f64)
       return SelectPostLoadLane(Node, 1, AArch64::LD1i64_POST);
     break;
   }
   case AArch64ISD::LD2LANEpost: {
     if (VT == MVT::v16i8 || VT == MVT::v8i8)
       return SelectPostLoadLane(Node, 2, AArch64::LD2i8_POST);
-    else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+             VT == MVT::v8f16)
       return SelectPostLoadLane(Node, 2, AArch64::LD2i16_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
              VT == MVT::v2f32)
       return SelectPostLoadLane(Node, 2, AArch64::LD2i32_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
              VT == MVT::v1f64)
       return SelectPostLoadLane(Node, 2, AArch64::LD2i64_POST);
     break;
   }
   case AArch64ISD::LD3LANEpost: {
     if (VT == MVT::v16i8 || VT == MVT::v8i8)
       return SelectPostLoadLane(Node, 3, AArch64::LD3i8_POST);
-    else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+             VT == MVT::v8f16)
       return SelectPostLoadLane(Node, 3, AArch64::LD3i16_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
              VT == MVT::v2f32)
       return SelectPostLoadLane(Node, 3, AArch64::LD3i32_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
              VT == MVT::v1f64)
       return SelectPostLoadLane(Node, 3, AArch64::LD3i64_POST);
     break;
   }
   case AArch64ISD::LD4LANEpost: {
     if (VT == MVT::v16i8 || VT == MVT::v8i8)
       return SelectPostLoadLane(Node, 4, AArch64::LD4i8_POST);
-    else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+             VT == MVT::v8f16)
       return SelectPostLoadLane(Node, 4, AArch64::LD4i16_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
              VT == MVT::v2f32)
       return SelectPostLoadLane(Node, 4, AArch64::LD4i32_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
              VT == MVT::v1f64)
       return SelectPostLoadLane(Node, 4, AArch64::LD4i64_POST);
     break;
   }
   case AArch64ISD::ST2post: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v8i8)
       return SelectPostStore(Node, 2, AArch64::ST2Twov8b_POST);
     else if (VT == MVT::v16i8)
       return SelectPostStore(Node, 2, AArch64::ST2Twov16b_POST);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostStore(Node, 2, AArch64::ST2Twov4h_POST);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostStore(Node, 2, AArch64::ST2Twov8h_POST);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostStore(Node, 2, AArch64::ST2Twov2s_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostStore(Node, 2, AArch64::ST2Twov4s_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostStore(Node, 2, AArch64::ST2Twov2d_POST);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostStore(Node, 2, AArch64::ST1Twov1d_POST);
     break;
   }
   case AArch64ISD::ST3post: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v8i8)
       return SelectPostStore(Node, 3, AArch64::ST3Threev8b_POST);
     else if (VT == MVT::v16i8)
       return SelectPostStore(Node, 3, AArch64::ST3Threev16b_POST);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostStore(Node, 3, AArch64::ST3Threev4h_POST);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostStore(Node, 3, AArch64::ST3Threev8h_POST);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostStore(Node, 3, AArch64::ST3Threev2s_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostStore(Node, 3, AArch64::ST3Threev4s_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostStore(Node, 3, AArch64::ST3Threev2d_POST);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostStore(Node, 3, AArch64::ST1Threev1d_POST);
     break;
   }
   case AArch64ISD::ST4post: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v8i8)
       return SelectPostStore(Node, 4, AArch64::ST4Fourv8b_POST);
     else if (VT == MVT::v16i8)
       return SelectPostStore(Node, 4, AArch64::ST4Fourv16b_POST);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostStore(Node, 4, AArch64::ST4Fourv4h_POST);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostStore(Node, 4, AArch64::ST4Fourv8h_POST);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostStore(Node, 4, AArch64::ST4Fourv2s_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostStore(Node, 4, AArch64::ST4Fourv4s_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostStore(Node, 4, AArch64::ST4Fourv2d_POST);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv1d_POST);
     break;
   }
   case AArch64ISD::ST1x2post: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v8i8)
       return SelectPostStore(Node, 2, AArch64::ST1Twov8b_POST);
     else if (VT == MVT::v16i8)
       return SelectPostStore(Node, 2, AArch64::ST1Twov16b_POST);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostStore(Node, 2, AArch64::ST1Twov4h_POST);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostStore(Node, 2, AArch64::ST1Twov8h_POST);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostStore(Node, 2, AArch64::ST1Twov2s_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostStore(Node, 2, AArch64::ST1Twov4s_POST);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostStore(Node, 2, AArch64::ST1Twov1d_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostStore(Node, 2, AArch64::ST1Twov2d_POST);
     break;
   }
   case AArch64ISD::ST1x3post: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v8i8)
       return SelectPostStore(Node, 3, AArch64::ST1Threev8b_POST);
     else if (VT == MVT::v16i8)
       return SelectPostStore(Node, 3, AArch64::ST1Threev16b_POST);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostStore(Node, 3, AArch64::ST1Threev4h_POST);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostStore(Node, 3, AArch64::ST1Threev8h_POST);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostStore(Node, 3, AArch64::ST1Threev2s_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostStore(Node, 3, AArch64::ST1Threev4s_POST);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostStore(Node, 3, AArch64::ST1Threev1d_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostStore(Node, 3, AArch64::ST1Threev2d_POST);
     break;
   }
   case AArch64ISD::ST1x4post: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v8i8)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv8b_POST);
     else if (VT == MVT::v16i8)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv16b_POST);
-    else if (VT == MVT::v4i16)
+    else if (VT == MVT::v4i16 || VT == MVT::v4f16)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv4h_POST);
-    else if (VT == MVT::v8i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v8f16)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv8h_POST);
     else if (VT == MVT::v2i32 || VT == MVT::v2f32)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv2s_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v4f32)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv4s_POST);
     else if (VT == MVT::v1i64 || VT == MVT::v1f64)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv1d_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v2f64)
       return SelectPostStore(Node, 4, AArch64::ST1Fourv2d_POST);
     break;
   }
   case AArch64ISD::ST2LANEpost: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v16i8 || VT == MVT::v8i8)
       return SelectPostStoreLane(Node, 2, AArch64::ST2i8_POST);
-    else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+             VT == MVT::v8f16)
       return SelectPostStoreLane(Node, 2, AArch64::ST2i16_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
              VT == MVT::v2f32)
       return SelectPostStoreLane(Node, 2, AArch64::ST2i32_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
              VT == MVT::v1f64)
       return SelectPostStoreLane(Node, 2, AArch64::ST2i64_POST);
     break;
   }
   case AArch64ISD::ST3LANEpost: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v16i8 || VT == MVT::v8i8)
       return SelectPostStoreLane(Node, 3, AArch64::ST3i8_POST);
-    else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+             VT == MVT::v8f16)
       return SelectPostStoreLane(Node, 3, AArch64::ST3i16_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
              VT == MVT::v2f32)
       return SelectPostStoreLane(Node, 3, AArch64::ST3i32_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
              VT == MVT::v1f64)
       return SelectPostStoreLane(Node, 3, AArch64::ST3i64_POST);
     break;
   }
   case AArch64ISD::ST4LANEpost: {
     VT = Node->getOperand(1).getValueType();
     if (VT == MVT::v16i8 || VT == MVT::v8i8)
       return SelectPostStoreLane(Node, 4, AArch64::ST4i8_POST);
-    else if (VT == MVT::v8i16 || VT == MVT::v4i16)
+    else if (VT == MVT::v8i16 || VT == MVT::v4i16 || VT == MVT::v4f16 ||
+             VT == MVT::v8f16)
       return SelectPostStoreLane(Node, 4, AArch64::ST4i16_POST);
     else if (VT == MVT::v4i32 || VT == MVT::v2i32 || VT == MVT::v4f32 ||
              VT == MVT::v2f32)
       return SelectPostStoreLane(Node, 4, AArch64::ST4i32_POST);
     else if (VT == MVT::v2i64 || VT == MVT::v1i64 || VT == MVT::v2f64 ||
              VT == MVT::v1f64)
       return SelectPostStoreLane(Node, 4, AArch64::ST4i64_POST);
     break;

lib/Target/AArch64/AArch64ISelLowering.cpp

if (Subtarget->hasNEON()) {
     addRegisterClass(MVT::v8i16, &AArch64::FPR16RegClass);
     // Someone set us up the NEON.
     addDRTypeForNEON(MVT::v2f32);
     addDRTypeForNEON(MVT::v8i8);
     addDRTypeForNEON(MVT::v4i16);
     addDRTypeForNEON(MVT::v2i32);
     addDRTypeForNEON(MVT::v1i64);
     addDRTypeForNEON(MVT::v1f64);
+    addDRTypeForNEON(MVT::v4f16);
 
     addQRTypeForNEON(MVT::v4f32);
     addQRTypeForNEON(MVT::v2f64);
     addQRTypeForNEON(MVT::v16i8);
     addQRTypeForNEON(MVT::v8i16);
     addQRTypeForNEON(MVT::v4i32);
     addQRTypeForNEON(MVT::v2i64);
+    addQRTypeForNEON(MVT::v8f16);
   }
 
   // Compute derived properties from the register classes
   computeRegisterProperties();
 
   // Provide all sorts of operation actions
   setOperationAction(ISD::GlobalAddress, MVT::i64, Custom);
   setOperationAction(ISD::GlobalTLSAddress, MVT::i64, Custom);
AArch64TargetLowering::AArch64TargetLowering(TargetMachine &TM)
   // valid, and ignore them otherwise. The operations not mentioned here will
   // fail to select, but this is not a major problem as no source language
   // should be emitting native f16 operations yet.
   setOperationAction(ISD::FADD, MVT::f16, Promote);
   setOperationAction(ISD::FDIV, MVT::f16, Promote);
   setOperationAction(ISD::FMUL, MVT::f16, Promote);
   setOperationAction(ISD::FSUB, MVT::f16, Promote);
 
+  // v4f16 is also a storage-only type, so promote it to v4f32 when that is
+  // known to be safe.
+  setOperationAction(ISD::FADD, MVT::v4f16, Promote);
+  setOperationAction(ISD::FSUB, MVT::v4f16, Promote);
+  setOperationAction(ISD::FMUL, MVT::v4f16, Promote);
+  setOperationAction(ISD::FDIV, MVT::v4f16, Promote);
+  setOperationAction(ISD::FP_EXTEND, MVT::v4f16, Promote);
+  setOperationAction(ISD::FP_ROUND, MVT::v4f16, Promote);
+  AddPromotedToType(ISD::FADD, MVT::v4f16, MVT::v4f32);
+  AddPromotedToType(ISD::FSUB, MVT::v4f16, MVT::v4f32);
+  AddPromotedToType(ISD::FMUL, MVT::v4f16, MVT::v4f32);
+  AddPromotedToType(ISD::FDIV, MVT::v4f16, MVT::v4f32);
+  AddPromotedToType(ISD::FP_EXTEND, MVT::v4f16, MVT::v4f32);
+  AddPromotedToType(ISD::FP_ROUND, MVT::v4f16, MVT::v4f32);
+
+  // Expand all other v4f16 operations.
+  // FIXME: We could generate better code by promoting some operations to
+  // a pair of v4f32s
+  setOperationAction(ISD::FABS, MVT::v4f16, Expand);
+  setOperationAction(ISD::FCEIL, MVT::v4f16, Expand);
+  setOperationAction(ISD::FCOPYSIGN, MVT::v4f16, Expand);
+  setOperationAction(ISD::FCOS, MVT::v4f16, Expand);
+  setOperationAction(ISD::FFLOOR, MVT::v4f16, Expand);
+  setOperationAction(ISD::FMA, MVT::v4f16, Expand);
+  setOperationAction(ISD::FNEARBYINT, MVT::v4f16, Expand);
+  setOperationAction(ISD::FNEG, MVT::v4f16, Expand);
+  setOperationAction(ISD::FPOW, MVT::v4f16, Expand);
+  setOperationAction(ISD::FPOWI, MVT::v4f16, Expand);
+  setOperationAction(ISD::FREM, MVT::v4f16, Expand);
+  setOperationAction(ISD::FROUND, MVT::v4f16, Expand);
+  setOperationAction(ISD::FRINT, MVT::v4f16, Expand);
+  setOperationAction(ISD::FSIN, MVT::v4f16, Expand);
+  setOperationAction(ISD::FSINCOS, MVT::v4f16, Expand);
+  setOperationAction(ISD::FSQRT, MVT::v4f16, Expand);
+  setOperationAction(ISD::FTRUNC, MVT::v4f16, Expand);
+  setOperationAction(ISD::SETCC, MVT::v4f16, Expand);
+  setOperationAction(ISD::BR_CC, MVT::v4f16, Expand);
+  setOperationAction(ISD::SELECT, MVT::v4f16, Expand);
+  setOperationAction(ISD::SELECT_CC, MVT::v4f16, Expand);
+  setOperationAction(ISD::FEXP, MVT::v4f16, Expand);
+  setOperationAction(ISD::FEXP2, MVT::v4f16, Expand);
+  setOperationAction(ISD::FLOG, MVT::v4f16, Expand);
+  setOperationAction(ISD::FLOG2, MVT::v4f16, Expand);
+  setOperationAction(ISD::FLOG10, MVT::v4f16, Expand);
+
+
+  // v8f16 is also a storage-only type, so expand it.
+  setOperationAction(ISD::FABS, MVT::v8f16, Expand);
+  setOperationAction(ISD::FADD, MVT::v8f16, Expand);
+  setOperationAction(ISD::FCEIL, MVT::v8f16, Expand);
+  setOperationAction(ISD::FCOPYSIGN, MVT::v8f16, Expand);
+  setOperationAction(ISD::FCOS, MVT::v8f16, Expand);
+  setOperationAction(ISD::FDIV, MVT::v8f16, Expand);
+  setOperationAction(ISD::FFLOOR, MVT::v8f16, Expand);
+  setOperationAction(ISD::FMA, MVT::v8f16, Expand);
+  setOperationAction(ISD::FMUL, MVT::v8f16, Expand);
+  setOperationAction(ISD::FNEARBYINT, MVT::v8f16, Expand);
+  setOperationAction(ISD::FNEG, MVT::v8f16, Expand);
+  setOperationAction(ISD::FPOW, MVT::v8f16, Expand);
+  setOperationAction(ISD::FPOWI, MVT::v8f16, Expand);
+  setOperationAction(ISD::FREM, MVT::v8f16, Expand);
+  setOperationAction(ISD::FROUND, MVT::v8f16, Expand);
+  setOperationAction(ISD::FRINT, MVT::v8f16, Expand);
+  setOperationAction(ISD::FSIN, MVT::v8f16, Expand);
+  setOperationAction(ISD::FSINCOS, MVT::v8f16, Expand);
+  setOperationAction(ISD::FSQRT, MVT::v8f16, Expand);
+  setOperationAction(ISD::FSUB, MVT::v8f16, Expand);
+  setOperationAction(ISD::FTRUNC, MVT::v8f16, Expand);
+  setOperationAction(ISD::SETCC, MVT::v8f16, Expand);
+  setOperationAction(ISD::BR_CC, MVT::v8f16, Expand);
+  setOperationAction(ISD::SELECT, MVT::v8f16, Expand);
+  setOperationAction(ISD::SELECT_CC, MVT::v8f16, Expand);
+  setOperationAction(ISD::FP_EXTEND, MVT::v8f16, Expand);
+  setOperationAction(ISD::FEXP, MVT::v8f16, Expand);
+  setOperationAction(ISD::FEXP2, MVT::v8f16, Expand);
+  setOperationAction(ISD::FLOG, MVT::v8f16, Expand);
+  setOperationAction(ISD::FLOG2, MVT::v8f16, Expand);
+  setOperationAction(ISD::FLOG10, MVT::v8f16, Expand);
+
   // AArch64 has implementations of a lot of rounding-like FP operations.
   static MVT RoundingTypes[] = { MVT::f32, MVT::f64};
   for (unsigned I = 0; I < array_lengthof(RoundingTypes); ++I) {
     MVT Ty = RoundingTypes[I];
     setOperationAction(ISD::FFLOOR, Ty, Legal);
     setOperationAction(ISD::FNEARBYINT, Ty, Legal);
     setOperationAction(ISD::FCEIL, Ty, Legal);
     setOperationAction(ISD::FRINT, Ty, Legal);
if (VT.getSizeInBits() < InVT.getSizeInBits()) {
     SDValue Cv =
         DAG.getNode(Op.getOpcode(), dl, InVT.changeVectorElementTypeToInteger(),
                     Op.getOperand(0));
     return DAG.getNode(ISD::TRUNCATE, dl, VT, Cv);
   }
 
   if (VT.getSizeInBits() > InVT.getSizeInBits()) {
     SDLoc dl(Op);
-    SDValue Ext = DAG.getNode(ISD::FP_EXTEND, dl, MVT::v2f64, Op.getOperand(0));
+    MVT ExtVT =
+        MVT::getVectorVT(MVT::getFloatingPointVT(VT.getScalarSizeInBits()),
+                         VT.getVectorNumElements());
+    SDValue Ext = DAG.getNode(ISD::FP_EXTEND, dl, ExtVT, Op.getOperand(0));
     return DAG.getNode(Op.getOpcode(), dl, VT, Ext);
   }
 
   // Type changing conversions are illegal.
   return Op;
 }
 
 SDValue AArch64TargetLowering::LowerFP_TO_INT(SDValue Op,
static SDValue GeneratePerfectShuffle(unsigned PFEntry, SDValue LHS,
   default:
     llvm_unreachable("Unknown shuffle opcode!");
   case OP_VREV:
     // VREV divides the vector in half and swaps within the half.
     if (VT.getVectorElementType() == MVT::i32 ||
         VT.getVectorElementType() == MVT::f32)
       return DAG.getNode(AArch64ISD::REV64, dl, VT, OpLHS);
     // vrev <4 x i16> -> REV32
-    if (VT.getVectorElementType() == MVT::i16)
+    if (VT.getVectorElementType() == MVT::i16 ||
+        VT.getVectorElementType() == MVT::f16)
       return DAG.getNode(AArch64ISD::REV32, dl, VT, OpLHS);
     // vrev <4 x i8> -> REV16
     assert(VT.getVectorElementType() == MVT::i8);
     return DAG.getNode(AArch64ISD::REV16, dl, VT, OpLHS);
   case OP_VDUP0:
   case OP_VDUP1:
   case OP_VDUP2:
   case OP_VDUP3: {
if (V2.getNode()->getOpcode() == ISD::UNDEF) {
     }
   }
   return DAG.getNode(ISD::BITCAST, DL, Op.getValueType(), Shuffle);
 }
 
 static unsigned getDUPLANEOp(EVT EltType) {
   if (EltType == MVT::i8)
     return AArch64ISD::DUPLANE8;
-  if (EltType == MVT::i16)
+  if (EltType == MVT::i16 || EltType == MVT::f16)
     return AArch64ISD::DUPLANE16;
   if (EltType == MVT::i32 || EltType == MVT::f32)
     return AArch64ISD::DUPLANE32;
   if (EltType == MVT::i64 || EltType == MVT::f64)
     return AArch64ISD::DUPLANE64;
 
   llvm_unreachable("Invalid vector element type?");
 }
if (isINSMask(ShuffleMask, NumInputElements, DstIsLeft, Anomaly)) {
     int SrcLane = ShuffleMask[Anomaly];
     if (SrcLane >= NumInputElements) {
       SrcVec = V2;
       SrcLane -= VT.getVectorNumElements();
     }
     SDValue SrcLaneV = DAG.getConstant(SrcLane, MVT::i64);
 
     EVT ScalarVT = VT.getVectorElementType();
-    if (ScalarVT.getSizeInBits() < 32)
+
+    if (ScalarVT.getSizeInBits() < 32 && ScalarVT.isInteger())
       ScalarVT = MVT::i32;
 
     return DAG.getNode(
         ISD::INSERT_VECTOR_ELT, dl, VT, DstVec,
         DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, ScalarVT, SrcVec, SrcLaneV),
         DstLaneV);
   }
 
SDValue AArch64TargetLowering::LowerINSERT_VECTOR_ELT(SDValue Op,
   EVT VT = Op.getOperand(0).getValueType();
   ConstantSDNode *CI = dyn_cast<ConstantSDNode>(Op.getOperand(2));
   if (!CI || CI->getZExtValue() >= VT.getVectorNumElements())
     return SDValue();
 
 
   // Insertion/extraction are legal for V128 types.
   if (VT == MVT::v16i8 || VT == MVT::v8i16 || VT == MVT::v4i32 ||
-      VT == MVT::v2i64 || VT == MVT::v4f32 || VT == MVT::v2f64)
+      VT == MVT::v2i64 || VT == MVT::v4f32 || VT == MVT::v2f64 ||
+      VT == MVT::v8f16)
     return Op;
 
   if (VT != MVT::v8i8 && VT != MVT::v4i16 && VT != MVT::v2i32 &&
-      VT != MVT::v1i64 && VT != MVT::v2f32)
+      VT != MVT::v1i64 && VT != MVT::v2f32 && VT != MVT::v4f16)
     return SDValue();
 
   // For V64 types, we perform insertion by expanding the value
   // to a V128 type and perform the insertion on that.
   SDLoc DL(Op);
   SDValue WideVec = WidenVector(Op.getOperand(0), DAG);
   EVT WideTy = WideVec.getValueType();
 
AArch64TargetLowering::LowerEXTRACT_VECTOR_ELT(SDValue Op,
   EVT VT = Op.getOperand(0).getValueType();
   ConstantSDNode *CI = dyn_cast<ConstantSDNode>(Op.getOperand(1));
   if (!CI || CI->getZExtValue() >= VT.getVectorNumElements())
     return SDValue();
 
 
   // Insertion/extraction are legal for V128 types.
   if (VT == MVT::v16i8 || VT == MVT::v8i16 || VT == MVT::v4i32 ||
-      VT == MVT::v2i64 || VT == MVT::v4f32 || VT == MVT::v2f64)
+      VT == MVT::v2i64 || VT == MVT::v4f32 || VT == MVT::v2f64 ||
+      VT == MVT::v8f16)
     return Op;
 
   if (VT != MVT::v8i8 && VT != MVT::v4i16 && VT != MVT::v2i32 &&
-      VT != MVT::v1i64 && VT != MVT::v2f32)
+      VT != MVT::v1i64 && VT != MVT::v2f32 && VT != MVT::v4f16)
     return SDValue();
 
   // For V64 types, we perform extraction by expanding the value
   // to a V128 type and perform the extraction on that.
   SDLoc DL(Op);
   SDValue WideVec = WidenVector(Op.getOperand(0), DAG);
   EVT WideTy = WideVec.getValueType();
 

lib/Target/AArch64/AArch64InstrFormats.td

def v8i16  : BaseSIMDZipVector<0b011, opc, V128,
       asm, ".8h", OpNode, v8i16>;
   def v2i32  : BaseSIMDZipVector<0b100, opc, V64,
       asm, ".2s", OpNode, v2i32>;
   def v4i32  : BaseSIMDZipVector<0b101, opc, V128,
       asm, ".4s", OpNode, v4i32>;
   def v2i64  : BaseSIMDZipVector<0b111, opc, V128,
       asm, ".2d", OpNode, v2i64>;
 
+  def : Pat<(v4f16 (OpNode V64:$Rn, V64:$Rm)),
+        (!cast<Instruction>(NAME#"v4i16") V64:$Rn, V64:$Rm)>;
+  def : Pat<(v8f16 (OpNode V128:$Rn, V128:$Rm)),
+        (!cast<Instruction>(NAME#"v8i16") V128:$Rn, V128:$Rm)>;
   def : Pat<(v2f32 (OpNode V64:$Rn, V64:$Rm)),
         (!cast<Instruction>(NAME#"v2i32") V64:$Rn, V64:$Rm)>;
   def : Pat<(v4f32 (OpNode V128:$Rn, V128:$Rm)),
         (!cast<Instruction>(NAME#"v4i32") V128:$Rn, V128:$Rm)>;
   def : Pat<(v2f64 (OpNode V128:$Rn, V128:$Rm)),
         (!cast<Instruction>(NAME#"v2i64") V128:$Rn, V128:$Rm)>;
 }
 

lib/Target/AArch64/AArch64InstrInfo.td

 
 let AddedComplexity = 10 in {
 defm : ScalToVecROLoadPat<ro8,  extloadi8,  i32, v8i8,  LDRBroW, LDRBroX, bsub>;
 defm : ScalToVecROLoadPat<ro8,  extloadi8,  i32, v16i8, LDRBroW, LDRBroX, bsub>;
 
 defm : ScalToVecROLoadPat<ro16, extloadi16, i32, v4i16, LDRHroW, LDRHroX, hsub>;
 defm : ScalToVecROLoadPat<ro16, extloadi16, i32, v8i16, LDRHroW, LDRHroX, hsub>;
 
+defm : ScalToVecROLoadPat<ro16, load,       i32, v4f16, LDRHroW, LDRHroX, hsub>;
+defm : ScalToVecROLoadPat<ro16, load,       i32, v8f16, LDRHroW, LDRHroX, hsub>;
+
 defm : ScalToVecROLoadPat<ro32, load,       i32, v2i32, LDRSroW, LDRSroX, ssub>;
 defm : ScalToVecROLoadPat<ro32, load,       i32, v4i32, LDRSroW, LDRSroX, ssub>;
 
 defm : ScalToVecROLoadPat<ro32, load,       f32, v2f32, LDRSroW, LDRSroX, ssub>;
 defm : ScalToVecROLoadPat<ro32, load,       f32, v4f32, LDRSroW, LDRSroX, ssub>;
 
 defm : ScalToVecROLoadPat<ro64, load,       i64, v2i64, LDRDroW, LDRDroX, dsub>;
 

 
 let AddedComplexity = 10 in {
 let Predicates = [IsLE] in {
   // We must do vector loads with LD1 in big-endian.
   defm : VecROLoadPat<ro64, v2i32, LDRDroW, LDRDroX>;
   defm : VecROLoadPat<ro64, v2f32, LDRDroW, LDRDroX>;
   defm : VecROLoadPat<ro64, v8i8,  LDRDroW, LDRDroX>;
   defm : VecROLoadPat<ro64, v4i16, LDRDroW, LDRDroX>;
+  defm : VecROLoadPat<ro64, v4f16, LDRDroW, LDRDroX>;
 }
 
 defm : VecROLoadPat<ro64, v1i64,  LDRDroW, LDRDroX>;
 defm : VecROLoadPat<ro64, v1f64,  LDRDroW, LDRDroX>;
 
 // Match all load 128 bits width whose type is compatible with FPR128
 let Predicates = [IsLE] in {
   // We must do vector loads with LD1 in big-endian.
   defm : VecROLoadPat<ro128, v2i64,  LDRQroW, LDRQroX>;
   defm : VecROLoadPat<ro128, v2f64,  LDRQroW, LDRQroX>;
   defm : VecROLoadPat<ro128, v4i32,  LDRQroW, LDRQroX>;
   defm : VecROLoadPat<ro128, v4f32,  LDRQroW, LDRQroX>;
   defm : VecROLoadPat<ro128, v8i16,  LDRQroW, LDRQroX>;
+  defm : VecROLoadPat<ro128, v8f16,  LDRQroW, LDRQroX>;
   defm : VecROLoadPat<ro128, v16i8,  LDRQroW, LDRQroX>;
 }
 } // AddedComplexity = 10
 
 // zextload -> i64
 multiclass ExtLoadTo64ROPat<ROAddrMode ro, SDPatternOperator loadop,
                             Instruction INSTW, Instruction INSTX> {
   def : Pat<(i64 (loadop (ro.Wpat GPR64sp:$Rn, GPR32:$Rm, ro.Wext:$extend))),
let Predicates = [IsLE] in {
   def : Pat<(v2f32 (load (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))),
             (LDRDui GPR64sp:$Rn, uimm12s8:$offset)>;
   def : Pat<(v8i8 (load (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))),
             (LDRDui GPR64sp:$Rn, uimm12s8:$offset)>;
   def : Pat<(v4i16 (load (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))),
             (LDRDui GPR64sp:$Rn, uimm12s8:$offset)>;
   def : Pat<(v2i32 (load (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))),
             (LDRDui GPR64sp:$Rn, uimm12s8:$offset)>;
+  def : Pat<(v4f16 (load (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))),
+            (LDRDui GPR64sp:$Rn, uimm12s8:$offset)>;
 }
 def : Pat<(v1f64 (load (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))),
           (LDRDui GPR64sp:$Rn, uimm12s8:$offset)>;
 def : Pat<(v1i64 (load (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))),
           (LDRDui GPR64sp:$Rn, uimm12s8:$offset)>;
 
 // Match all load 128 bits width whose type is compatible with FPR128
 let Predicates = [IsLE] in {
   // We must use LD1 to perform vector loads in big-endian.
   def : Pat<(v4f32 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
             (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
   def : Pat<(v2f64 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
             (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
   def : Pat<(v16i8 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
             (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
   def : Pat<(v8i16 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
             (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
   def : Pat<(v4i32 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
             (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
   def : Pat<(v2i64 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
             (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
+  def : Pat<(v8f16 (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
+            (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
 }
 def : Pat<(f128  (load (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset))),
           (LDRQui GPR64sp:$Rn, uimm12s16:$offset)>;
 
 defm LDRHH : LoadUI<0b01, 0, 0b01, GPR32, uimm12s2, "ldrh",
                     [(set GPR32:$Rt,
                           (zextloadi16 (am_indexed16 GPR64sp:$Rn,
                                                      uimm12s2:$offset)))]>;
let Predicates = [IsLE] in {
   def : Pat<(v2f32 (load (am_unscaled64 GPR64sp:$Rn, simm9:$offset))),
             (LDURDi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v2i32 (load (am_unscaled64 GPR64sp:$Rn, simm9:$offset))),
             (LDURDi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v4i16 (load (am_unscaled64 GPR64sp:$Rn, simm9:$offset))),
             (LDURDi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v8i8 (load (am_unscaled64 GPR64sp:$Rn, simm9:$offset))),
             (LDURDi GPR64sp:$Rn, simm9:$offset)>;
+  def : Pat<(v4f16 (load (am_unscaled64 GPR64sp:$Rn, simm9:$offset))),
+            (LDURDi GPR64sp:$Rn, simm9:$offset)>;
 }
 def : Pat<(v1f64 (load (am_unscaled64 GPR64sp:$Rn, simm9:$offset))),
           (LDURDi GPR64sp:$Rn, simm9:$offset)>;
 def : Pat<(v1i64 (load (am_unscaled64 GPR64sp:$Rn, simm9:$offset))),
           (LDURDi GPR64sp:$Rn, simm9:$offset)>;
 
 // Match all load 128 bits width whose type is compatible with FPR128
 let Predicates = [IsLE] in {
   def : Pat<(v2f64 (load (am_unscaled128 GPR64sp:$Rn, simm9:$offset))),
             (LDURQi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v2i64 (load (am_unscaled128 GPR64sp:$Rn, simm9:$offset))),
             (LDURQi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v4f32 (load (am_unscaled128 GPR64sp:$Rn, simm9:$offset))),
             (LDURQi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v4i32 (load (am_unscaled128 GPR64sp:$Rn, simm9:$offset))),
             (LDURQi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v8i16 (load (am_unscaled128 GPR64sp:$Rn, simm9:$offset))),
             (LDURQi GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(v16i8 (load (am_unscaled128 GPR64sp:$Rn, simm9:$offset))),
             (LDURQi GPR64sp:$Rn, simm9:$offset)>;
+  def : Pat<(v8f16 (load (am_unscaled128 GPR64sp:$Rn, simm9:$offset))),
+            (LDURQi GPR64sp:$Rn, simm9:$offset)>;
 }
 
 //  anyext -> zext
 def : Pat<(i32 (extloadi16 (am_unscaled16 GPR64sp:$Rn, simm9:$offset))),
           (LDURHHi GPR64sp:$Rn, simm9:$offset)>;
 def : Pat<(i32 (extloadi8 (am_unscaled8 GPR64sp:$Rn, simm9:$offset))),
           (LDURBBi GPR64sp:$Rn, simm9:$offset)>;
 def : Pat<(i32 (extloadi1 (am_unscaled8 GPR64sp:$Rn, simm9:$offset))),

 let AddedComplexity = 10 in {
 // Match all store 64 bits width whose type is compatible with FPR64
 let Predicates = [IsLE] in {
   // We must use ST1 to store vectors in big-endian.
   defm : VecROStorePat<ro64, v2i32, FPR64, STRDroW, STRDroX>;
   defm : VecROStorePat<ro64, v2f32, FPR64, STRDroW, STRDroX>;
   defm : VecROStorePat<ro64, v4i16, FPR64, STRDroW, STRDroX>;
   defm : VecROStorePat<ro64, v8i8, FPR64, STRDroW, STRDroX>;
+  defm : VecROStorePat<ro64, v4f16, FPR64, STRDroW, STRDroX>;
 }
 
 defm : VecROStorePat<ro64, v1i64, FPR64, STRDroW, STRDroX>;
 defm : VecROStorePat<ro64, v1f64, FPR64, STRDroW, STRDroX>;
 
 // Match all store 128 bits width whose type is compatible with FPR128
 let Predicates = [IsLE] in {
   // We must use ST1 to store vectors in big-endian.
   defm : VecROStorePat<ro128, v2i64, FPR128, STRQroW, STRQroX>;
   defm : VecROStorePat<ro128, v2f64, FPR128, STRQroW, STRQroX>;
   defm : VecROStorePat<ro128, v4i32, FPR128, STRQroW, STRQroX>;
   defm : VecROStorePat<ro128, v4f32, FPR128, STRQroW, STRQroX>;
   defm : VecROStorePat<ro128, v8i16, FPR128, STRQroW, STRQroX>;
   defm : VecROStorePat<ro128, v16i8, FPR128, STRQroW, STRQroX>;
+  defm : VecROStorePat<ro128, v8f16, FPR128, STRQroW, STRQroX>;
 }
 } // AddedComplexity = 10
 
 //---
 // (unsigned immediate)
 defm STRX : StoreUI<0b11, 0, 0b00, GPR64, uimm12s8, "str",
                    [(store GPR64:$Rt,
                             (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset))]>;
def : Pat<(store (v8i8 FPR64:$Rt),
                    (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset)),
             (STRDui FPR64:$Rt, GPR64sp:$Rn, uimm12s8:$offset)>;
   def : Pat<(store (v4i16 FPR64:$Rt),
                    (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset)),
             (STRDui FPR64:$Rt, GPR64sp:$Rn, uimm12s8:$offset)>;
   def : Pat<(store (v2i32 FPR64:$Rt),
                    (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset)),
             (STRDui FPR64:$Rt, GPR64sp:$Rn, uimm12s8:$offset)>;
+  def : Pat<(store (v4f16 FPR64:$Rt),
+                   (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset)),
+            (STRDui FPR64:$Rt, GPR64sp:$Rn, uimm12s8:$offset)>;
 }
 def : Pat<(store (v1f64 FPR64:$Rt),
                  (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset)),
           (STRDui FPR64:$Rt, GPR64sp:$Rn, uimm12s8:$offset)>;
 def : Pat<(store (v1i64 FPR64:$Rt),
                  (am_indexed64 GPR64sp:$Rn, uimm12s8:$offset)),
           (STRDui FPR64:$Rt, GPR64sp:$Rn, uimm12s8:$offset)>;
 
def : Pat<(store (v8i16 FPR128:$Rt),
                    (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset)),
             (STRQui FPR128:$Rt, GPR64sp:$Rn, uimm12s16:$offset)>;
   def : Pat<(store (v4i32 FPR128:$Rt),
                    (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset)),
             (STRQui FPR128:$Rt, GPR64sp:$Rn, uimm12s16:$offset)>;
   def : Pat<(store (v2i64 FPR128:$Rt),
                    (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset)),
             (STRQui FPR128:$Rt, GPR64sp:$Rn, uimm12s16:$offset)>;
+  def : Pat<(store (v8f16 FPR128:$Rt),
+                   (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset)),
+            (STRQui FPR128:$Rt, GPR64sp:$Rn, uimm12s16:$offset)>;
 }
 def : Pat<(store (f128  FPR128:$Rt),
                  (am_indexed128 GPR64sp:$Rn, uimm12s16:$offset)),
           (STRQui FPR128:$Rt, GPR64sp:$Rn, uimm12s16:$offset)>;
 
 // truncstore i64
 def : Pat<(truncstorei32 GPR64:$Rt,
                          (am_indexed32 GPR64sp:$Rn, uimm12s4:$offset)),
def : Pat<(store (v8i8 FPR64:$Rt),
                    (am_unscaled64 GPR64sp:$Rn, simm9:$offset)),
             (STURDi FPR64:$Rt, GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(store (v4i16 FPR64:$Rt),
                    (am_unscaled64 GPR64sp:$Rn, simm9:$offset)),
             (STURDi FPR64:$Rt, GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(store (v2i32 FPR64:$Rt),
                    (am_unscaled64 GPR64sp:$Rn, simm9:$offset)),
             (STURDi FPR64:$Rt, GPR64sp:$Rn, simm9:$offset)>;
+  def : Pat<(store (v4f16 FPR64:$Rt),
+                   (am_unscaled64 GPR64sp:$Rn, simm9:$offset)),
+            (STURDi FPR64:$Rt, GPR64sp:$Rn, simm9:$offset)>;
 }
 def : Pat<(store (v1f64 FPR64:$Rt), (am_unscaled64 GPR64sp:$Rn, simm9:$offset)),
           (STURDi FPR64:$Rt, GPR64sp:$Rn, simm9:$offset)>;
 def : Pat<(store (v1i64 FPR64:$Rt), (am_unscaled64 GPR64sp:$Rn, simm9:$offset)),
           (STURDi FPR64:$Rt, GPR64sp:$Rn, simm9:$offset)>;
 
 // Match all store 128 bits width whose type is compatible with FPR128
 let Predicates = [IsLE] in {
def : Pat<(store (v4i32 FPR128:$Rt),
                    (am_unscaled128 GPR64sp:$Rn, simm9:$offset)),
             (STURQi FPR128:$Rt, GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(store (v2i64 FPR128:$Rt),
                    (am_unscaled128 GPR64sp:$Rn, simm9:$offset)),
             (STURQi FPR128:$Rt, GPR64sp:$Rn, simm9:$offset)>;
   def : Pat<(store (v2f64 FPR128:$Rt),
                    (am_unscaled128 GPR64sp:$Rn, simm9:$offset)),
             (STURQi FPR128:$Rt, GPR64sp:$Rn, simm9:$offset)>;
+  def : Pat<(store (v8f16 FPR128:$Rt),
+                   (am_unscaled128 GPR64sp:$Rn, simm9:$offset)),
+            (STURQi FPR128:$Rt, GPR64sp:$Rn, simm9:$offset)>;
 }
 
 // unscaled i64 truncating stores
 def : Pat<(truncstorei32 GPR64:$Rt, (am_unscaled32 GPR64sp:$Rn, simm9:$offset)),
   (STURWi (EXTRACT_SUBREG GPR64:$Rt, sub_32), GPR64sp:$Rn, simm9:$offset)>;
 def : Pat<(truncstorei16 GPR64:$Rt, (am_unscaled16 GPR64sp:$Rn, simm9:$offset)),
   (STURHHi (EXTRACT_SUBREG GPR64:$Rt, sub_32), GPR64sp:$Rn, simm9:$offset)>;
 def : Pat<(truncstorei8 GPR64:$Rt, (am_unscaled8 GPR64sp:$Rn, simm9:$offset)),

 def : Pat<(pre_store (v2i32 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpre FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v2f32 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpre FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v1i64 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpre FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v1f64 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpre FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
+def : Pat<(pre_store (v4f16 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
+          (STRDpre FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 
 def : Pat<(pre_store (v16i8 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpre FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v8i16 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpre FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v4i32 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpre FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v4f32 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpre FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v2i64 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpre FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(pre_store (v2f64 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpre FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
+def : Pat<(pre_store (v8f16 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
+          (STRQpre FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 
 //---
 // (immediate post-indexed)
 def STRWpost : StorePostIdx<0b10, 0, 0b00, GPR32,  "str", post_store, i32>;
 def STRXpost : StorePostIdx<0b11, 0, 0b00, GPR64,  "str", post_store, i64>;
 def STRBpost : StorePostIdx<0b00, 1, 0b00, FPR8,   "str", post_store, untyped>;
 def STRHpost : StorePostIdx<0b01, 1, 0b00, FPR16,  "str", post_store, f16>;
 def STRSpost : StorePostIdx<0b10, 1, 0b00, FPR32,  "str", post_store, f32>;

 def : Pat<(post_store (v2i32 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpost FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v2f32 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpost FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v1i64 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpost FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v1f64 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
           (STRDpost FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
+def : Pat<(post_store (v4f16 FPR64:$Rt), GPR64sp:$addr, simm9:$off),
+          (STRDpost FPR64:$Rt, GPR64sp:$addr, simm9:$off)>;
 
 def : Pat<(post_store (v16i8 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpost FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v8i16 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpost FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v4i32 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpost FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v4f32 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpost FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v2i64 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpost FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 def : Pat<(post_store (v2f64 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
           (STRQpost FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
+def : Pat<(post_store (v8f16 FPR128:$Rt), GPR64sp:$addr, simm9:$off),
+          (STRQpost FPR128:$Rt, GPR64sp:$addr, simm9:$off)>;
 
 //===----------------------------------------------------------------------===//
 // Load/store exclusive instructions.
 //===----------------------------------------------------------------------===//
 
 def LDARW  : LoadAcquire   <0b10, 1, 1, 0, 1, GPR32, "ldar">;
 def LDARX  : LoadAcquire   <0b11, 1, 1, 0, 1, GPR64, "ldar">;
 def LDARB  : LoadAcquire   <0b00, 1, 1, 0, 1, GPR32, "ldarb">;

 def : Pat<(v4f32 (int_aarch64_neon_vcvthf2fp (extract_subvector (v8i16 V128:$Rn),
                                                               (i64 4)))),
           (FCVTLv8i16 V128:$Rn)>;
 def : Pat<(v2f64 (fextend (v2f32 V64:$Rn))), (FCVTLv2i32 V64:$Rn)>;
 def : Pat<(v2f64 (fextend (v2f32 (extract_subvector (v4f32 V128:$Rn),
                                                     (i64 2))))),
           (FCVTLv4i32 V128:$Rn)>;
 
+def : Pat<(v4f32 (fextend (v4f16 V64:$Rn))), (FCVTLv4i16 V64:$Rn)>;
+def : Pat<(v4f32 (fextend (v4f16 (extract_subvector (v8f16 V128:$Rn),
+                                                    (i64 4))))),
+          (FCVTLv8i16 V128:$Rn)>;
+
 defm FCVTMS : SIMDTwoVectorFPToInt<0,0,0b11011, "fcvtms",int_aarch64_neon_fcvtms>;
 defm FCVTMU : SIMDTwoVectorFPToInt<1,0,0b11011, "fcvtmu",int_aarch64_neon_fcvtmu>;
 defm FCVTNS : SIMDTwoVectorFPToInt<0,0,0b11010, "fcvtns",int_aarch64_neon_fcvtns>;
 defm FCVTNU : SIMDTwoVectorFPToInt<1,0,0b11010, "fcvtnu",int_aarch64_neon_fcvtnu>;
 defm FCVTN  : SIMDFPNarrowTwoVector<0, 0, 0b10110, "fcvtn">;
 def : Pat<(v4i16 (int_aarch64_neon_vcvtfp2hf (v4f32 V128:$Rn))),
           (FCVTNv4i16 V128:$Rn)>;
 def : Pat<(concat_vectors V64:$Rd,
                           (v4i16 (int_aarch64_neon_vcvtfp2hf (v4f32 V128:$Rn)))),
           (FCVTNv8i16 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Rd, dsub), V128:$Rn)>;
 def : Pat<(v2f32 (fround (v2f64 V128:$Rn))), (FCVTNv2i32 V128:$Rn)>;
+def : Pat<(v4f16 (fround (v4f32 V128:$Rn))), (FCVTNv4i16 V128:$Rn)>;
 def : Pat<(concat_vectors V64:$Rd, (v2f32 (fround (v2f64 V128:$Rn)))),
           (FCVTNv4i32 (INSERT_SUBREG (IMPLICIT_DEF), V64:$Rd, dsub), V128:$Rn)>;
 defm FCVTPS : SIMDTwoVectorFPToInt<0,1,0b11010, "fcvtps",int_aarch64_neon_fcvtps>;
 defm FCVTPU : SIMDTwoVectorFPToInt<1,1,0b11010, "fcvtpu",int_aarch64_neon_fcvtpu>;
 defm FCVTXN : SIMDFPInexactCvtTwoVector<1, 0, 0b10110, "fcvtxn",
                                         int_aarch64_neon_fcvtxn>;
 defm FCVTZS : SIMDTwoVectorFPToInt<0, 1, 0b11011, "fcvtzs", fp_to_sint>;
 defm FCVTZU : SIMDTwoVectorFPToInt<1, 1, 0b11011, "fcvtzu", fp_to_uint>;
defm UADDLP : SIMDLongTwoVector<1, 0b00010, "uaddlp",
                     int_aarch64_neon_uaddlp>;
 defm UCVTF  : SIMDTwoVectorIntToFP<1, 0, 0b11101, "ucvtf", uint_to_fp>;
 defm UQXTN  : SIMDMixedTwoVector<1, 0b10100, "uqxtn", int_aarch64_neon_uqxtn>;
 defm URECPE : SIMDTwoVectorS<0, 1, 0b11100, "urecpe", int_aarch64_neon_urecpe>;
 defm URSQRTE: SIMDTwoVectorS<1, 1, 0b11100, "ursqrte", int_aarch64_neon_ursqrte>;
 defm USQADD : SIMDTwoVectorBHSDTied<1, 0b00011, "usqadd",int_aarch64_neon_usqadd>;
 defm XTN    : SIMDMixedTwoVector<0, 0b10010, "xtn", trunc>;
 
+def : Pat<(v4f16 (AArch64rev32 V64:$Rn)), (REV32v4i16 V64:$Rn)>;
+def : Pat<(v4f16 (AArch64rev64 V64:$Rn)), (REV64v4i16 V64:$Rn)>;
+def : Pat<(v8f16 (AArch64rev32 V128:$Rn)), (REV32v8i16 V128:$Rn)>;
+def : Pat<(v8f16 (AArch64rev64 V128:$Rn)), (REV64v8i16 V128:$Rn)>;
 def : Pat<(v2f32 (AArch64rev64 V64:$Rn)), (REV64v2i32 V64:$Rn)>;
 def : Pat<(v4f32 (AArch64rev64 V128:$Rn)), (REV64v4i32 V128:$Rn)>;
 
 // Patterns for vector long shift (by element width). These need to match all
 // three of zext, sext and anyext so it's easier to pull the patterns out of the
 // definition.
 multiclass SIMDVectorLShiftLongBySizeBHSPats<SDPatternOperator ext> {
   def : Pat<(AArch64vshl (v8i16 (ext (v8i8 V64:$Rn))), (i32 8)),

 def : Pat<(v4i32 (AArch64ext V128:$Rn, V128:$Rm, (i32 imm:$imm))),
           (EXTv16i8 V128:$Rn, V128:$Rm, imm:$imm)>;
 def : Pat<(v4f32 (AArch64ext V128:$Rn, V128:$Rm, (i32 imm:$imm))),
           (EXTv16i8 V128:$Rn, V128:$Rm, imm:$imm)>;
 def : Pat<(v2i64 (AArch64ext V128:$Rn, V128:$Rm, (i32 imm:$imm))),
           (EXTv16i8 V128:$Rn, V128:$Rm, imm:$imm)>;
 def : Pat<(v2f64 (AArch64ext V128:$Rn, V128:$Rm, (i32 imm:$imm))),
           (EXTv16i8 V128:$Rn, V128:$Rm, imm:$imm)>;
+def : Pat<(v4f16 (AArch64ext V64:$Rn, V64:$Rm, (i32 imm:$imm))),
+          (EXTv8i8 V64:$Rn, V64:$Rm, imm:$imm)>;
+def : Pat<(v8f16 (AArch64ext V128:$Rn, V128:$Rm, (i32 imm:$imm))),
+          (EXTv16i8 V128:$Rn, V128:$Rm, imm:$imm)>;
 
 // We use EXT to handle extract_subvector to copy the upper 64-bits of a
 // 128-bit vector.
 def : Pat<(v8i8  (extract_subvector V128:$Rn, (i64 8))),
           (EXTRACT_SUBREG (EXTv16i8 V128:$Rn, V128:$Rn, 8), dsub)>;
 def : Pat<(v4i16 (extract_subvector V128:$Rn, (i64 4))),
           (EXTRACT_SUBREG (EXTv16i8 V128:$Rn, V128:$Rn, 8), dsub)>;
 def : Pat<(v2i32 (extract_subvector V128:$Rn, (i64 2))),
           (EXTRACT_SUBREG (EXTv16i8 V128:$Rn, V128:$Rn, 8), dsub)>;
 def : Pat<(v1i64 (extract_subvector V128:$Rn, (i64 1))),
           (EXTRACT_SUBREG (EXTv16i8 V128:$Rn, V128:$Rn, 8), dsub)>;
+def : Pat<(v4f16 (extract_subvector V128:$Rn, (i64 4))),
+          (EXTRACT_SUBREG (EXTv16i8 V128:$Rn, V128:$Rn, 8), dsub)>;
 def : Pat<(v2f32 (extract_subvector V128:$Rn, (i64 2))),
           (EXTRACT_SUBREG (EXTv16i8 V128:$Rn, V128:$Rn, 8), dsub)>;
 def : Pat<(v1f64 (extract_subvector V128:$Rn, (i64 1))),
           (EXTRACT_SUBREG (EXTv16i8 V128:$Rn, V128:$Rn, 8), dsub)>;
 
 
 //----------------------------------------------------------------------------
 // AdvSIMD zip vector

 def : Pat<(v4f32 (AArch64dup (f32 FPR32:$Rn))),
           (v4f32 (DUPv4i32lane
             (INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), FPR32:$Rn, ssub),
             (i64 0)))>;
 def : Pat<(v2f64 (AArch64dup (f64 FPR64:$Rn))),
           (v2f64 (DUPv2i64lane
             (INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), FPR64:$Rn, dsub),
             (i64 0)))>;
+def : Pat<(v4f16 (AArch64dup (f16 FPR16:$Rn))),
+          (v4f16 (DUPv4i16lane
+            (INSERT_SUBREG (v8i16 (IMPLICIT_DEF)), FPR16:$Rn, hsub),
+            (i64 0)))>;
+def : Pat<(v8f16 (AArch64dup (f16 FPR16:$Rn))),
+          (v8f16 (DUPv8i16lane
+            (INSERT_SUBREG (v8i16 (IMPLICIT_DEF)), FPR16:$Rn, hsub),
+            (i64 0)))>;
+
+def : Pat<(v4f16 (AArch64duplane16 (v8f16 V128:$Rn), VectorIndexH:$imm)),
+          (DUPv4i16lane V128:$Rn, VectorIndexH:$imm)>;
+def : Pat<(v8f16 (AArch64duplane16 (v8f16 V128:$Rn), VectorIndexH:$imm)),
+          (DUPv8i16lane V128:$Rn, VectorIndexH:$imm)>;
 
 def : Pat<(v2f32 (AArch64duplane32 (v4f32 V128:$Rn), VectorIndexS:$imm)),
           (DUPv2i32lane V128:$Rn, VectorIndexS:$imm)>;
 def : Pat<(v4f32 (AArch64duplane32 (v4f32 V128:$Rn), VectorIndexS:$imm)),
          (DUPv4i32lane V128:$Rn, VectorIndexS:$imm)>;
 def : Pat<(v2f64 (AArch64duplane64 (v2f64 V128:$Rn), VectorIndexD:$imm)),
           (DUPv2i64lane V128:$Rn, VectorIndexD:$imm)>;
 

 
 def : Pat<(v4f32 (scalar_to_vector (f32 FPR32:$Rn))),
           (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FPR32:$Rn, ssub)>;
 def : Pat<(v2f32 (scalar_to_vector (f32 FPR32:$Rn))),
           (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)), FPR32:$Rn, ssub)>;
 def : Pat<(v2f64 (scalar_to_vector (f64 FPR64:$Rn))),
           (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FPR64:$Rn, dsub)>;
 
+def : Pat<(v4f16 (vector_insert (v4f16 V64:$Rn),
+            (f16 FPR16:$Rm), (i64 VectorIndexS:$imm))),
+          (EXTRACT_SUBREG
+            (INSvi16lane
+              (v8f16 (INSERT_SUBREG (v8f16 (IMPLICIT_DEF)), V64:$Rn, dsub)),
+              VectorIndexS:$imm,
+              (v8f16 (INSERT_SUBREG (v8f16 (IMPLICIT_DEF)), FPR16:$Rm, hsub)),
+              (i64 0)),
+            dsub)>;
+
+def : Pat<(v8f16 (vector_insert (v8f16 V128:$Rn),
+            (f16 FPR16:$Rm), (i64 VectorIndexH:$imm))),
+          (INSvi16lane
+            V128:$Rn, VectorIndexH:$imm,
+            (v8f16 (INSERT_SUBREG (v8f16 (IMPLICIT_DEF)), FPR16:$Rm, hsub)),
+            (i64 0))>;
+
 def : Pat<(v2f32 (vector_insert (v2f32 V64:$Rn),
             (f32 FPR32:$Rm), (i64 VectorIndexS:$imm))),
           (EXTRACT_SUBREG
             (INSvi32lane
               (v4f32 (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), V64:$Rn, dsub)),
               VectorIndexS:$imm,
               (v4f32 (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FPR32:$Rm, ssub)),
               (i64 0)),
def : Pat<(VT64 (vector_insert V64:$src,
                         (VTScal (vector_extract (VT64 V64:$Rn), imm:$Immn)),
                         imm:$Immd)),
             (EXTRACT_SUBREG
                 (INS (SUBREG_TO_REG (i64 0), V64:$src, dsub), imm:$Immd,
                      (SUBREG_TO_REG (i64 0), V64:$Rn, dsub), imm:$Immn),
                 dsub)>;
 }
 
+defm : Neon_INS_elt_pattern<v8f16, v4f16, f16, INSvi16lane>;
 defm : Neon_INS_elt_pattern<v4f32, v2f32, f32, INSvi32lane>;
 defm : Neon_INS_elt_pattern<v2f64, v1f64, f64, INSvi64lane>;
 defm : Neon_INS_elt_pattern<v16i8, v8i8,  i32, INSvi8lane>;
 defm : Neon_INS_elt_pattern<v8i16, v4i16, i32, INSvi16lane>;
 defm : Neon_INS_elt_pattern<v4i32, v2i32, i32, INSvi32lane>;
 defm : Neon_INS_elt_pattern<v2i64, v1i64, i64, INSvi32lane>;
 
 
 // Floating point vector extractions are codegen'd as either a sequence of
 // subregister extractions, possibly fed by an INS if the lane number is
 // anything other than zero.
 def : Pat<(vector_extract (v2f64 V128:$Rn), 0),
           (f64 (EXTRACT_SUBREG V128:$Rn, dsub))>;
 def : Pat<(vector_extract (v4f32 V128:$Rn), 0),
           (f32 (EXTRACT_SUBREG V128:$Rn, ssub))>;
+def : Pat<(vector_extract (v8f16 V128:$Rn), 0),
+          (f16 (EXTRACT_SUBREG V128:$Rn, hsub))>;
 def : Pat<(vector_extract (v2f64 V128:$Rn), VectorIndexD:$idx),
           (f64 (EXTRACT_SUBREG
             (INSvi64lane (v2f64 (IMPLICIT_DEF)), 0,
                          V128:$Rn, VectorIndexD:$idx),
             dsub))>;
 def : Pat<(vector_extract (v4f32 V128:$Rn), VectorIndexS:$idx),
           (f32 (EXTRACT_SUBREG
             (INSvi32lane (v4f32 (IMPLICIT_DEF)), 0,
                          V128:$Rn, VectorIndexS:$idx),
             ssub))>;
+def : Pat<(vector_extract (v8f16 V128:$Rn), VectorIndexH:$idx),
+          (f16 (EXTRACT_SUBREG
+            (INSvi16lane (v8f16 (IMPLICIT_DEF)), 0,
+                         V128:$Rn, VectorIndexH:$idx),
+            hsub))>;
 
 // All concat_vectors operations are canonicalised to act on i64 vectors for
 // AArch64. In the general case we need an instruction, which had just as well be
 // INS.
 class ConcatPat<ValueType DstTy, ValueType SrcTy>
   : Pat<(DstTy (concat_vectors (SrcTy V64:$Rd), V64:$Rn)),
         (INSvi64lane (INSERT_SUBREG (IMPLICIT_DEF), V64:$Rd, dsub), 1,
                      (INSERT_SUBREG (IMPLICIT_DEF), V64:$Rn, dsub), 0)>;
 
 def : ConcatPat<v2i64, v1i64>;
 def : ConcatPat<v2f64, v1f64>;
 def : ConcatPat<v4i32, v2i32>;
 def : ConcatPat<v4f32, v2f32>;
 def : ConcatPat<v8i16, v4i16>;
+def : ConcatPat<v8f16, v4f16>;
 def : ConcatPat<v16i8, v8i8>;
 
 // If the high lanes are undef, though, we can just ignore them:
 class ConcatUndefPat<ValueType DstTy, ValueType SrcTy>
   : Pat<(DstTy (concat_vectors (SrcTy V64:$Rn), undef)),
         (INSERT_SUBREG (IMPLICIT_DEF), V64:$Rn, dsub)>;
 
 def : ConcatUndefPat<v2i64, v1i64>;

 def : Pat<(v2f32 (AArch64dup (f32 (load GPR64sp:$Rn)))),
           (LD1Rv2s GPR64sp:$Rn)>;
 def : Pat<(v4f32 (AArch64dup (f32 (load GPR64sp:$Rn)))),
           (LD1Rv4s GPR64sp:$Rn)>;
 def : Pat<(v2f64 (AArch64dup (f64 (load GPR64sp:$Rn)))),
           (LD1Rv2d GPR64sp:$Rn)>;
 def : Pat<(v1f64 (AArch64dup (f64 (load GPR64sp:$Rn)))),
           (LD1Rv1d GPR64sp:$Rn)>;
+def : Pat<(v4f16 (AArch64dup (f16 (load GPR64sp:$Rn)))),
+          (LD1Rv4h GPR64sp:$Rn)>;
+def : Pat<(v8f16 (AArch64dup (f16 (load GPR64sp:$Rn)))),
+          (LD1Rv8h GPR64sp:$Rn)>;
 
 class Ld1Lane128Pat<SDPatternOperator scalar_load, Operand VecIndex,
                     ValueType VTy, ValueType STy, Instruction LD1>
   : Pat<(vector_insert (VTy VecListOne128:$Rd),
            (STy (scalar_load GPR64sp:$Rn)), VecIndex:$idx),
         (LD1 VecListOne128:$Rd, VecIndex:$idx, GPR64sp:$Rn)>;
 
 def : Ld1Lane128Pat<extloadi8,  VectorIndexB, v16i8, i32, LD1i8>;
 def : Ld1Lane128Pat<extloadi16, VectorIndexH, v8i16, i32, LD1i16>;
 def : Ld1Lane128Pat<load,       VectorIndexS, v4i32, i32, LD1i32>;
 def : Ld1Lane128Pat<load,       VectorIndexS, v4f32, f32, LD1i32>;
 def : Ld1Lane128Pat<load,       VectorIndexD, v2i64, i64, LD1i64>;
 def : Ld1Lane128Pat<load,       VectorIndexD, v2f64, f64, LD1i64>;
+def : Ld1Lane128Pat<load,       VectorIndexH, v8f16, f16, LD1i16>;
 
 class Ld1Lane64Pat<SDPatternOperator scalar_load, Operand VecIndex,
                    ValueType VTy, ValueType STy, Instruction LD1>
   : Pat<(vector_insert (VTy VecListOne64:$Rd),
            (STy (scalar_load GPR64sp:$Rn)), VecIndex:$idx),
         (EXTRACT_SUBREG
             (LD1 (SUBREG_TO_REG (i32 0), VecListOne64:$Rd, dsub),
                           VecIndex:$idx, GPR64sp:$Rn),
             dsub)>;
 
 def : Ld1Lane64Pat<extloadi8,  VectorIndexB, v8i8,  i32, LD1i8>;
 def : Ld1Lane64Pat<extloadi16, VectorIndexH, v4i16, i32, LD1i16>;
 def : Ld1Lane64Pat<load,       VectorIndexS, v2i32, i32, LD1i32>;
 def : Ld1Lane64Pat<load,       VectorIndexS, v2f32, f32, LD1i32>;
+def : Ld1Lane64Pat<load,       VectorIndexH, v4f16, f16, LD1i16>;
 
 
 defm LD1 : SIMDLdSt1SingleAliases<"ld1">;
 defm LD2 : SIMDLdSt2SingleAliases<"ld2">;
 defm LD3 : SIMDLdSt3SingleAliases<"ld3">;
 defm LD4 : SIMDLdSt4SingleAliases<"ld4">;
 
 // Stores
: Pat<(scalar_store
         (ST1 VecListOne128:$Vt, VecIndex:$idx, GPR64sp:$Rn)>;
 
 def : St1Lane128Pat<truncstorei8,  VectorIndexB, v16i8, i32, ST1i8>;
 def : St1Lane128Pat<truncstorei16, VectorIndexH, v8i16, i32, ST1i16>;
 def : St1Lane128Pat<store,         VectorIndexS, v4i32, i32, ST1i32>;
 def : St1Lane128Pat<store,         VectorIndexS, v4f32, f32, ST1i32>;
 def : St1Lane128Pat<store,         VectorIndexD, v2i64, i64, ST1i64>;
 def : St1Lane128Pat<store,         VectorIndexD, v2f64, f64, ST1i64>;
+def : St1Lane128Pat<store,         VectorIndexH, v8f16, f16, ST1i16>;
 
 let AddedComplexity = 15 in
 class St1Lane64Pat<SDPatternOperator scalar_store, Operand VecIndex,
                    ValueType VTy, ValueType STy, Instruction ST1>
   : Pat<(scalar_store
              (STy (vector_extract (VTy VecListOne64:$Vt), VecIndex:$idx)),
              GPR64sp:$Rn),
         (ST1 (SUBREG_TO_REG (i32 0), VecListOne64:$Vt, dsub),
              VecIndex:$idx, GPR64sp:$Rn)>;
 
 def : St1Lane64Pat<truncstorei8,  VectorIndexB, v8i8, i32, ST1i8>;
 def : St1Lane64Pat<truncstorei16, VectorIndexH, v4i16, i32, ST1i16>;
 def : St1Lane64Pat<store,         VectorIndexS, v2i32, i32, ST1i32>;
 def : St1Lane64Pat<store,         VectorIndexS, v2f32, f32, ST1i32>;
+def : St1Lane64Pat<store,         VectorIndexH, v4f16, f16, ST1i16>;
 
 multiclass St1LanePost64Pat<SDPatternOperator scalar_store, Operand VecIndex,
                              ValueType VTy, ValueType STy, Instruction ST1,
                              int offset> {
   def : Pat<(scalar_store
               (STy (vector_extract (VTy VecListOne64:$Vt), VecIndex:$idx)),
               GPR64sp:$Rn, offset),
         (ST1 (SUBREG_TO_REG (i32 0), VecListOne64:$Vt, dsub),
              VecIndex:$idx, GPR64sp:$Rn, XZR)>;
 
   def : Pat<(scalar_store
               (STy (vector_extract (VTy VecListOne64:$Vt), VecIndex:$idx)),
               GPR64sp:$Rn, GPR64:$Rm),
         (ST1 (SUBREG_TO_REG (i32 0), VecListOne64:$Vt, dsub),
              VecIndex:$idx, GPR64sp:$Rn, $Rm)>;
 }
 
 defm : St1LanePost64Pat<post_truncsti8, VectorIndexB, v8i8, i32, ST1i8_POST, 1>;
 defm : St1LanePost64Pat<post_truncsti16, VectorIndexH, v4i16, i32, ST1i16_POST,
                         2>;
 defm : St1LanePost64Pat<post_store, VectorIndexS, v2i32, i32, ST1i32_POST, 4>;
 defm : St1LanePost64Pat<post_store, VectorIndexS, v2f32, f32, ST1i32_POST, 4>;
 defm : St1LanePost64Pat<post_store, VectorIndexD, v1i64, i64, ST1i64_POST, 8>;
 defm : St1LanePost64Pat<post_store, VectorIndexD, v1f64, f64, ST1i64_POST, 8>;
+defm : St1LanePost64Pat<post_store, VectorIndexH, v4f16, f16, ST1i16_POST, 2>;
 
 multiclass St1LanePost128Pat<SDPatternOperator scalar_store, Operand VecIndex,
                              ValueType VTy, ValueType STy, Instruction ST1,
                              int offset> {
   def : Pat<(scalar_store
               (STy (vector_extract (VTy VecListOne128:$Vt), VecIndex:$idx)),
               GPR64sp:$Rn, offset),
         (ST1 VecListOne128:$Vt, VecIndex:$idx, GPR64sp:$Rn, XZR)>;
 
   def : Pat<(scalar_store
               (STy (vector_extract (VTy VecListOne128:$Vt), VecIndex:$idx)),
               GPR64sp:$Rn, GPR64:$Rm),
         (ST1 VecListOne128:$Vt, VecIndex:$idx, GPR64sp:$Rn, $Rm)>;
 }
 
 defm : St1LanePost128Pat<post_truncsti8, VectorIndexB, v16i8, i32, ST1i8_POST,
                          1>;
 defm : St1LanePost128Pat<post_truncsti16, VectorIndexH, v8i16, i32, ST1i16_POST,
                          2>;
 defm : St1LanePost128Pat<post_store, VectorIndexS, v4i32, i32, ST1i32_POST, 4>;
 defm : St1LanePost128Pat<post_store, VectorIndexS, v4f32, f32, ST1i32_POST, 4>;
 defm : St1LanePost128Pat<post_store, VectorIndexD, v2i64, i64, ST1i64_POST, 8>;
 defm : St1LanePost128Pat<post_store, VectorIndexD, v2f64, f64, ST1i64_POST, 8>;
+defm : St1LanePost128Pat<post_store, VectorIndexH, v8f16, f16, ST1i16_POST, 2>;
 
 let mayStore = 1, neverHasSideEffects = 1 in {
 defm ST2 : SIMDStSingleB<1, 0b000,       "st2", VecListTwob,   GPR64pi2>;
 defm ST2 : SIMDStSingleH<1, 0b010, 0,    "st2", VecListTwoh,   GPR64pi4>;
 defm ST2 : SIMDStSingleS<1, 0b100, 0b00, "st2", VecListTwos,   GPR64pi8>;
 defm ST2 : SIMDStSingleD<1, 0b100, 0b01, "st2", VecListTwod,   GPR64pi16>;
 defm ST3 : SIMDStSingleB<0, 0b001,       "st3", VecListThreeb, GPR64pi3>;
 defm ST3 : SIMDStSingleH<0, 0b011, 0,    "st3", VecListThreeh, GPR64pi6>;

 //   a) Identity conversions -  vNfX <-> vNiX
 //   b) Single-lane-to-scalar - v1fX <-> fX or v1iX <-> iX
 //
 
 let Predicates = [IsLE] in {
 def : Pat<(v8i8  (bitconvert GPR64:$Xn)), (COPY_TO_REGCLASS GPR64:$Xn, FPR64)>;
 def : Pat<(v4i16 (bitconvert GPR64:$Xn)), (COPY_TO_REGCLASS GPR64:$Xn, FPR64)>;
 def : Pat<(v2i32 (bitconvert GPR64:$Xn)), (COPY_TO_REGCLASS GPR64:$Xn, FPR64)>;
+def : Pat<(v4f16 (bitconvert GPR64:$Xn)), (COPY_TO_REGCLASS GPR64:$Xn, FPR64)>;
 def : Pat<(v2f32 (bitconvert GPR64:$Xn)), (COPY_TO_REGCLASS GPR64:$Xn, FPR64)>;
 
 def : Pat<(i64 (bitconvert (v8i8  V64:$Vn))),
           (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
 def : Pat<(i64 (bitconvert (v4i16 V64:$Vn))),
           (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
 def : Pat<(i64 (bitconvert (v2i32 V64:$Vn))),
           (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
+def : Pat<(i64 (bitconvert (v4f16 V64:$Vn))),
+          (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
 def : Pat<(i64 (bitconvert (v2f32 V64:$Vn))),
           (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
 def : Pat<(i64 (bitconvert (v1f64 V64:$Vn))),
           (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v8i8  (bitconvert GPR64:$Xn)),
                  (REV64v8i8 (COPY_TO_REGCLASS GPR64:$Xn, FPR64))>;
 def : Pat<(v4i16 (bitconvert GPR64:$Xn)),
                  (REV64v4i16 (COPY_TO_REGCLASS GPR64:$Xn, FPR64))>;
 def : Pat<(v2i32 (bitconvert GPR64:$Xn)),
                  (REV64v2i32 (COPY_TO_REGCLASS GPR64:$Xn, FPR64))>;
+def : Pat<(v4f16 (bitconvert GPR64:$Xn)),
+                 (REV64v4i16 (COPY_TO_REGCLASS GPR64:$Xn, FPR64))>;
 def : Pat<(v2f32 (bitconvert GPR64:$Xn)),
                  (REV64v2i32 (COPY_TO_REGCLASS GPR64:$Xn, FPR64))>;
 
 def : Pat<(i64 (bitconvert (v8i8  V64:$Vn))),
           (REV64v8i8 (COPY_TO_REGCLASS V64:$Vn, GPR64))>;
 def : Pat<(i64 (bitconvert (v4i16 V64:$Vn))),
           (REV64v4i16 (COPY_TO_REGCLASS V64:$Vn, GPR64))>;
 def : Pat<(i64 (bitconvert (v2i32 V64:$Vn))),
           (REV64v2i32 (COPY_TO_REGCLASS V64:$Vn, GPR64))>;
+def : Pat<(i64 (bitconvert (v4f16 V64:$Vn))),
+          (REV64v4i16 (COPY_TO_REGCLASS V64:$Vn, GPR64))>;
 def : Pat<(i64 (bitconvert (v2f32 V64:$Vn))),
           (REV64v2i32 (COPY_TO_REGCLASS V64:$Vn, GPR64))>;
 }
 def : Pat<(v1i64 (bitconvert GPR64:$Xn)), (COPY_TO_REGCLASS GPR64:$Xn, FPR64)>;
 def : Pat<(v1f64 (bitconvert GPR64:$Xn)), (COPY_TO_REGCLASS GPR64:$Xn, FPR64)>;
 def : Pat<(i64 (bitconvert (v1i64 V64:$Vn))),
           (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
 def : Pat<(v1i64 (scalar_to_vector GPR64:$Xn)),
def : Pat<(i64 (bitconvert (f64 FPR64:$Xn))),
           (COPY_TO_REGCLASS FPR64:$Xn, GPR64)>;
 def : Pat<(i64 (bitconvert (v1f64 V64:$Vn))),
           (COPY_TO_REGCLASS V64:$Vn, GPR64)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v1i64 (bitconvert (v2i32 FPR64:$src))), (v1i64 FPR64:$src)>;
 def : Pat<(v1i64 (bitconvert (v4i16 FPR64:$src))), (v1i64 FPR64:$src)>;
 def : Pat<(v1i64 (bitconvert (v8i8  FPR64:$src))), (v1i64 FPR64:$src)>;
+def : Pat<(v1i64 (bitconvert (v4f16 FPR64:$src))), (v1i64 FPR64:$src)>;
 def : Pat<(v1i64 (bitconvert (v2f32 FPR64:$src))), (v1i64 FPR64:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v1i64 (bitconvert (v2i32 FPR64:$src))),
                              (v1i64 (REV64v2i32 FPR64:$src))>;
 def : Pat<(v1i64 (bitconvert (v4i16 FPR64:$src))),
                              (v1i64 (REV64v4i16 FPR64:$src))>;
 def : Pat<(v1i64 (bitconvert (v8i8  FPR64:$src))),
                              (v1i64 (REV64v8i8 FPR64:$src))>;
+def : Pat<(v1i64 (bitconvert (v4f16 FPR64:$src))),
+                             (v1i64 (REV64v4i16 FPR64:$src))>;
 def : Pat<(v1i64 (bitconvert (v2f32 FPR64:$src))),
                              (v1i64 (REV64v2i32 FPR64:$src))>;
 }
 def : Pat<(v1i64 (bitconvert (v1f64 FPR64:$src))), (v1i64 FPR64:$src)>;
 def : Pat<(v1i64 (bitconvert (f64   FPR64:$src))), (v1i64 FPR64:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v2i32 (bitconvert (v1i64 FPR64:$src))), (v2i32 FPR64:$src)>;
 def : Pat<(v2i32 (bitconvert (v4i16 FPR64:$src))), (v2i32 FPR64:$src)>;
 def : Pat<(v2i32 (bitconvert (v8i8  FPR64:$src))), (v2i32 FPR64:$src)>;
 def : Pat<(v2i32 (bitconvert (f64   FPR64:$src))), (v2i32 FPR64:$src)>;
 def : Pat<(v2i32 (bitconvert (v1f64 FPR64:$src))), (v2i32 FPR64:$src)>;
+def : Pat<(v2i32 (bitconvert (v4f16 FPR64:$src))), (v2i32 FPR64:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v2i32 (bitconvert (v1i64 FPR64:$src))),
                              (v2i32 (REV64v2i32 FPR64:$src))>;
 def : Pat<(v2i32 (bitconvert (v4i16 FPR64:$src))),
                              (v2i32 (REV32v4i16 FPR64:$src))>;
 def : Pat<(v2i32 (bitconvert (v8i8  FPR64:$src))),
                              (v2i32 (REV32v8i8 FPR64:$src))>;
 def : Pat<(v2i32 (bitconvert (f64   FPR64:$src))),
                              (v2i32 (REV64v2i32 FPR64:$src))>;
 def : Pat<(v2i32 (bitconvert (v1f64 FPR64:$src))),
                              (v2i32 (REV64v2i32 FPR64:$src))>;
+def : Pat<(v2i32 (bitconvert (v4f16 FPR64:$src))),
+                             (v2i32 (REV64v4i16 FPR64:$src))>;
 }
 def : Pat<(v2i32 (bitconvert (v2f32 FPR64:$src))), (v2i32 FPR64:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v4i16 (bitconvert (v1i64 FPR64:$src))), (v4i16 FPR64:$src)>;
 def : Pat<(v4i16 (bitconvert (v2i32 FPR64:$src))), (v4i16 FPR64:$src)>;
 def : Pat<(v4i16 (bitconvert (v8i8  FPR64:$src))), (v4i16 FPR64:$src)>;
 def : Pat<(v4i16 (bitconvert (f64   FPR64:$src))), (v4i16 FPR64:$src)>;
+def : Pat<(v4i16 (bitconvert (v4f16 FPR64:$src))), (v4i16 FPR64:$src)>;
 def : Pat<(v4i16 (bitconvert (v2f32 FPR64:$src))), (v4i16 FPR64:$src)>;
 def : Pat<(v4i16 (bitconvert (v1f64 FPR64:$src))), (v4i16 FPR64:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v4i16 (bitconvert (v1i64 FPR64:$src))),
                              (v4i16 (REV64v4i16 FPR64:$src))>;
 def : Pat<(v4i16 (bitconvert (v2i32 FPR64:$src))),
                              (v4i16 (REV32v4i16 FPR64:$src))>;
 def : Pat<(v4i16 (bitconvert (v8i8  FPR64:$src))),
                              (v4i16 (REV16v8i8 FPR64:$src))>;
 def : Pat<(v4i16 (bitconvert (f64   FPR64:$src))),
                              (v4i16 (REV64v4i16 FPR64:$src))>;
+def : Pat<(v4i16 (bitconvert (v4f16 FPR64:$src))),
+                             (v4i16 (REV32v4i16 FPR64:$src))>;
 def : Pat<(v4i16 (bitconvert (v2f32 FPR64:$src))),
                              (v4i16 (REV32v4i16 FPR64:$src))>;
 def : Pat<(v4i16 (bitconvert (v1f64 FPR64:$src))),
                              (v4i16 (REV64v4i16 FPR64:$src))>;
 }
 
 let Predicates = [IsLE] in {
+def : Pat<(v4f16 (bitconvert (v1i64 FPR64:$src))), (v4f16 FPR64:$src)>;
+def : Pat<(v4f16 (bitconvert (v2i32 FPR64:$src))), (v4f16 FPR64:$src)>;
+def : Pat<(v4f16 (bitconvert (v4i16 FPR64:$src))), (v4f16 FPR64:$src)>;
+def : Pat<(v4f16 (bitconvert (v8i8  FPR64:$src))), (v4f16 FPR64:$src)>;
+def : Pat<(v4f16 (bitconvert (f64   FPR64:$src))), (v4f16 FPR64:$src)>;
+def : Pat<(v4f16 (bitconvert (v2f32 FPR64:$src))), (v4f16 FPR64:$src)>;
+def : Pat<(v4f16 (bitconvert (v1f64 FPR64:$src))), (v4f16 FPR64:$src)>;
+}
+let Predicates = [IsBE] in {
+def : Pat<(v4f16 (bitconvert (v1i64 FPR64:$src))),
+                             (v4f16 (REV64v4i16 FPR64:$src))>;
+def : Pat<(v4f16 (bitconvert (v2i32 FPR64:$src))),
+                             (v4f16 (REV64v4i16 FPR64:$src))>;
+def : Pat<(v4f16 (bitconvert (v4i16 FPR64:$src))),
+                             (v4f16 (REV64v4i16 FPR64:$src))>;
+def : Pat<(v4f16 (bitconvert (v8i8  FPR64:$src))),
+                             (v4f16 (REV16v8i8 FPR64:$src))>;
+def : Pat<(v4f16 (bitconvert (f64   FPR64:$src))),
+                             (v4f16 (REV64v4i16 FPR64:$src))>;
+def : Pat<(v4f16 (bitconvert (v2f32 FPR64:$src))),
+                             (v4f16 (REV64v4i16 FPR64:$src))>;
+def : Pat<(v4f16 (bitconvert (v1f64 FPR64:$src))),
+                             (v4f16 (REV64v4i16 FPR64:$src))>;
+}
+
+
+
+let Predicates = [IsLE] in {
 def : Pat<(v8i8  (bitconvert (v1i64 FPR64:$src))), (v8i8  FPR64:$src)>;
 def : Pat<(v8i8  (bitconvert (v2i32 FPR64:$src))), (v8i8  FPR64:$src)>;
 def : Pat<(v8i8  (bitconvert (v4i16 FPR64:$src))), (v8i8  FPR64:$src)>;
 def : Pat<(v8i8  (bitconvert (f64   FPR64:$src))), (v8i8  FPR64:$src)>;
 def : Pat<(v8i8  (bitconvert (v2f32 FPR64:$src))), (v8i8  FPR64:$src)>;
 def : Pat<(v8i8  (bitconvert (v1f64 FPR64:$src))), (v8i8  FPR64:$src)>;
+def : Pat<(v8i8  (bitconvert (v4f16 FPR64:$src))), (v8i8  FPR64:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v8i8  (bitconvert (v1i64 FPR64:$src))),
                              (v8i8 (REV64v8i8 FPR64:$src))>;
 def : Pat<(v8i8  (bitconvert (v2i32 FPR64:$src))),
                              (v8i8 (REV32v8i8 FPR64:$src))>;
 def : Pat<(v8i8  (bitconvert (v4i16 FPR64:$src))),
                              (v8i8 (REV16v8i8 FPR64:$src))>;
 def : Pat<(v8i8  (bitconvert (f64   FPR64:$src))),
                              (v8i8 (REV64v8i8 FPR64:$src))>;
 def : Pat<(v8i8  (bitconvert (v2f32 FPR64:$src))),
                              (v8i8 (REV32v8i8 FPR64:$src))>;
 def : Pat<(v8i8  (bitconvert (v1f64 FPR64:$src))),
                              (v8i8 (REV64v8i8 FPR64:$src))>;
+def : Pat<(v8i8  (bitconvert (v4f16 FPR64:$src))),
+                             (v8i8 (REV16v8i8 FPR64:$src))>;
 }
 
 let Predicates = [IsLE] in {
 def : Pat<(f64   (bitconvert (v2i32 FPR64:$src))), (f64   FPR64:$src)>;
 def : Pat<(f64   (bitconvert (v4i16 FPR64:$src))), (f64   FPR64:$src)>;
 def : Pat<(f64   (bitconvert (v2f32 FPR64:$src))), (f64   FPR64:$src)>;
 def : Pat<(f64   (bitconvert (v8i8  FPR64:$src))), (f64   FPR64:$src)>;
+def : Pat<(f64   (bitconvert (v4f16 FPR64:$src))), (f64   FPR64:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(f64   (bitconvert (v2i32 FPR64:$src))),
                              (f64 (REV64v2i32 FPR64:$src))>;
 def : Pat<(f64   (bitconvert (v4i16 FPR64:$src))),
                              (f64 (REV64v4i16 FPR64:$src))>;
 def : Pat<(f64   (bitconvert (v2f32 FPR64:$src))),
                              (f64 (REV64v2i32 FPR64:$src))>;
 def : Pat<(f64   (bitconvert (v8i8  FPR64:$src))),
                              (f64 (REV64v8i8 FPR64:$src))>;
+def : Pat<(f64   (bitconvert (v4f16 FPR64:$src))),
+                             (f64 (REV64v4i16 FPR64:$src))>;
 }
 def : Pat<(f64   (bitconvert (v1i64 FPR64:$src))), (f64   FPR64:$src)>;
 def : Pat<(f64   (bitconvert (v1f64 FPR64:$src))), (f64   FPR64:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v1f64 (bitconvert (v2i32 FPR64:$src))), (v1f64 FPR64:$src)>;
 def : Pat<(v1f64 (bitconvert (v4i16 FPR64:$src))), (v1f64 FPR64:$src)>;
 def : Pat<(v1f64 (bitconvert (v8i8  FPR64:$src))), (v1f64 FPR64:$src)>;
 def : Pat<(v1f64 (bitconvert (v2f32 FPR64:$src))), (v1f64 FPR64:$src)>;
+def : Pat<(v1f64 (bitconvert (v4f16 FPR64:$src))), (v1f64 FPR64:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v1f64 (bitconvert (v2i32 FPR64:$src))),
                              (v1f64 (REV64v2i32 FPR64:$src))>;
 def : Pat<(v1f64 (bitconvert (v4i16 FPR64:$src))),
                              (v1f64 (REV64v4i16 FPR64:$src))>;
 def : Pat<(v1f64 (bitconvert (v8i8  FPR64:$src))),
                              (v1f64 (REV64v8i8 FPR64:$src))>;
 def : Pat<(v1f64 (bitconvert (v2f32 FPR64:$src))),
                              (v1f64 (REV64v2i32 FPR64:$src))>;
+def : Pat<(v1f64 (bitconvert (v4f16 FPR64:$src))),
+                             (v1f64 (REV64v4i16 FPR64:$src))>;
 }
 def : Pat<(v1f64 (bitconvert (v1i64 FPR64:$src))), (v1f64 FPR64:$src)>;
 def : Pat<(v1f64 (bitconvert (f64   FPR64:$src))), (v1f64 FPR64:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v2f32 (bitconvert (v1i64 FPR64:$src))), (v2f32 FPR64:$src)>;
 def : Pat<(v2f32 (bitconvert (v4i16 FPR64:$src))), (v2f32 FPR64:$src)>;
 def : Pat<(v2f32 (bitconvert (v8i8  FPR64:$src))), (v2f32 FPR64:$src)>;
 def : Pat<(v2f32 (bitconvert (v1f64 FPR64:$src))), (v2f32 FPR64:$src)>;
 def : Pat<(v2f32 (bitconvert (f64   FPR64:$src))), (v2f32 FPR64:$src)>;
+def : Pat<(v2f32 (bitconvert (v4f16 FPR64:$src))), (v2f32 FPR64:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v2f32 (bitconvert (v1i64 FPR64:$src))),
                              (v2f32 (REV64v2i32 FPR64:$src))>;
 def : Pat<(v2f32 (bitconvert (v4i16 FPR64:$src))),
                              (v2f32 (REV32v4i16 FPR64:$src))>;
 def : Pat<(v2f32 (bitconvert (v8i8  FPR64:$src))),
                              (v2f32 (REV32v8i8 FPR64:$src))>;
 def : Pat<(v2f32 (bitconvert (v1f64 FPR64:$src))),
                              (v2f32 (REV64v2i32 FPR64:$src))>;
 def : Pat<(v2f32 (bitconvert (f64   FPR64:$src))),
                              (v2f32 (REV64v2i32 FPR64:$src))>;
+def : Pat<(v2f32 (bitconvert (v4f16 FPR64:$src))),
+                             (v2f32 (REV64v4i16 FPR64:$src))>;
 }
 def : Pat<(v2f32 (bitconvert (v2i32 FPR64:$src))), (v2f32 FPR64:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(f128 (bitconvert (v2i64 FPR128:$src))), (f128 FPR128:$src)>;
 def : Pat<(f128 (bitconvert (v4i32 FPR128:$src))), (f128 FPR128:$src)>;
 def : Pat<(f128 (bitconvert (v8i16 FPR128:$src))), (f128 FPR128:$src)>;
 def : Pat<(f128 (bitconvert (v2f64 FPR128:$src))), (f128 FPR128:$src)>;
 def : Pat<(f128 (bitconvert (v4f32 FPR128:$src))), (f128 FPR128:$src)>;
+def : Pat<(f128 (bitconvert (v8f16 FPR128:$src))), (f128 FPR128:$src)>;
 def : Pat<(f128 (bitconvert (v16i8 FPR128:$src))), (f128 FPR128:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(f128 (bitconvert (v2i64 FPR128:$src))),
                             (f128 (EXTv16i8 FPR128:$src, FPR128:$src, (i32 8)))>;
 def : Pat<(f128 (bitconvert (v4i32 FPR128:$src))),
                             (f128 (EXTv16i8 (REV64v4i32 FPR128:$src),
                                             (REV64v4i32 FPR128:$src), (i32 8)))>;

 def : Pat<(v4f32 (bitconvert (v4i32 FPR128:$src))), (v4f32 FPR128:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v2i64 (bitconvert (f128  FPR128:$src))), (v2i64 FPR128:$src)>;
 def : Pat<(v2i64 (bitconvert (v4i32 FPR128:$src))), (v2i64 FPR128:$src)>;
 def : Pat<(v2i64 (bitconvert (v8i16 FPR128:$src))), (v2i64 FPR128:$src)>;
 def : Pat<(v2i64 (bitconvert (v16i8 FPR128:$src))), (v2i64 FPR128:$src)>;
 def : Pat<(v2i64 (bitconvert (v4f32 FPR128:$src))), (v2i64 FPR128:$src)>;
+def : Pat<(v2i64 (bitconvert (v8f16 FPR128:$src))), (v2i64 FPR128:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v2i64 (bitconvert (f128  FPR128:$src))),
                              (v2i64 (EXTv16i8 FPR128:$src,
                                               FPR128:$src, (i32 8)))>;
 def : Pat<(v2i64 (bitconvert (v4i32 FPR128:$src))),
                              (v2i64 (REV64v4i32 FPR128:$src))>;
 def : Pat<(v2i64 (bitconvert (v8i16 FPR128:$src))),
                              (v2i64 (REV64v8i16 FPR128:$src))>;
 def : Pat<(v2i64 (bitconvert (v16i8 FPR128:$src))),
                              (v2i64 (REV64v16i8 FPR128:$src))>;
 def : Pat<(v2i64 (bitconvert (v4f32 FPR128:$src))),
                              (v2i64 (REV64v4i32 FPR128:$src))>;
+def : Pat<(v2i64 (bitconvert (v8f16 FPR128:$src))),
+                             (v2i64 (REV64v8i16 FPR128:$src))>;
 }
 def : Pat<(v2i64 (bitconvert (v2f64 FPR128:$src))), (v2i64 FPR128:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v4i32 (bitconvert (f128  FPR128:$src))), (v4i32 FPR128:$src)>;
 def : Pat<(v4i32 (bitconvert (v2i64 FPR128:$src))), (v4i32 FPR128:$src)>;
 def : Pat<(v4i32 (bitconvert (v8i16 FPR128:$src))), (v4i32 FPR128:$src)>;
 def : Pat<(v4i32 (bitconvert (v16i8 FPR128:$src))), (v4i32 FPR128:$src)>;
 def : Pat<(v4i32 (bitconvert (v2f64 FPR128:$src))), (v4i32 FPR128:$src)>;
+def : Pat<(v4i32 (bitconvert (v8f16 FPR128:$src))), (v4i32 FPR128:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v4i32 (bitconvert (f128  FPR128:$src))),
                              (v4i32 (EXTv16i8 (REV64v4i32 FPR128:$src),
                                               (REV64v4i32 FPR128:$src),
                                               (i32 8)))>;
 def : Pat<(v4i32 (bitconvert (v2i64 FPR128:$src))),
                              (v4i32 (REV64v4i32 FPR128:$src))>;
 def : Pat<(v4i32 (bitconvert (v8i16 FPR128:$src))),
                              (v4i32 (REV32v8i16 FPR128:$src))>;
 def : Pat<(v4i32 (bitconvert (v16i8 FPR128:$src))),
                              (v4i32 (REV32v16i8 FPR128:$src))>;
 def : Pat<(v4i32 (bitconvert (v2f64 FPR128:$src))),
                              (v4i32 (REV64v4i32 FPR128:$src))>;
+def : Pat<(v4i32 (bitconvert (v8f16 FPR128:$src))),
+                             (v4i32 (REV32v8i16 FPR128:$src))>;
 }
 def : Pat<(v4i32 (bitconvert (v4f32 FPR128:$src))), (v4i32 FPR128:$src)>;
 
 let Predicates = [IsLE] in {
 def : Pat<(v8i16 (bitconvert (f128  FPR128:$src))), (v8i16 FPR128:$src)>;
 def : Pat<(v8i16 (bitconvert (v2i64 FPR128:$src))), (v8i16 FPR128:$src)>;
 def : Pat<(v8i16 (bitconvert (v4i32 FPR128:$src))), (v8i16 FPR128:$src)>;
 def : Pat<(v8i16 (bitconvert (v16i8 FPR128:$src))), (v8i16 FPR128:$src)>;
 def : Pat<(v8i16 (bitconvert (v2f64 FPR128:$src))), (v8i16 FPR128:$src)>;
 def : Pat<(v8i16 (bitconvert (v4f32 FPR128:$src))), (v8i16 FPR128:$src)>;
+def : Pat<(v8i16 (bitconvert (v8f16 FPR128:$src))), (v8i16 FPR128:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v8i16 (bitconvert (f128  FPR128:$src))),
                              (v8i16 (EXTv16i8 (REV64v8i16 FPR128:$src),
                                               (REV64v8i16 FPR128:$src),
                                               (i32 8)))>;
 def : Pat<(v8i16 (bitconvert (v2i64 FPR128:$src))),
                              (v8i16 (REV64v8i16 FPR128:$src))>;
 def : Pat<(v8i16 (bitconvert (v4i32 FPR128:$src))),
                              (v8i16 (REV32v8i16 FPR128:$src))>;
 def : Pat<(v8i16 (bitconvert (v16i8 FPR128:$src))),
                              (v8i16 (REV16v16i8 FPR128:$src))>;
 def : Pat<(v8i16 (bitconvert (v2f64 FPR128:$src))),
                              (v8i16 (REV64v8i16 FPR128:$src))>;
 def : Pat<(v8i16 (bitconvert (v4f32 FPR128:$src))),
                              (v8i16 (REV32v8i16 FPR128:$src))>;
+def : Pat<(v8i16 (bitconvert (v8f16 FPR128:$src))),
+                             (v8i16 (REV32v8i16 FPR128:$src))>;
+}
+
+let Predicates = [IsLE] in {
+def : Pat<(v8f16 (bitconvert (f128  FPR128:$src))), (v8f16 FPR128:$src)>;
+def : Pat<(v8f16 (bitconvert (v2i64 FPR128:$src))), (v8f16 FPR128:$src)>;
+def : Pat<(v8f16 (bitconvert (v4i32 FPR128:$src))), (v8f16 FPR128:$src)>;
+def : Pat<(v8f16 (bitconvert (v8i16 FPR128:$src))), (v8f16 FPR128:$src)>;
+def : Pat<(v8f16 (bitconvert (v16i8 FPR128:$src))), (v8f16 FPR128:$src)>;
+def : Pat<(v8f16 (bitconvert (v2f64 FPR128:$src))), (v8f16 FPR128:$src)>;
+def : Pat<(v8f16 (bitconvert (v4f32 FPR128:$src))), (v8f16 FPR128:$src)>;
+}
+let Predicates = [IsBE] in {
+def : Pat<(v8f16 (bitconvert (f128  FPR128:$src))),
+                             (v8f16 (EXTv16i8 (REV64v8i16 FPR128:$src),
+                                              (REV64v8i16 FPR128:$src),
+                                              (i32 8)))>;
+def : Pat<(v8f16 (bitconvert (v2i64 FPR128:$src))),
+                             (v8f16 (REV64v8i16 FPR128:$src))>;
+def : Pat<(v8f16 (bitconvert (v4i32 FPR128:$src))),
+                             (v8f16 (REV32v8i16 FPR128:$src))>;
+def : Pat<(v8f16 (bitconvert (v8i16 FPR128:$src))),
+                             (v8f16 (REV64v8i16 FPR128:$src))>;
+def : Pat<(v8f16 (bitconvert (v16i8 FPR128:$src))),
+                             (v8f16 (REV16v16i8 FPR128:$src))>;
+def : Pat<(v8f16 (bitconvert (v2f64 FPR128:$src))),
+                             (v8f16 (REV64v8i16 FPR128:$src))>;
+def : Pat<(v8f16 (bitconvert (v4f32 FPR128:$src))),
+                             (v8f16 (REV32v8i16 FPR128:$src))>;
 }
 
 let Predicates = [IsLE] in {
 def : Pat<(v16i8 (bitconvert (f128  FPR128:$src))), (v16i8 FPR128:$src)>;
 def : Pat<(v16i8 (bitconvert (v2i64 FPR128:$src))), (v16i8 FPR128:$src)>;
 def : Pat<(v16i8 (bitconvert (v4i32 FPR128:$src))), (v16i8 FPR128:$src)>;
 def : Pat<(v16i8 (bitconvert (v8i16 FPR128:$src))), (v16i8 FPR128:$src)>;
 def : Pat<(v16i8 (bitconvert (v2f64 FPR128:$src))), (v16i8 FPR128:$src)>;
 def : Pat<(v16i8 (bitconvert (v4f32 FPR128:$src))), (v16i8 FPR128:$src)>;
+def : Pat<(v16i8 (bitconvert (v8f16 FPR128:$src))), (v16i8 FPR128:$src)>;
 }
 let Predicates = [IsBE] in {
 def : Pat<(v16i8 (bitconvert (f128  FPR128:$src))),
                              (v16i8 (EXTv16i8 (REV64v16i8 FPR128:$src),
                                               (REV64v16i8 FPR128:$src),
                                               (i32 8)))>;
 def : Pat<(v16i8 (bitconvert (v2i64 FPR128:$src))),
                              (v16i8 (REV64v16i8 FPR128:$src))>;
 def : Pat<(v16i8 (bitconvert (v4i32 FPR128:$src))),
                              (v16i8 (REV32v16i8 FPR128:$src))>;
 def : Pat<(v16i8 (bitconvert (v8i16 FPR128:$src))),
                              (v16i8 (REV16v16i8 FPR128:$src))>;
 def : Pat<(v16i8 (bitconvert (v2f64 FPR128:$src))),
                              (v16i8 (REV64v16i8 FPR128:$src))>;
 def : Pat<(v16i8 (bitconvert (v4f32 FPR128:$src))),
                              (v16i8 (REV32v16i8 FPR128:$src))>;
+def : Pat<(v16i8 (bitconvert (v8f16 FPR128:$src))),
+                             (v16i8 (REV16v16i8 FPR128:$src))>;
 }
 
 def : Pat<(v8i8 (extract_subvector (v16i8 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;
 def : Pat<(v4i16 (extract_subvector (v8i16 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;
 def : Pat<(v2i32 (extract_subvector (v4i32 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;
 def : Pat<(v1i64 (extract_subvector (v2i64 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;
 
 // A 64-bit subvector insert to the first 128-bit vector position
 // is a subregister copy that needs no instruction.
 def : Pat<(insert_subvector undef, (v1i64 FPR64:$src), (i32 0)),
           (INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
 def : Pat<(insert_subvector undef, (v1f64 FPR64:$src), (i32 0)),
           (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
 def : Pat<(insert_subvector undef, (v2i32 FPR64:$src), (i32 0)),
           (INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
 def : Pat<(insert_subvector undef, (v2f32 FPR64:$src), (i32 0)),
           (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
 def : Pat<(insert_subvector undef, (v4i16 FPR64:$src), (i32 0)),
           (INSERT_SUBREG (v8i16 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
+def : Pat<(insert_subvector undef, (v4f16 FPR64:$src), (i32 0)),
+          (INSERT_SUBREG (v8f16 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
 def : Pat<(insert_subvector undef, (v8i8 FPR64:$src), (i32 0)),
           (INSERT_SUBREG (v16i8 (IMPLICIT_DEF)), FPR64:$src, dsub)>;
 
 // Use pair-wise add instructions when summing up the lanes for v2f64, v2i64
 // or v2f32.
 def : Pat<(i64 (add (vector_extract (v2i64 FPR128:$Rn), (i64 0)),
                     (vector_extract (v2i64 FPR128:$Rn), (i64 1)))),
            (i64 (ADDPv2i64p (v2i64 FPR128:$Rn)))>;

lib/Target/AArch64/AArch64RegisterInfo.td

 def FPR8  : RegisterClass<"AArch64", [untyped], 8, (sequence "B%u", 0, 31)> {
   let Size = 8;
 }
 def FPR16 : RegisterClass<"AArch64", [f16], 16, (sequence "H%u", 0, 31)> {
   let Size = 16;
 }
 def FPR32 : RegisterClass<"AArch64", [f32, i32], 32,(sequence "S%u", 0, 31)>;
 def FPR64 : RegisterClass<"AArch64", [f64, i64, v2f32, v1f64, v8i8, v4i16, v2i32,
-                                    v1i64],
+                                    v1i64, v4f16],
                                     64, (sequence "D%u", 0, 31)>;
 // We don't (yet) have an f128 legal type, so don't use that here. We
 // normalize 128-bit vectors to v2f64 for arg passing and such, so use
 // that here.
 def FPR128 : RegisterClass<"AArch64",
-                           [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64, f128],
+                           [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64, f128,
+                            v8f16],
                            128, (sequence "Q%u", 0, 31)>;
 
 // The lower 16 vector registers.  Some instructions can only take registers
 // in this range.
 def FPR128_lo : RegisterClass<"AArch64",
                               [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
                               128, (trunc FPR128, 16)>;
 

test/CodeGen/AArch64/arm64-aapcs.ll

 ; CHECK-LABEL: test_vreg_stack:
 ; CHECK: ldr {{q[0-9]+}}, [sp]
   ret <2 x double> %varg_stack;
 }
 
 ; Check that f16 can be passed and returned (ACLE 2.0 extension)
 define half @test_half(float, half %arg) {
 ; CHECK-LABEL: test_half:
-; CHECK: mov v0.16b, v{{[0-9]+}}.16b
+; CHECK: mov v0.16b, v1.16b
   ret half %arg;
 }
 
 ; Check that f16 constants are materialized correctly
 define half @test_half_const() {
 ; CHECK-LABEL: test_half_const:
 ; CHECK: ldr h0, [x{{[0-9]+}}, :lo12:{{.*}}]
   ret half 0xH4248
 }
+
+; Check that v4f16 can be passed and returned in registers
+define <4 x half> @test_v4_half_register(float, <4 x half> %arg) {
+; CHECK-LABEL: test_v4_half_register:
+; CHECK: mov v0.16b, v1.16b
+  ret <4 x half> %arg;
+}
+
+; Check that v8f16 can be passed and returned in registers
+define <8 x half> @test_v8_half_register(float, <8 x half> %arg) {
+; CHECK-LABEL: test_v8_half_register:
+; CHECK: mov v0.16b, v1.16b
+  ret <8 x half> %arg;
+}
+
+; Check that v4f16 can be passed and returned on the stack
+define <4 x half> @test_v4_half_stack([8 x <2 x double>], <4 x half> %arg) {
+; CHECK-LABEL: test_v4_half_stack:
+; CHECK: ldr d0, [sp]
+  ret <4 x half> %arg;
+}
+
+; Check that v8f16 can be passed and returned on the stack
+define <8 x half> @test_v8_half_stack([8 x <2 x double>], <8 x half> %arg) {
+; CHECK-LABEL: test_v8_half_stack:
+; CHECK: ldr q0, [sp]
+  ret <8 x half> %arg;
+}

test/CodeGen/AArch64/fp16-instructions.ll

 ; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
 
 define half @add_h(half %a, half %b) {
 entry:
 ; CHECK-LABEL: add_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK: fadd
-; CHECK: fcvt
+; CHECK-DAG: fcvt [[OP1:s[0-9]+]], h0
+; CHECK-DAG: fcvt [[OP2:s[0-9]+]], h1
+; CHECK: fadd [[RES:s[0-9]+]], [[OP1]], [[OP2]]
+; CHECK: fcvt h0, [[RES]]
   %0 = fadd half %a, %b
   ret half %0
 }
 
 
 define half @sub_h(half %a, half %b) {
 entry:
 ; CHECK-LABEL: sub_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK: fsub
-; CHECK: fcvt
+; CHECK-DAG: fcvt [[OP1:s[0-9]+]], h0
+; CHECK-DAG: fcvt [[OP2:s[0-9]+]], h1
+; CHECK: fsub [[RES:s[0-9]+]], [[OP1]], [[OP2]]
+; CHECK: fcvt h0, [[RES]]
   %0 = fsub half %a, %b
   ret half %0
 }
 
 
 define half @mul_h(half %a, half %b) {
 entry:
 ; CHECK-LABEL: mul_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK: fmul
-; CHECK: fcvt
+; CHECK-DAG: fcvt [[OP1:s[0-9]+]], h0
+; CHECK-DAG: fcvt [[OP2:s[0-9]+]], h1
+; CHECK: fmul [[RES:s[0-9]+]], [[OP1]], [[OP2]]
+; CHECK: fcvt h0, [[RES]]
   %0 = fmul half %a, %b
   ret half %0
 }
 
 
 define half @div_h(half %a, half %b) {
 entry:
 ; CHECK-LABEL: div_h:
-; CHECK: fcvt
-; CHECK: fcvt
-; CHECK: fdiv
-; CHECK: fcvt
+; CHECK-DAG: fcvt [[OP1:s[0-9]+]], h0
+; CHECK-DAG: fcvt [[OP2:s[0-9]+]], h1
+; CHECK: fdiv [[RES:s[0-9]+]], [[OP1]], [[OP2]]
+; CHECK: fcvt h0, [[RES]]
   %0 = fdiv half %a, %b
   ret half %0
 }
 
 
 define half @load_h(half* %a) {
 entry:
 ; CHECK-LABEL: load_h:
-; CHECK: ldr h
+; CHECK: ldr h0, [x0]
   %0 = load half* %a, align 4
   ret half %0
 }
 
 
 define void @store_h(half* %a, half %b) {
 entry:
 ; CHECK-LABEL: store_h:
-; CHECK: str h
+; CHECK: str h0, [x0]
   store half %b, half* %a, align 4
   ret void
 }
 
 define half @s_to_h(float %a) {
 ; CHECK-LABEL: s_to_h:
-; CHECK: fcvt
+; CHECK: fcvt h0, s0
   %1 = fptrunc float %a to half
   ret half %1
 }
 
 define half @d_to_h(double %a) {
 ; CHECK-LABEL: d_to_h:
-; CHECK: fcvt
+; CHECK: fcvt h0, d0
   %1 = fptrunc double %a to half
   ret half %1
 }
 
 define float @h_to_s(half %a) {
 ; CHECK-LABEL: h_to_s:
-; CHECK: fcvt
+; CHECK: fcvt s0, h0
   %1 = fpext half %a to float
   ret float %1
 }
 
 define double @h_to_d(half %a) {
 ; CHECK-LABEL: h_to_d:
-; CHECK: fcvt
+; CHECK: fcvt d0, h0
   %1 = fpext half %a to double
   ret double %1
 }
 
 define half @bitcast_i_to_h(i16 %a) {
 ; CHECK-LABEL: bitcast_i_to_h:
-; CHECK: fmov
+; CHECK: fmov s0, w0
   %1 = bitcast i16 %a to half
   ret half %1
 }
 
 
 define i16 @bitcast_h_to_i(half %a) {
 ; CHECK-LABEL: bitcast_h_to_i:
-; CHECK: fmov
+; CHECK: fmov w0, s0
   %1 = bitcast half %a to i16
   ret i16 %1
 }

test/CodeGen/AArch64/fp16-v4-instructions.ll

+; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
+
+define <4 x half> @add_h(<4 x half> %a, <4 x half> %b) {
+entry:
+; CHECK-LABEL: add_h:
+; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK: fadd [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK: fcvtn v0.4h, [[RES]]
+  %0 = fadd <4 x half> %a, %b
+  ret <4 x half> %0
+}
+
+
+define <4 x half> @sub_h(<4 x half> %a, <4 x half> %b) {
+entry:
+; CHECK-LABEL: sub_h:
+; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK: fsub [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK: fcvtn v0.4h, [[RES]]
+  %0 = fsub <4 x half> %a, %b
+  ret <4 x half> %0
+}
+
+
+define <4 x half> @mul_h(<4 x half> %a, <4 x half> %b) {
+entry:
+; CHECK-LABEL: mul_h:
+; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK: fmul [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK: fcvtn v0.4h, [[RES]]
+  %0 = fmul <4 x half> %a, %b
+  ret <4 x half> %0
+}
+
+
+define <4 x half> @div_h(<4 x half> %a, <4 x half> %b) {
+entry:
+; CHECK-LABEL: div_h:
+; CHECK-DAG: fcvtl [[OP1:v[0-9]+\.4s]], v0.4h
+; CHECK-DAG: fcvtl [[OP2:v[0-9]+\.4s]], v1.4h
+; CHECK: fdiv [[RES:v[0-9]+.4s]], [[OP1]], [[OP2]]
+; CHECK: fcvtn v0.4h, [[RES]]
+  %0 = fdiv <4 x half> %a, %b
+  ret <4 x half> %0
+}
+
+
+define <4 x half> @load_h(<4 x half>* %a) {
+entry:
+; CHECK-LABEL: load_h:
+; CHECK: ldr d0, [x0]
+  %0 = load <4 x half>* %a, align 4
+  ret <4 x half> %0
+}
+
+
+define void @store_h(<4 x half>* %a, <4 x half> %b) {
+entry:
+; CHECK-LABEL: store_h:
+; CHECK: str d0, [x0]
+  store <4 x half> %b, <4 x half>* %a, align 4
+  ret void
+}
+
+define <4 x half> @s_to_h(<4 x float> %a) {
+; CHECK-LABEL: s_to_h:
+; CHECK: fcvtn v0.4h, v0.4s
+  %1 = fptrunc <4 x float> %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x half> @d_to_h(<4 x double> %a) {
+; CHECK-LABEL: d_to_h:
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+  %1 = fptrunc <4 x double> %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x float> @h_to_s(<4 x half> %a) {
+; CHECK-LABEL: h_to_s:
+; CHECK: fcvtl v0.4s, v0.4h
+  %1 = fpext <4 x half> %a to <4 x float>
+  ret <4 x float> %1
+}
+
+define <4 x double> @h_to_d(<4 x half> %a) {
+; CHECK-LABEL: h_to_d:
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+  %1 = fpext <4 x half> %a to <4 x double>
+  ret <4 x double> %1
+}
+
+define <4 x half> @bitcast_i_to_h(float, <4 x i16> %a) {
+; CHECK-LABEL: bitcast_i_to_h:
+; CHECK: mov v0.16b, v1.16b
+  %2 = bitcast <4 x i16> %a to <4 x half>
+  ret <4 x half> %2
+}
+
+define <4 x i16> @bitcast_h_to_i(float, <4 x half> %a) {
+; CHECK-LABEL: bitcast_h_to_i:
+; CHECK: mov v0.16b, v1.16b
+  %2 = bitcast <4 x half> %a to <4 x i16>
+  ret <4 x i16> %2
+}

test/CodeGen/AArch64/fp16-v8-instructions.ll

+; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
+
+define <8 x half> @add_h(<8 x half> %a, <8 x half> %b) {
+entry:
+; CHECK-LABEL: add_h:
+; CHECK: fcvt
+; CHECK: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fadd
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK: fcvt
+  %0 = fadd <8 x half> %a, %b
+  ret <8 x half> %0
+}
+
+
+define <8 x half> @sub_h(<8 x half> %a, <8 x half> %b) {
+entry:
+; CHECK-LABEL: sub_h:
+; CHECK: fcvt
+; CHECK: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fsub
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK: fcvt
+  %0 = fsub <8 x half> %a, %b
+  ret <8 x half> %0
+}
+
+
+define <8 x half> @mul_h(<8 x half> %a, <8 x half> %b) {
+entry:
+; CHECK-LABEL: mul_h:
+; CHECK: fcvt
+; CHECK: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fmul
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK: fcvt
+  %0 = fmul <8 x half> %a, %b
+  ret <8 x half> %0
+}
+
+
+define <8 x half> @div_h(<8 x half> %a, <8 x half> %b) {
+entry:
+; CHECK-LABEL: div_h:
+; CHECK: fcvt
+; CHECK: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fdiv
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK-DAG: fcvt
+; CHECK: fcvt
+  %0 = fdiv <8 x half> %a, %b
+  ret <8 x half> %0
+}
+
+
+define <8 x half> @load_h(<8 x half>* %a) {
+entry:
+; CHECK-LABEL: load_h:
+; CHECK: ldr q0, [x0]
+  %0 = load <8 x half>* %a, align 4
+  ret <8 x half> %0
+}
+
+
+define void @store_h(<8 x half>* %a, <8 x half> %b) {
+entry:
+; CHECK-LABEL: store_h:
+; CHECK: str q0, [x0]
+  store <8 x half> %b, <8 x half>* %a, align 4
+  ret void
+}
+
+define <8 x half> @s_to_h(<8 x float> %a) {
+; CHECK-LABEL: s_to_h:
+; CHECK-DAG: fcvtn v0.4h, v0.4s
+; CHECK-DAG: fcvtn [[REG:v[0-9+]]].4h, v1.4s
+; CHECK: ins v0.d[1], [[REG]].d[0]
+  %1 = fptrunc <8 x float> %a to <8 x half>
+  ret <8 x half> %1
+}
+
+define <8 x half> @d_to_h(<8 x double> %a) {
+; CHECK-LABEL: d_to_h:
+; CHECK-DAG: ins v{{[0-9]+}}.d
+; CHECK-DAG: ins v{{[0-9]+}}.d
+; CHECK-DAG: ins v{{[0-9]+}}.d
+; CHECK-DAG: ins v{{[0-9]+}}.d
+; CHECK-DAG: fcvt h
+; CHECK-DAG: fcvt h
+; CHECK-DAG: fcvt h
+; CHECK-DAG: fcvt h
+; CHECK-DAG: fcvt h
+; CHECK-DAG: fcvt h
+; CHECK-DAG: fcvt h
+; CHECK-DAG: fcvt h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+; CHECK-DAG: ins v{{[0-9]+}}.h
+  %1 = fptrunc <8 x double> %a to <8 x half>
+  ret <8 x half> %1
+}
+
+define <8 x float> @h_to_s(<8 x half> %a) {
+; CHECK-LABEL: h_to_s:
+; CHECK: fcvtl2 v1.4s, v0.8h
+; CHECK: fcvtl v0.4s, v0.4h
+  %1 = fpext <8 x half> %a to <8 x float>
+  ret <8 x float> %1
+}
+
+define <8 x double> @h_to_d(<8 x half> %a) {
+; CHECK-LABEL: h_to_d:
+; CHECK-DAG: fcvt d
+; CHECK-DAG: fcvt d
+; CHECK-DAG: fcvt d
+; CHECK-DAG: fcvt d
+; CHECK-DAG: fcvt d
+; CHECK-DAG: fcvt d
+; CHECK-DAG: fcvt d
+; CHECK-DAG: fcvt d
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+; CHECK-DAG: ins
+  %1 = fpext <8 x half> %a to <8 x double>
+  ret <8 x double> %1
+}
+
+
+define <8 x half> @bitcast_i_to_h(float, <8 x i16> %a) {
+; CHECK-LABEL: bitcast_i_to_h:
+; CHECK: mov v0.16b, v1.16b
+  %2 = bitcast <8 x i16> %a to <8 x half>
+  ret <8 x half> %2
+}
+
+define <8 x i16> @bitcast_h_to_i(float, <8 x half> %a) {
+; CHECK-LABEL: bitcast_h_to_i:
+; CHECK: mov v0.16b, v1.16b
+  %2 = bitcast <8 x half> %a to <8 x i16>
+  ret <8 x i16> %2
+}
+

test/CodeGen/AArch64/fp16-vector-bitcast.ll

+; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
+
+define <4 x i16> @v4f16_to_v4i16(float, <4 x half> %a) #0 {
+; CHECK-LABEL: v4f16_to_v4i16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x half> %a to <4 x i16>
+  ret <4 x i16> %1
+}
+
+define <2 x i32> @v4f16_to_v2i32(float, <4 x half> %a) #0 {
+; CHECK-LABEL: v4f16_to_v2i32:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x half> %a to <2 x i32>
+  ret <2 x i32> %1
+}
+
+define <1 x i64> @v4f16_to_v1i64(float, <4 x half> %a) #0 {
+; CHECK-LABEL: v4f16_to_v1i64:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x half> %a to <1 x i64>
+  ret <1 x i64> %1
+}
+
+define i64 @v4f16_to_i64(float, <4 x half> %a) #0 {
+; CHECK-LABEL: v4f16_to_i64:
+; CHECK: fmov x0, d1
+entry:
+  %1 = bitcast <4 x half> %a to i64
+  ret i64 %1
+}
+
+define <2 x float> @v4f16_to_v2float(float, <4 x half> %a) #0 {
+; CHECK-LABEL: v4f16_to_v2float:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x half> %a to <2 x float>
+  ret <2 x float> %1
+}
+
+define <1 x double> @v4f16_to_v1double(float, <4 x half> %a) #0 {
+; CHECK-LABEL: v4f16_to_v1double:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x half> %a to <1 x double>
+  ret <1 x double> %1
+}
+
+define double @v4f16_to_double(float, <4 x half> %a) #0 {
+; CHECK-LABEL: v4f16_to_double:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x half> %a to double
+  ret double %1
+}
+
+
+define <4 x half> @v4i16_to_v4f16(float, <4 x i16> %a) #0 {
+; CHECK-LABEL: v4i16_to_v4f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x i16> %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x half> @v2i32_to_v4f16(float, <2 x i32> %a) #0 {
+; CHECK-LABEL: v2i32_to_v4f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <2 x i32> %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x half> @v1i64_to_v4f16(float, <1 x i64> %a) #0 {
+; CHECK-LABEL: v1i64_to_v4f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <1 x i64> %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x half> @i64_to_v4f16(float, i64 %a) #0 {
+; CHECK-LABEL: i64_to_v4f16:
+; CHECK: fmov d0, x0
+entry:
+  %1 = bitcast i64 %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x half> @v2float_to_v4f16(float, <2 x float> %a) #0 {
+; CHECK-LABEL: v2float_to_v4f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <2 x float> %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x half> @v1double_to_v4f16(float, <1 x double> %a) #0 {
+; CHECK-LABEL: v1double_to_v4f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <1 x double> %a to <4 x half>
+  ret <4 x half> %1
+}
+
+define <4 x half> @double_to_v4f16(float, double %a) #0 {
+; CHECK-LABEL: double_to_v4f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast double %a to <4 x half>
+  ret <4 x half> %1
+}
+
+
+
+
+
+
+
+
+
+
+define <8 x i16> @v8f16_to_v8i16(float, <8 x half> %a) #0 {
+; CHECK-LABEL: v8f16_to_v8i16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <8 x half> %a to <8 x i16>
+  ret <8 x i16> %1
+}
+
+define <4 x i32> @v8f16_to_v4i32(float, <8 x half> %a) #0 {
+; CHECK-LABEL: v8f16_to_v4i32:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <8 x half> %a to <4 x i32>
+  ret <4 x i32> %1
+}
+
+define <2 x i64> @v8f16_to_v2i64(float, <8 x half> %a) #0 {
+; CHECK-LABEL: v8f16_to_v2i64:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <8 x half> %a to <2 x i64>
+  ret <2 x i64> %1
+}
+
+define <4 x float> @v8f16_to_v4float(float, <8 x half> %a) #0 {
+; CHECK-LABEL: v8f16_to_v4float:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <8 x half> %a to <4 x float>
+  ret <4 x float> %1
+}
+
+define <2 x double> @v8f16_to_v2double(float, <8 x half> %a) #0 {
+; CHECK-LABEL: v8f16_to_v2double:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <8 x half> %a to <2 x double>
+  ret <2 x double> %1
+}
+
+define <8 x half> @v8i16_to_v8f16(float, <8 x i16> %a) #0 {
+; CHECK-LABEL: v8i16_to_v8f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <8 x i16> %a to <8 x half>
+  ret <8 x half> %1
+}
+
+define <8 x half> @v4i32_to_v8f16(float, <4 x i32> %a) #0 {
+; CHECK-LABEL: v4i32_to_v8f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x i32> %a to <8 x half>
+  ret <8 x half> %1
+}
+
+define <8 x half> @v2i64_to_v8f16(float, <2 x i64> %a) #0 {
+; CHECK-LABEL: v2i64_to_v8f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <2 x i64> %a to <8 x half>
+  ret <8 x half> %1
+}
+
+define <8 x half> @v4float_to_v8f16(float, <4 x float> %a) #0 {
+; CHECK-LABEL: v4float_to_v8f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <4 x float> %a to <8 x half>
+  ret <8 x half> %1
+}
+
+define <8 x half> @v2double_to_v8f16(float, <2 x double> %a) #0 {
+; CHECK-LABEL: v2double_to_v8f16:
+; CHECK: mov v0.16b, v1.16b
+entry:
+  %1 = bitcast <2 x double> %a to <8 x half>
+  ret <8 x half> %1
+}

test/CodeGen/AArch64/fp16-vector-load-store.ll

+; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
+
+; Simple load of v4i16
+define <4 x half> @load_64(<4 x half>* nocapture readonly %a) #0 {
+; CHECK-LABEL: load_64:
+; CHECK: ldr d0, [x0]
+entry:
+  %0 = load <4 x half>* %a, align 8
+  ret <4 x half> %0
+}
+
+; Simple load of v8i16
+define <8 x half> @load_128(<8 x half>* nocapture readonly %a) #0 {
+; CHECK-LABEL: load_128:
+; CHECK: ldr q0, [x0]
+entry:
+  %0 = load <8 x half>* %a, align 16
+  ret <8 x half> %0
+}
+
+; Duplicating load to v4i16
+define <4 x half> @load_dup_64(half* nocapture readonly %a) #0 {
+; CHECK-LABEL: load_dup_64:
+; CHECK: ld1r { v0.4h }, [x0]
+entry:
+  %0 = load half* %a, align 2
+  %1 = insertelement <4 x half> undef, half %0, i32 0
+  %2 = shufflevector <4 x half> %1, <4 x half> undef, <4 x i32> zeroinitializer
+  ret <4 x half> %2
+}
+
+; Duplicating load to v8i16
+define <8 x half> @load_dup_128(half* nocapture readonly %a) #0 {
+; CHECK-LABEL: load_dup_128:
+; CHECK: ld1r { v0.8h }, [x0]
+entry:
+  %0 = load half* %a, align 2
+  %1 = insertelement <8 x half> undef, half %0, i32 0
+  %2 = shufflevector <8 x half> %1, <8 x half> undef, <8 x i32> zeroinitializer
+  ret <8 x half> %2
+}
+
+; Load to one lane of v4f16
+define <4 x half> @load_lane_64(half* nocapture readonly %a, <4 x half> %b) #0 {
+; CHECK-LABEL: load_lane_64:
+; CHECK: ld1 { v0.h }[2], [x0]
+entry:
+  %0 = load half* %a, align 2
+  %1 = insertelement <4 x half> %b, half %0, i32 2
+  ret <4 x half> %1
+}
+
+; Load to one lane of v8f16
+define <8 x half> @load_lane_128(half* nocapture readonly %a, <8 x half> %b) #0 {
+; CHECK-LABEL: load_lane_128:
+; CHECK: ld1 { v0.h }[5], [x0]
+entry:
+  %0 = load half* %a, align 2
+  %1 = insertelement <8 x half> %b, half %0, i32 5
+  ret <8 x half> %1
+}
+
+; Simple store of v4f16
+define void @store_64(<4 x half>* nocapture %a, <4 x half> %b) #1 {
+; CHECK-LABEL: store_64:
+; CHECK: str d0, [x0]
+entry:
+  store <4 x half> %b, <4 x half>* %a, align 8
+  ret void
+}
+
+; Simple store of v8f16
+define void @store_128(<8 x half>* nocapture %a, <8 x half> %b) #1 {
+; CHECK-LABEL: store_128:
+; CHECK: str q0, [x0]
+entry:
+  store <8 x half> %b, <8 x half>* %a, align 16
+  ret void
+}
+
+; Store from one lane of v4f16
+define void @store_lane_64(half* nocapture %a, <4 x half> %b) #1 {
+; CHECK-LABEL: store_lane_64:
+; CHECK: st1 { v0.h }[2], [x0]
+entry:
+  %0 = extractelement <4 x half> %b, i32 2
+  store half %0, half* %a, align 2
+  ret void
+}
+
+; Store from one lane of v8f16
+define void @store_lane_128(half* nocapture %a, <8 x half> %b) #1 {
+; CHECK-LABEL: store_lane_128:
+; CHECK: st1 { v0.h }[5], [x0]
+entry:
+  %0 = extractelement <8 x half> %b, i32 5
+  store half %0, half* %a, align 2
+  ret void
+}
+
+; NEON intrinsics - (de-)interleaving loads and stores
+declare { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2.v4f16.p0v4f16(<4 x half>*)
+declare { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3.v4f16.p0v4f16(<4 x half>*)
+declare { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4.v4f16.p0v4f16(<4 x half>*)
+declare void @llvm.aarch64.neon.st2.v4f16.p0v4f16(<4 x half>, <4 x half>, <4 x half>*)
+declare void @llvm.aarch64.neon.st3.v4f16.p0v4f16(<4 x half>, <4 x half>, <4 x half>, <4 x half>*)
+declare void @llvm.aarch64.neon.st4.v4f16.p0v4f16(<4 x half>, <4 x half>, <4 x half>, <4 x half>, <4 x half>*)
+declare { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2.v8f16.p0v8f16(<8 x half>*)
+declare { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3.v8f16.p0v8f16(<8 x half>*)
+declare { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4.v8f16.p0v8f16(<8 x half>*)
+declare void @llvm.aarch64.neon.st2.v8f16.p0v8f16(<8 x half>, <8 x half>, <8 x half>*)
+declare void @llvm.aarch64.neon.st3.v8f16.p0v8f16(<8 x half>, <8 x half>, <8 x half>, <8 x half>*)
+declare void @llvm.aarch64.neon.st4.v8f16.p0v8f16(<8 x half>, <8 x half>, <8 x half>, <8 x half>, <8 x half>*)
+
+; Load 2 x v4f16 with de-interleaving
+define { <4 x half>, <4 x half> } @load_interleave_64_2(<4 x half>* %a) #0 {
+; CHECK-LABEL: load_interleave_64_2:
+; CHECK: ld2 { v0.4h, v1.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2.v4f16.p0v4f16(<4 x half>* %a)
+  ret { <4 x half>, <4 x half> } %0
+}
+
+; Load 3 x v4f16 with de-interleaving
+define { <4 x half>, <4 x half>, <4 x half> } @load_interleave_64_3(<4 x half>* %a) #0 {
+; CHECK-LABEL: load_interleave_64_3:
+; CHECK: ld3 { v0.4h, v1.4h, v2.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3.v4f16.p0v4f16(<4 x half>* %a)
+  ret { <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Load 4 x v4f16 with de-interleaving
+define { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @load_interleave_64_4(<4 x half>* %a) #0 {
+; CHECK-LABEL: load_interleave_64_4:
+; CHECK: ld4 { v0.4h, v1.4h, v2.4h, v3.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4.v4f16.p0v4f16(<4 x half>* %a)
+  ret { <4 x half>, <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Store 2 x v4f16 with interleaving
+define void @store_interleave_64_2(<4 x half>* %a, <4 x half> %b, <4 x half> %c) #0 {
+; CHECK-LABEL: store_interleave_64_2:
+; CHECK: st2 { v0.4h, v1.4h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st2.v4f16.p0v4f16(<4 x half> %b, <4 x half> %c, <4 x half>* %a)
+  ret void
+}
+
+; Store 3 x v4f16 with interleaving
+define void @store_interleave_64_3(<4 x half>* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d) #0 {
+; CHECK-LABEL: store_interleave_64_3:
+; CHECK: st3 { v0.4h, v1.4h, v2.4h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st3.v4f16.p0v4f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half>* %a)
+  ret void
+}
+
+; Store 4 x v4f16 with interleaving
+define void @store_interleave_64_4(<4 x half>* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e) #0 {
+; CHECK-LABEL: store_interleave_64_4:
+; CHECK: st4 { v0.4h, v1.4h, v2.4h, v3.4h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st4.v4f16.p0v4f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e, <4 x half>* %a)
+  ret void
+}
+
+; Load 2 x v8f16 with de-interleaving
+define { <8 x half>, <8 x half> } @load_interleave_128_2(<8 x half>* %a) #0 {
+; CHECK-LABEL: load_interleave_128_2:
+; CHECK: ld2 { v0.8h, v1.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2.v8f16.p0v8f16(<8 x half>* %a)
+  ret { <8 x half>, <8 x half> } %0
+}
+
+; Load 3 x v8f16 with de-interleaving
+define { <8 x half>, <8 x half>, <8 x half> } @load_interleave_128_3(<8 x half>* %a) #0 {
+; CHECK-LABEL: load_interleave_128_3:
+; CHECK: ld3 { v0.8h, v1.8h, v2.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3.v8f16.p0v8f16(<8 x half>* %a)
+  ret { <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+; Load 8 x v8f16 with de-interleaving
+define { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @load_interleave_128_4(<8 x half>* %a) #0 {
+; CHECK-LABEL: load_interleave_128_4:
+; CHECK: ld4 { v0.8h, v1.8h, v2.8h, v3.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4.v8f16.p0v8f16(<8 x half>* %a)
+  ret { <8 x half>, <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+; Store 2 x v8f16 with interleaving
+define void @store_interleave_128_2(<8 x half>* %a, <8 x half> %b, <8 x half> %c) #0 {
+; CHECK-LABEL: store_interleave_128_2:
+; CHECK: st2 { v0.8h, v1.8h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st2.v8f16.p0v8f16(<8 x half> %b, <8 x half> %c, <8 x half>* %a)
+  ret void
+}
+
+; Store 3 x v8f16 with interleaving
+define void @store_interleave_128_3(<8 x half>* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d) #0 {
+; CHECK-LABEL: store_interleave_128_3:
+; CHECK: st3 { v0.8h, v1.8h, v2.8h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st3.v8f16.p0v8f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half>* %a)
+  ret void
+}
+
+; Store 8 x v8f16 with interleaving
+define void @store_interleave_128_4(<8 x half>* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e) #0 {
+; CHECK-LABEL: store_interleave_128_4:
+; CHECK: st4 { v0.8h, v1.8h, v2.8h, v3.8h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st4.v8f16.p0v8f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e, <8 x half>* %a)
+  ret void
+}
+
+; NEON intrinsics - duplicating loads
+declare { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2r.v4f16.p0f16(half*)
+declare { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3r.v4f16.p0f16(half*)
+declare { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4r.v4f16.p0f16(half*)
+declare { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2r.v8f16.p0f16(half*)
+declare { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3r.v8f16.p0f16(half*)
+declare { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4r.v8f16.p0f16(half*)
+
+; Load 2 x v4f16 with duplication
+define { <4 x half>, <4 x half> } @load_dup_64_2(half* %a) #0 {
+; CHECK-LABEL: load_dup_64_2:
+; CHECK: ld2r { v0.4h, v1.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2r.v4f16.p0f16(half* %a)
+  ret { <4 x half>, <4 x half> } %0
+}
+
+; Load 3 x v4f16 with duplication
+define { <4 x half>, <4 x half>, <4 x half> } @load_dup_64_3(half* %a) #0 {
+; CHECK-LABEL: load_dup_64_3:
+; CHECK: ld3r { v0.4h, v1.4h, v2.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3r.v4f16.p0f16(half* %a)
+  ret { <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Load 4 x v4f16 with duplication
+define { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @load_dup_64_4(half* %a) #0 {
+; CHECK-LABEL: load_dup_64_4:
+; CHECK: ld4r { v0.4h, v1.4h, v2.4h, v3.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4r.v4f16.p0f16(half* %a)
+  ret { <4 x half>, <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Load 2 x v8f16 with duplication
+define { <8 x half>, <8 x half> } @load_dup_128_2(half* %a) #0 {
+; CHECK-LABEL: load_dup_128_2:
+; CHECK: ld2r { v0.8h, v1.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2r.v8f16.p0f16(half* %a)
+  ret { <8 x half>, <8 x half> } %0
+}
+
+; Load 3 x v8f16 with duplication
+define { <8 x half>, <8 x half>, <8 x half> } @load_dup_128_3(half* %a) #0 {
+; CHECK-LABEL: load_dup_128_3:
+; CHECK: ld3r { v0.8h, v1.8h, v2.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3r.v8f16.p0f16(half* %a)
+  ret { <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+; Load 8 x v8f16 with duplication
+define { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @load_dup_128_4(half* %a) #0 {
+; CHECK-LABEL: load_dup_128_4:
+; CHECK: ld4r { v0.8h, v1.8h, v2.8h, v3.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4r.v8f16.p0f16(half* %a)
+  ret { <8 x half>, <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+
+; NEON intrinsics - loads and stores to/from one lane
+declare { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2lane.v4f16.p0f16(<4 x half>, <4 x half>, i64, half*)
+declare { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3lane.v4f16.p0f16(<4 x half>, <4 x half>, <4 x half>, i64, half*)
+declare { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4lane.v4f16.p0f16(<4 x half>, <4 x half>, <4 x half>, <4 x half>, i64, half*)
+declare void @llvm.aarch64.neon.st2lane.v4f16.p0f16(<4 x half>, <4 x half>, i64, half*)
+declare void @llvm.aarch64.neon.st3lane.v4f16.p0f16(<4 x half>, <4 x half>, <4 x half>, i64, half*)
+declare void @llvm.aarch64.neon.st4lane.v4f16.p0f16(<4 x half>, <4 x half>, <4 x half>, <4 x half>, i64, half*)
+declare { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2lane.v8f16.p0f16(<8 x half>, <8 x half>, i64, half*)
+declare { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3lane.v8f16.p0f16(<8 x half>, <8 x half>, <8 x half>, i64, half*)
+declare { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4lane.v8f16.p0f16(<8 x half>, <8 x half>, <8 x half>, <8 x half>, i64, half*)
+declare void @llvm.aarch64.neon.st2lane.v8f16.p0f16(<8 x half>, <8 x half>, i64, half*)
+declare void @llvm.aarch64.neon.st3lane.v8f16.p0f16(<8 x half>, <8 x half>, <8 x half>, i64, half*)
+declare void @llvm.aarch64.neon.st4lane.v8f16.p0f16(<8 x half>, <8 x half>, <8 x half>, <8 x half>, i64, half*)
+
+; Load one lane of 2 x v4f16
+define { <4 x half>, <4 x half> } @load_lane_64_2(half* %a, <4 x half> %b, <4 x half> %c) #0 {
+; CHECK-LABEL: load_lane_64_2:
+; CHECK: ld2 { v0.h, v1.h }[2], [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld2lane.v4f16.p0f16(<4 x half> %b, <4 x half> %c, i64 2, half* %a)
+  ret { <4 x half>, <4 x half> } %0
+}
+
+; Load one lane of 3 x v4f16
+define { <4 x half>, <4 x half>, <4 x half> } @load_lane_64_3(half* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d) #0 {
+; CHECK-LABEL: load_lane_64_3:
+; CHECK: ld3 { v0.h, v1.h, v2.h }[2], [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld3lane.v4f16.p0f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, i64 2, half* %a)
+  ret { <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Load one lane of 4 x v4f16
+define { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @load_lane_64_4(half* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e) #0 {
+; CHECK-LABEL: load_lane_64_4:
+; CHECK: ld4 { v0.h, v1.h, v2.h, v3.h }[2], [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld4lane.v4f16.p0f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e, i64 2, half* %a)
+  ret { <4 x half>, <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Store one lane of 2 x v4f16
+define void @store_lane_64_2(half* %a, <4 x half> %b, <4 x half> %c) #0 {
+; CHECK-LABEL: store_lane_64_2:
+; CHECK: st2 { v0.h, v1.h }[2], [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st2lane.v4f16.p0f16(<4 x half> %b, <4 x half> %c, i64 2, half* %a)
+  ret void
+}
+
+; Store one lane of 3 x v4f16
+define void @store_lane_64_3(half* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d) #0 {
+; CHECK-LABEL: store_lane_64_3:
+; CHECK: st3 { v0.h, v1.h, v2.h }[2], [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st3lane.v4f16.p0f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, i64 2, half* %a)
+  ret void
+}
+
+; Store one lane of 4 x v4f16
+define void @store_lane_64_4(half* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e) #0 {
+; CHECK-LABEL: store_lane_64_4:
+; CHECK: st4 { v0.h, v1.h, v2.h, v3.h }[2], [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st4lane.v4f16.p0f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e, i64 2, half* %a)
+  ret void
+}
+
+; Load one lane of 2 x v8f16
+define { <8 x half>, <8 x half> } @load_lane_128_2(half* %a, <8 x half> %b, <8 x half> %c) #0 {
+; CHECK-LABEL: load_lane_128_2:
+; CHECK: ld2 { v0.h, v1.h }[2], [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld2lane.v8f16.p0f16(<8 x half> %b, <8 x half> %c, i64 2, half* %a)
+  ret { <8 x half>, <8 x half> } %0
+}
+
+; Load one lane of 3 x v8f16
+define { <8 x half>, <8 x half>, <8 x half> } @load_lane_128_3(half* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d) #0 {
+; CHECK-LABEL: load_lane_128_3:
+; CHECK: ld3 { v0.h, v1.h, v2.h }[2], [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld3lane.v8f16.p0f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, i64 2, half* %a)
+  ret { <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+; Load one lane of 8 x v8f16
+define { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @load_lane_128_4(half* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e) #0 {
+; CHECK-LABEL: load_lane_128_4:
+; CHECK: ld4 { v0.h, v1.h, v2.h, v3.h }[2], [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld4lane.v8f16.p0f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e, i64 2, half* %a)
+  ret { <8 x half>, <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+; Store one lane of 2 x v8f16
+define void @store_lane_128_2(half* %a, <8 x half> %b, <8 x half> %c) #0 {
+; CHECK-LABEL: store_lane_128_2:
+; CHECK: st2 { v0.h, v1.h }[2], [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st2lane.v8f16.p0f16(<8 x half> %b, <8 x half> %c, i64 2, half* %a)
+  ret void
+}
+
+; Store one lane of 3 x v8f16
+define void @store_lane_128_3(half* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d) #0 {
+; CHECK-LABEL: store_lane_128_3:
+; CHECK: st3 { v0.h, v1.h, v2.h }[2], [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st3lane.v8f16.p0f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, i64 2, half* %a)
+  ret void
+}
+
+; Store one lane of 8 x v8f16
+define void @store_lane_128_4(half* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e) #0 {
+; CHECK-LABEL: store_lane_128_4:
+; CHECK: st4 { v0.h, v1.h, v2.h, v3.h }[2], [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st4lane.v8f16.p0f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e, i64 2, half* %a)
+  ret void
+}
+
+; NEON intrinsics - load/store without interleaving
+declare { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld1x2.v4f16.p0v4f16(<4 x half>*)
+declare { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld1x3.v4f16.p0v4f16(<4 x half>*)
+declare { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld1x4.v4f16.p0v4f16(<4 x half>*)
+declare void @llvm.aarch64.neon.st1x2.v4f16.p0v4f16(<4 x half>, <4 x half>, <4 x half>*)
+declare void @llvm.aarch64.neon.st1x3.v4f16.p0v4f16(<4 x half>, <4 x half>, <4 x half>, <4 x half>*)
+declare void @llvm.aarch64.neon.st1x4.v4f16.p0v4f16(<4 x half>, <4 x half>, <4 x half>, <4 x half>, <4 x half>*)
+declare { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld1x2.v8f16.p0v8f16(<8 x half>*)
+declare { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld1x3.v8f16.p0v8f16(<8 x half>*)
+declare { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld1x4.v8f16.p0v8f16(<8 x half>*)
+declare void @llvm.aarch64.neon.st1x2.v8f16.p0v8f16(<8 x half>, <8 x half>, <8 x half>*)
+declare void @llvm.aarch64.neon.st1x3.v8f16.p0v8f16(<8 x half>, <8 x half>, <8 x half>, <8 x half>*)
+declare void @llvm.aarch64.neon.st1x4.v8f16.p0v8f16(<8 x half>, <8 x half>, <8 x half>, <8 x half>, <8 x half>*)
+
+; Load 2 x v4f16 without de-interleaving
+define { <4 x half>, <4 x half> } @load_64_2(<4 x half>* %a) #0 {
+; CHECK-LABEL: load_64_2:
+; CHECK: ld1 { v0.4h, v1.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half> } @llvm.aarch64.neon.ld1x2.v4f16.p0v4f16(<4 x half>* %a)
+  ret { <4 x half>, <4 x half> } %0
+}
+
+; Load 3 x v4f16 without de-interleaving
+define { <4 x half>, <4 x half>, <4 x half> } @load_64_3(<4 x half>* %a) #0 {
+; CHECK-LABEL: load_64_3:
+; CHECK: ld1 { v0.4h, v1.4h, v2.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld1x3.v4f16.p0v4f16(<4 x half>* %a)
+  ret { <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Load 4 x v4f16 without de-interleaving
+define { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @load_64_4(<4 x half>* %a) #0 {
+; CHECK-LABEL: load_64_4:
+; CHECK: ld1 { v0.4h, v1.4h, v2.4h, v3.4h }, [x0]
+entry:
+  %0 = tail call { <4 x half>, <4 x half>, <4 x half>, <4 x half> } @llvm.aarch64.neon.ld1x4.v4f16.p0v4f16(<4 x half>* %a)
+  ret { <4 x half>, <4 x half>, <4 x half>, <4 x half> } %0
+}
+
+; Store 2 x v4f16 without interleaving
+define void @store_64_2(<4 x half>* %a, <4 x half> %b, <4 x half> %c) #0 {
+; CHECK-LABEL: store_64_2:
+; CHECK: st1 { v0.4h, v1.4h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st1x2.v4f16.p0v4f16(<4 x half> %b, <4 x half> %c, <4 x half>* %a)
+  ret void
+}
+
+; Store 3 x v4f16 without interleaving
+define void @store_64_3(<4 x half>* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d) #0 {
+; CHECK-LABEL: store_64_3:
+; CHECK: st1 { v0.4h, v1.4h, v2.4h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st1x3.v4f16.p0v4f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half>* %a)
+  ret void
+}
+
+; Store 4 x v4f16 without interleaving
+define void @store_64_4(<4 x half>* %a, <4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e) #0 {
+; CHECK-LABEL: store_64_4:
+; CHECK: st1 { v0.4h, v1.4h, v2.4h, v3.4h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st1x4.v4f16.p0v4f16(<4 x half> %b, <4 x half> %c, <4 x half> %d, <4 x half> %e, <4 x half>* %a)
+  ret void
+}
+
+; Load 2 x v8f16 without de-interleaving
+define { <8 x half>, <8 x half> } @load_128_2(<8 x half>* %a) #0 {
+; CHECK-LABEL: load_128_2:
+; CHECK: ld1 { v0.8h, v1.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half> } @llvm.aarch64.neon.ld1x2.v8f16.p0v8f16(<8 x half>* %a)
+  ret { <8 x half>, <8 x half> } %0
+}
+
+; Load 3 x v8f16 without de-interleaving
+define { <8 x half>, <8 x half>, <8 x half> } @load_128_3(<8 x half>* %a) #0 {
+; CHECK-LABEL: load_128_3:
+; CHECK: ld1 { v0.8h, v1.8h, v2.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld1x3.v8f16.p0v8f16(<8 x half>* %a)
+  ret { <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+; Load 8 x v8f16 without de-interleaving
+define { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @load_128_4(<8 x half>* %a) #0 {
+; CHECK-LABEL: load_128_4:
+; CHECK: ld1 { v0.8h, v1.8h, v2.8h, v3.8h }, [x0]
+entry:
+  %0 = tail call { <8 x half>, <8 x half>, <8 x half>, <8 x half> } @llvm.aarch64.neon.ld1x4.v8f16.p0v8f16(<8 x half>* %a)
+  ret { <8 x half>, <8 x half>, <8 x half>, <8 x half> } %0
+}
+
+; Store 2 x v8f16 without interleaving
+define void @store_128_2(<8 x half>* %a, <8 x half> %b, <8 x half> %c) #0 {
+; CHECK-LABEL: store_128_2:
+; CHECK: st1 { v0.8h, v1.8h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st1x2.v8f16.p0v8f16(<8 x half> %b, <8 x half> %c, <8 x half>* %a)
+  ret void
+}
+
+; Store 3 x v8f16 without interleaving
+define void @store_128_3(<8 x half>* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d) #0 {
+; CHECK-LABEL: store_128_3:
+; CHECK: st1 { v0.8h, v1.8h, v2.8h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st1x3.v8f16.p0v8f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half>* %a)
+  ret void
+}
+
+; Store 8 x v8f16 without interleaving
+define void @store_128_4(<8 x half>* %a, <8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e) #0 {
+; CHECK-LABEL: store_128_4:
+; CHECK: st1 { v0.8h, v1.8h, v2.8h, v3.8h }, [x0]
+entry:
+  tail call void @llvm.aarch64.neon.st1x4.v8f16.p0v8f16(<8 x half> %b, <8 x half> %c, <8 x half> %d, <8 x half> %e, <8 x half>* %a)
+  ret void
+}

test/CodeGen/AArch64/fp16-vector-shuffle.ll

+; RUN: llc < %s -mtriple=aarch64-none-eabi | FileCheck %s
+
+; float16x4_t select_64(float16x4_t a, float16x4_t b, uint16x4_t c) { return vbsl_u16(c, a, b); }
+define <4 x half> @select_64(<4 x half> %a, <4 x half> %b, <4 x i16> %c) #0 {
+; CHECK-LABEL: select_64:
+; CHECK: bsl
+entry:
+  %0 = bitcast <4 x half> %a to <4 x i16>
+  %1 = bitcast <4 x half> %b to <4 x i16>
+  %vbsl3.i = and <4 x i16> %0, %c
+  %2 = xor <4 x i16> %c, <i16 -1, i16 -1, i16 -1, i16 -1>
+  %vbsl4.i = and <4 x i16> %1, %2
+  %vbsl5.i = or <4 x i16> %vbsl3.i, %vbsl4.i
+  %3 = bitcast <4 x i16> %vbsl5.i to <4 x half>
+  ret <4 x half> %3
+}
+
+; float16x8_t select_128(float16x8_t a, float16x8_t b, uint16x8_t c) { return vbslq_u16(c, a, b); }
+define <8 x half> @select_128(<8 x half> %a, <8 x half> %b, <8 x i16> %c) #0 {
+; CHECK-LABEL: select_128:
+; CHECK: bsl
+entry:
+  %0 = bitcast <8 x half> %a to <8 x i16>
+  %1 = bitcast <8 x half> %b to <8 x i16>
+  %vbsl3.i = and <8 x i16> %0, %c
+  %2 = xor <8 x i16> %c, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
+  %vbsl4.i = and <8 x i16> %1, %2
+  %vbsl5.i = or <8 x i16> %vbsl3.i, %vbsl4.i
+  %3 = bitcast <8 x i16> %vbsl5.i to <8 x half>
+  ret <8 x half> %3
+}
+
+; float16x4_t lane_64_64(float16x4_t a, float16x4_t b) {
+;  return vcopy_lane_s16(a, 1, b, 2);
+; }
+define <4 x half> @lane_64_64(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-LABEL: lane_64_64:
+; CHECK: ins
+entry:
+  %0 = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 0, i32 6, i32 2, i32 3>
+  ret <4 x half> %0
+}
+
+; float16x8_t lane_128_64(float16x8_t a, float16x4_t b) {
+;   return vcopyq_lane_s16(a, 1, b, 2);
+; }
+define <8 x half> @lane_128_64(<8 x half> %a, <4 x half> %b) #0 {
+; CHECK-LABEL: lane_128_64:
+; CHECK: ins
+entry:
+  %0 = bitcast <4 x half> %b to <4 x i16>
+  %vget_lane = extractelement <4 x i16> %0, i32 2
+  %1 = bitcast <8 x half> %a to <8 x i16>
+  %vset_lane = insertelement <8 x i16> %1, i16 %vget_lane, i32 1
+  %2 = bitcast <8 x i16> %vset_lane to <8 x half>
+  ret <8 x half> %2
+}
+
+; float16x4_t lane_64_128(float16x4_t a, float16x8_t b) {
+;   return vcopy_laneq_s16(a, 3, b, 5);
+; }
+define <4 x half> @lane_64_128(<4 x half> %a, <8 x half> %b) #0 {
+; CHECK-LABEL: lane_64_128:
+; CHECK: ins
+entry:
+  %0 = bitcast <8 x half> %b to <8 x i16>
+  %vgetq_lane = extractelement <8 x i16> %0, i32 5
+  %1 = bitcast <4 x half> %a to <4 x i16>
+  %vset_lane = insertelement <4 x i16> %1, i16 %vgetq_lane, i32 3
+  %2 = bitcast <4 x i16> %vset_lane to <4 x half>
+  ret <4 x half> %2
+}
+
+; float16x8_t lane_128_128(float16x8_t a, float16x8_t b) {
+;   return vcopyq_laneq_s16(a, 3, b, 5);
+; }
+define <8 x half> @lane_128_128(<8 x half> %a, <8 x half> %b) #0 {
+; CHECK-LABEL: lane_128_128:
+; CHECK: ins
+entry:
+  %0 = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 13, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x half> %0
+}
+
+; float16x4_t ext_64(float16x4_t a, float16x4_t b) {
+;   return vext_s16(a, b, 3);
+; }
+define <4 x half> @ext_64(<4 x half> %a, <4 x half> %b) #0 {
+; CHECK-LABEL: ext_64:
+; CHECK: ext
+entry:
+  %0 = shufflevector <4 x half> %a, <4 x half> %b, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
+  ret <4 x half> %0
+}
+
+; float16x8_t ext_128(float16x8_t a, float16x8_t b) {
+;   return vextq_s16(a, b, 3);
+; }
+define <8 x half> @ext_128(<8 x half> %a, <8 x half> %b) #0 {
+; CHECK-LABEL: ext_128:
+; CHECK: ext
+entry:
+  %0 = shufflevector <8 x half> %a, <8 x half> %b, <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
+  ret <8 x half> %0
+}
+
+; float16x4_t rev32_64(float16x4_t a) {
+;   return vrev32_s16(a);
+; }
+define <4 x half> @rev32_64(<4 x half> %a) #0 {
+entry:
+; CHECK-LABEL: rev32_64:
+; CHECK: rev32
+  %0 = shufflevector <4 x half> %a, <4 x half> undef, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
+  ret <4 x half> %0
+}
+
+; float16x4_t rev64_64(float16x4_t a) {
+;   return vrev64_s16(a);
+; }
+define <4 x half> @rev64_64(<4 x half> %a) #0 {
+entry:
+; CHECK-LABEL: rev64_64:
+; CHECK: rev64
+  %0 = shufflevector <4 x half> %a, <4 x half> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+  ret <4 x half> %0
+}
+
+; float16x8_t rev32_128(float16x8_t a) {
+;   return vrev32q_s16(a);
+; }
+define <8 x half> @rev32_128(<8 x half> %a) #0 {
+entry:
+; CHECK-LABEL: rev32_128:
+; CHECK: rev32
+  %0 = shufflevector <8 x half> %a, <8 x half> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
+  ret <8 x half> %0
+}
+
+; float16x8_t rev64_128(float16x8_t a) {
+;   return vrev64q_s16(a);
+; }
+define <8 x half> @rev64_128(<8 x half> %a) #0 {
+entry:
+; CHECK-LABEL: rev64_128:
+; CHECK: rev64
+  %0 = shufflevector <8 x half> %a, <8 x half> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
+  ret <8 x half> %0
+}
+
+; float16x4_t create_64(long long a) { return vcreate_f16(a); }
+define <4 x half> @create_64(i64 %a) #0 {
+; CHECK-LABEL: create_64:
+; CHECK: fmov
+entry:
+  %0 = bitcast i64 %a to <4 x half>
+  ret <4 x half> %0
+}
+
+; float16x4_t dup_64(__fp16 a) { return vdup_n_f16(a); }
+define <4 x half> @dup_64(half %a) #0 {
+; CHECK-LABEL: dup_64:
+; CHECK: dup
+entry:
+  %vecinit = insertelement <4 x half> undef, half %a, i32 0
+  %vecinit1 = insertelement <4 x half> %vecinit, half %a, i32 1
+  %vecinit2 = insertelement <4 x half> %vecinit1, half %a, i32 2
+  %vecinit3 = insertelement <4 x half> %vecinit2, half %a, i32 3
+  ret <4 x half> %vecinit3
+}
+
+; float16x8_t dup_128(__fp16 a) { return vdupq_n_f16(a); }
+define <8 x half> @dup_128(half %a) #0 {
+entry:
+; CHECK-LABEL: dup_128:
+; CHECK: dup
+  %vecinit = insertelement <8 x half> undef, half %a, i32 0
+  %vecinit1 = insertelement <8 x half> %vecinit, half %a, i32 1
+  %vecinit2 = insertelement <8 x half> %vecinit1, half %a, i32 2
+  %vecinit3 = insertelement <8 x half> %vecinit2, half %a, i32 3
+  %vecinit4 = insertelement <8 x half> %vecinit3, half %a, i32 4
+  %vecinit5 = insertelement <8 x half> %vecinit4, half %a, i32 5
+  %vecinit6 = insertelement <8 x half> %vecinit5, half %a, i32 6
+  %vecinit7 = insertelement <8 x half> %vecinit6, half %a, i32 7
+  ret <8 x half> %vecinit7
+}
+
+; float16x4_t dup_lane_64(float16x4_t a) { return vdup_lane_f16(a, 2); }
+define <4 x half> @dup_lane_64(<4 x half> %a) #0 {
+entry:
+; CHECK-LABEL: dup_lane_64:
+; CHECK: dup
+  %shuffle = shufflevector <4 x half> %a, <4 x half> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
+  ret <4 x half> %shuffle
+}
+
+; float16x8_t dup_lane_128(float16x4_t a) { return vdupq_lane_f16(a, 2); }
+define <8 x half> @dup_lane_128(<4 x half> %a) #0 {
+entry:
+; CHECK-LABEL: dup_lane_128:
+; CHECK: dup
+  %shuffle = shufflevector <4 x half> %a, <4 x half> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+  ret <8 x half> %shuffle
+}
+
+; float16x4_t dup_laneq_64(float16x8_t a) { return vdup_laneq_f16(a, 2); }
+define <4 x half> @dup_laneq_64(<8 x half> %a) #0 {
+entry:
+; CHECK-LABEL: dup_laneq_64:
+; CHECK: dup
+  %shuffle = shufflevector <8 x half> %a, <8 x half> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
+  ret <4 x half> %shuffle
+}
+
+; float16x8_t dup_laneq_128(float16x8_t a) { return vdupq_laneq_f16(a, 2); }
+define <8 x half> @dup_laneq_128(<8 x half> %a) #0 {
+entry:
+; CHECK-LABEL: dup_laneq_128:
+; CHECK: dup
+  %shuffle = shufflevector <8 x half> %a, <8 x half> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+  ret <8 x half> %shuffle
+}
+
+; float16x8_t vcombine(float16x4_t a, float16x4_t b) { return vcombine_f16(a, b); }
+define <8 x half> @vcombine(<4 x half> %a, <4 x half> %b) #0 {
+entry:
+; CHECK-LABEL: vcombine:
+; CHECK: ins
+  %shuffle.i = shufflevector <4 x half> %a, <4 x half> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+  ret <8 x half> %shuffle.i
+}
+
+; float16x4_t get_high(float16x8_t a) { return vget_high_f16(a); }
+define <4 x half> @get_high(<8 x half> %a) #0 {
+; CHECK-LABEL: get_high:
+; CHECK: ext
+entry:
+  %shuffle.i = shufflevector <8 x half> %a, <8 x half> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+  ret <4 x half> %shuffle.i
+}
+
+
+; float16x4_t get_low(float16x8_t a) { return vget_low_f16(a); }
+define <4 x half> @get_low(<8 x half> %a) #0 {
+; CHECK-LABEL: get_low:
+; CHECK-NOT: ext
+entry:
+  %shuffle.i = shufflevector <8 x half> %a, <8 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  ret <4 x half> %shuffle.i
+}
+
+; float16x4_t set_lane_64(float16x4_t a, __fp16 b) { return vset_lane_f16(b, a, 2); }
+define <4 x half> @set_lane_64(<4 x half> %a, half %b) #0 {
+; CHECK-LABEL: set_lane_64:
+; CHECK: fmov
+; CHECK: ins
+entry:
+  %0 = bitcast half %b to i16
+  %1 = bitcast <4 x half> %a to <4 x i16>
+  %vset_lane = insertelement <4 x i16> %1, i16 %0, i32 2
+  %2 = bitcast <4 x i16> %vset_lane to <4 x half>
+  ret <4 x half> %2
+}
+
+
+; float16x8_t set_lane_128(float16x8_t a, __fp16 b) { return vsetq_lane_f16(b, a, 2); }
+define <8 x half> @set_lane_128(<8 x half> %a, half %b) #0 {
+; CHECK-LABEL: set_lane_128:
+; CHECK: fmov
+; CHECK: ins
+entry:
+  %0 = bitcast half %b to i16
+  %1 = bitcast <8 x half> %a to <8 x i16>
+  %vset_lane = insertelement <8 x i16> %1, i16 %0, i32 2
+  %2 = bitcast <8 x i16> %vset_lane to <8 x half>
+  ret <8 x half> %2
+}
+
+; __fp16 get_lane_64(float16x4_t a) { return vget_lane_f16(a, 2); }
+define half @get_lane_64(<4 x half> %a) #0 {
+; CHECK-LABEL: get_lane_64:
+; CHECK: umov
+; CHECK: fmov
+entry:
+  %0 = bitcast <4 x half> %a to <4 x i16>
+  %vget_lane = extractelement <4 x i16> %0, i32 2
+  %1 = bitcast i16 %vget_lane to half
+  ret half %1
+}
+
+; __fp16 get_lane_128(float16x8_t a) { return vgetq_lane_f16(a, 2); }
+define half @get_lane_128(<8 x half> %a) #0 {
+; CHECK-LABEL: get_lane_128:
+; CHECK: umov
+; CHECK: fmov
+entry:
+  %0 = bitcast <8 x half> %a to <8 x i16>
+  %vgetq_lane = extractelement <8 x i16> %0, i32 2
+  %1 = bitcast i16 %vgetq_lane to half
+  ret half %1
+}