diff --git a/llvm/include/llvm/CodeGen/SelectionDAG.h b/llvm/include/llvm/CodeGen/SelectionDAG.h
--- a/llvm/include/llvm/CodeGen/SelectionDAG.h
+++ b/llvm/include/llvm/CodeGen/SelectionDAG.h
@@ -452,6 +452,9 @@
   const DataLayout &getDataLayout() const { return MF->getDataLayout(); }
   const TargetMachine &getTarget() const { return TM; }
   const TargetSubtargetInfo &getSubtarget() const { return MF->getSubtarget(); }
+  template <typename STC> const STC &getSubtarget() const {
+    return MF->getSubtarget<STC>();
+  }
   const TargetLowering &getTargetLoweringInfo() const { return *TLI; }
   const TargetLibraryInfo &getLibInfo() const { return *LibInfo; }
   const SelectionDAGTargetInfo &getSelectionDAGInfo() const { return *TSI; }
diff --git a/llvm/lib/Target/AArch64/AArch64AsmPrinter.cpp b/llvm/lib/Target/AArch64/AArch64AsmPrinter.cpp
--- a/llvm/lib/Target/AArch64/AArch64AsmPrinter.cpp
+++ b/llvm/lib/Target/AArch64/AArch64AsmPrinter.cpp
@@ -132,7 +132,7 @@
 
   bool runOnMachineFunction(MachineFunction &MF) override {
     AArch64FI = MF.getInfo<AArch64FunctionInfo>();
-    STI = static_cast<const AArch64Subtarget*>(&MF.getSubtarget());
+    STI = &MF.getSubtarget<AArch64Subtarget>();
 
     SetupMachineFunction(MF);
 
diff --git a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
--- a/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FalkorHWPFFix.cpp
@@ -813,7 +813,7 @@
 }
 
 bool FalkorHWPFFix::runOnMachineFunction(MachineFunction &Fn) {
-  auto &ST = static_cast<const AArch64Subtarget &>(Fn.getSubtarget());
+  auto &ST = Fn.getSubtarget<AArch64Subtarget>();
   if (ST.getProcFamily() != AArch64Subtarget::Falkor)
     return false;
 
diff --git a/llvm/lib/Target/AArch64/AArch64FastISel.cpp b/llvm/lib/Target/AArch64/AArch64FastISel.cpp
--- a/llvm/lib/Target/AArch64/AArch64FastISel.cpp
+++ b/llvm/lib/Target/AArch64/AArch64FastISel.cpp
@@ -283,8 +283,7 @@
   explicit AArch64FastISel(FunctionLoweringInfo &FuncInfo,
                            const TargetLibraryInfo *LibInfo)
       : FastISel(FuncInfo, LibInfo, /*SkipTargetIndependentISel=*/true) {
-    Subtarget =
-        &static_cast<const AArch64Subtarget &>(FuncInfo.MF->getSubtarget());
+    Subtarget = &FuncInfo.MF->getSubtarget<AArch64Subtarget>();
     Context = &FuncInfo.Fn->getContext();
   }
 
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -2576,8 +2576,7 @@
   EVT VT = LHS.getValueType();
   assert(VT != MVT::f128);
 
-  const bool FullFP16 =
-      static_cast<const AArch64Subtarget &>(DAG.getSubtarget()).hasFullFP16();
+  const bool FullFP16 = DAG.getSubtarget<AArch64Subtarget>().hasFullFP16();
 
   if (VT == MVT::f16 && !FullFP16) {
     LHS = DAG.getNode(ISD::STRICT_FP_EXTEND, dl, {MVT::f32, MVT::Other},
@@ -2595,8 +2594,7 @@
 static SDValue emitComparison(SDValue LHS, SDValue RHS, ISD::CondCode CC,
                               const SDLoc &dl, SelectionDAG &DAG) {
   EVT VT = LHS.getValueType();
-  const bool FullFP16 =
-    static_cast<const AArch64Subtarget &>(DAG.getSubtarget()).hasFullFP16();
+  const bool FullFP16 = DAG.getSubtarget<AArch64Subtarget>().hasFullFP16();
 
   if (VT.isFloatingPoint()) {
     assert(VT != MVT::f128);
@@ -2704,8 +2702,7 @@
                                          AArch64CC::CondCode OutCC,
                                          const SDLoc &DL, SelectionDAG &DAG) {
   unsigned Opcode = 0;
-  const bool FullFP16 =
-    static_cast<const AArch64Subtarget &>(DAG.getSubtarget()).hasFullFP16();
+  const bool FullFP16 = DAG.getSubtarget<AArch64Subtarget>().hasFullFP16();
 
   if (LHS.getValueType().isFloatingPoint()) {
     assert(LHS.getValueType() != MVT::f128);
@@ -11823,8 +11820,7 @@
     return DAG.getSExtOrTrunc(Cmp, dl, Op.getValueType());
   }
 
-  const bool FullFP16 =
-    static_cast<const AArch64Subtarget &>(DAG.getSubtarget()).hasFullFP16();
+  const bool FullFP16 = DAG.getSubtarget<AArch64Subtarget>().hasFullFP16();
 
   // Make v4f16 (only) fcmp operations utilise vector instructions
   // v8f16 support will be a litle more complicated
@@ -11958,7 +11954,7 @@
 
 SDValue AArch64TargetLowering::LowerATOMIC_LOAD_SUB(SDValue Op,
                                                     SelectionDAG &DAG) const {
-  auto &Subtarget = static_cast<const AArch64Subtarget &>(DAG.getSubtarget());
+  auto &Subtarget = DAG.getSubtarget<AArch64Subtarget>();
   if (!Subtarget.hasLSE() && !Subtarget.outlineAtomics())
     return SDValue();
 
@@ -11975,7 +11971,7 @@
 
 SDValue AArch64TargetLowering::LowerATOMIC_LOAD_AND(SDValue Op,
                                                     SelectionDAG &DAG) const {
-  auto &Subtarget = static_cast<const AArch64Subtarget &>(DAG.getSubtarget());
+  auto &Subtarget = DAG.getSubtarget<AArch64Subtarget>();
   if (!Subtarget.hasLSE() && !Subtarget.outlineAtomics())
     return SDValue();
 
@@ -14732,8 +14728,7 @@
   ConstantSDNode *ConstantN1 = dyn_cast<ConstantSDNode>(N1);
 
   EVT VT = N->getValueType(0);
-  const bool FullFP16 =
-      static_cast<const AArch64Subtarget &>(DAG.getSubtarget()).hasFullFP16();
+  const bool FullFP16 = DAG.getSubtarget<AArch64Subtarget>().hasFullFP16();
   bool IsStrict = N0->isStrictFPOpcode();
 
   // Rewrite for pairwise fadd pattern
@@ -15918,8 +15913,7 @@
   // If we're compiling for a specific vector-length, we can check if the
   // pattern's VL equals that of the scalable vector at runtime.
   if (N.getOpcode() == AArch64ISD::PTRUE) {
-    const auto &Subtarget =
-        static_cast<const AArch64Subtarget &>(DAG.getSubtarget());
+    const auto &Subtarget = DAG.getSubtarget<AArch64Subtarget>();
     unsigned MinSVESize = Subtarget.getMinSVEVectorSizeInBits();
     unsigned MaxSVESize = Subtarget.getMaxSVEVectorSizeInBits();
     if (MaxSVESize && MinSVESize == MaxSVESize) {
@@ -17140,8 +17134,7 @@
       Stride > std::numeric_limits<int32_t>::max())
     return Changed;
 
-  const auto &Subtarget =
-      static_cast<const AArch64Subtarget &>(DAG.getSubtarget());
+  const auto &Subtarget = DAG.getSubtarget<AArch64Subtarget>();
   unsigned MaxVScale =
       Subtarget.getMaxSVEVectorSizeInBits() / AArch64::SVEBitsPerBlock;
   int64_t LastElementOffset =
@@ -20074,8 +20067,7 @@
   // For vectors that are exactly getMaxSVEVectorSizeInBits big, we can use
   // AArch64SVEPredPattern::all, which can enable the use of unpredicated
   // variants of instructions when available.
-  const auto &Subtarget =
-      static_cast<const AArch64Subtarget &>(DAG.getSubtarget());
+  const auto &Subtarget = DAG.getSubtarget<AArch64Subtarget>();
   unsigned MinSVESize = Subtarget.getMinSVEVectorSizeInBits();
   unsigned MaxSVESize = Subtarget.getMaxSVEVectorSizeInBits();
   if (MaxSVESize && MinSVESize == MaxSVESize &&
diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -2308,7 +2308,7 @@
   if (skipFunction(Fn.getFunction()))
     return false;
 
-  Subtarget = &static_cast<const AArch64Subtarget &>(Fn.getSubtarget());
+  Subtarget = &Fn.getSubtarget<AArch64Subtarget>();
   TII = static_cast<const AArch64InstrInfo *>(Subtarget->getInstrInfo());
   TRI = Subtarget->getRegisterInfo();
   AA = &getAnalysis<AAResultsWrapperPass>().getAAResults();
diff --git a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
--- a/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64InstructionSelector.cpp
@@ -2325,8 +2325,7 @@
   MachineFunction &MF = *MBB.getParent();
   MachineRegisterInfo &MRI = MF.getRegInfo();
 
-  const AArch64Subtarget *Subtarget =
-      &static_cast<const AArch64Subtarget &>(MF.getSubtarget());
+  const AArch64Subtarget *Subtarget = &MF.getSubtarget<AArch64Subtarget>();
   if (Subtarget->requiresStrictAlign()) {
     // We don't support this feature yet.
     LLVM_DEBUG(dbgs() << "AArch64 GISel does not support strict-align yet\n");
diff --git a/llvm/lib/Target/ARM/ARMBlockPlacement.cpp b/llvm/lib/Target/ARM/ARMBlockPlacement.cpp
--- a/llvm/lib/Target/ARM/ARMBlockPlacement.cpp
+++ b/llvm/lib/Target/ARM/ARMBlockPlacement.cpp
@@ -213,7 +213,7 @@
 bool ARMBlockPlacement::runOnMachineFunction(MachineFunction &MF) {
   if (skipFunction(MF.getFunction()))
     return false;
-  const ARMSubtarget &ST = static_cast<const ARMSubtarget &>(MF.getSubtarget());
+  const ARMSubtarget &ST = MF.getSubtarget<ARMSubtarget>();
   if (!ST.hasLOB())
     return false;
   LLVM_DEBUG(dbgs() << DEBUG_PREFIX << "Running on " << MF.getName() << "\n");
diff --git a/llvm/lib/Target/ARM/ARMConstantIslandPass.cpp b/llvm/lib/Target/ARM/ARMConstantIslandPass.cpp
--- a/llvm/lib/Target/ARM/ARMConstantIslandPass.cpp
+++ b/llvm/lib/Target/ARM/ARMConstantIslandPass.cpp
@@ -396,7 +396,7 @@
                     << MCP->getConstants().size() << " CP entries, aligned to "
                     << MCP->getConstantPoolAlign().value() << " bytes *****\n");
 
-  STI = &static_cast<const ARMSubtarget &>(MF->getSubtarget());
+  STI = &MF->getSubtarget<ARMSubtarget>();
   TII = STI->getInstrInfo();
   isPositionIndependentOrROPI =
       STI->getTargetLowering()->isPositionIndependent() || STI->isROPI();
diff --git a/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp b/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp
--- a/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp
+++ b/llvm/lib/Target/ARM/ARMExpandPseudoInsts.cpp
@@ -3132,7 +3132,7 @@
 }
 
 bool ARMExpandPseudo::runOnMachineFunction(MachineFunction &MF) {
-  STI = &static_cast<const ARMSubtarget &>(MF.getSubtarget());
+  STI = &MF.getSubtarget<ARMSubtarget>();
   TII = STI->getInstrInfo();
   TRI = STI->getRegisterInfo();
   AFI = MF.getInfo<ARMFunctionInfo>();
diff --git a/llvm/lib/Target/ARM/ARMFastISel.cpp b/llvm/lib/Target/ARM/ARMFastISel.cpp
--- a/llvm/lib/Target/ARM/ARMFastISel.cpp
+++ b/llvm/lib/Target/ARM/ARMFastISel.cpp
@@ -122,8 +122,7 @@
     explicit ARMFastISel(FunctionLoweringInfo &funcInfo,
                          const TargetLibraryInfo *libInfo)
         : FastISel(funcInfo, libInfo),
-          Subtarget(
-              &static_cast<const ARMSubtarget &>(funcInfo.MF->getSubtarget())),
+          Subtarget(&funcInfo.MF->getSubtarget<ARMSubtarget>()),
           M(const_cast<Module &>(*funcInfo.Fn->getParent())),
           TM(funcInfo.MF->getTarget()), TII(*Subtarget->getInstrInfo()),
           TLI(*Subtarget->getTargetLowering()) {
@@ -156,7 +155,7 @@
                              const LoadInst *LI) override;
     bool fastLowerArguments() override;
 
-  #include "ARMGenFastISel.inc"
+#include "ARMGenFastISel.inc"
 
     // Instruction selection routines.
 
diff --git a/llvm/lib/Target/ARM/ARMFrameLowering.cpp b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
--- a/llvm/lib/Target/ARM/ARMFrameLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMFrameLowering.cpp
@@ -392,8 +392,7 @@
                                      const DebugLoc &DL, const unsigned Reg,
                                      const Align Alignment,
                                      const bool MustBeSingleInstruction) {
-  const ARMSubtarget &AST =
-      static_cast<const ARMSubtarget &>(MF.getSubtarget());
+  const ARMSubtarget &AST = MF.getSubtarget<ARMSubtarget>();
   const bool CanUseBFC = AST.hasV6T2Ops() || AST.hasV7Ops();
   const unsigned AlignMask = Alignment.value() - 1U;
   const unsigned NrBitsToZero = Log2(Alignment);
@@ -1768,7 +1767,7 @@
     return;
 
   // We are planning to use NEON instructions vst1 / vld1.
-  if (!static_cast<const ARMSubtarget &>(MF.getSubtarget()).hasNEON())
+  if (!MF.getSubtarget<ARMSubtarget>().hasNEON())
     return;
 
   // Don't bother if the default stack alignment is sufficiently high.
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -5809,8 +5809,7 @@
     return DAG.UnrollVectorOp(Op.getNode());
   }
 
-  const bool HasFullFP16 =
-    static_cast<const ARMSubtarget&>(DAG.getSubtarget()).hasFullFP16();
+  const bool HasFullFP16 = DAG.getSubtarget<ARMSubtarget>().hasFullFP16();
 
   EVT NewTy;
   const EVT OpTy = Op.getOperand(0).getValueType();
@@ -5920,8 +5919,7 @@
           Op.getOperand(0).getValueType() == MVT::v8i16) &&
          "Invalid type for custom lowering!");
 
-  const bool HasFullFP16 =
-    static_cast<const ARMSubtarget&>(DAG.getSubtarget()).hasFullFP16();
+  const bool HasFullFP16 = DAG.getSubtarget<ARMSubtarget>().hasFullFP16();
 
   EVT DestVecType;
   if (VT == MVT::v4f32)
@@ -9884,7 +9882,7 @@
   if (N->getOpcode() != ISD::SDIV)
     return SDValue();
 
-  const auto &ST = static_cast<const ARMSubtarget&>(DAG.getSubtarget());
+  const auto &ST = DAG.getSubtarget<ARMSubtarget>();
   const bool MinSize = ST.hasMinSize();
   const bool HasDivide = ST.isThumb() ? ST.hasDivideInThumbMode()
                                       : ST.hasDivideInARMMode();
diff --git a/llvm/lib/Target/ARM/ARMLoadStoreOptimizer.cpp b/llvm/lib/Target/ARM/ARMLoadStoreOptimizer.cpp
--- a/llvm/lib/Target/ARM/ARMLoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/ARM/ARMLoadStoreOptimizer.cpp
@@ -2110,7 +2110,7 @@
     return false;
 
   MF = &Fn;
-  STI = &static_cast<const ARMSubtarget &>(Fn.getSubtarget());
+  STI = &Fn.getSubtarget<ARMSubtarget>();
   TL = STI->getTargetLowering();
   AFI = Fn.getInfo<ARMFunctionInfo>();
   TII = STI->getInstrInfo();
@@ -2201,7 +2201,7 @@
     return false;
 
   TD = &Fn.getDataLayout();
-  STI = &static_cast<const ARMSubtarget &>(Fn.getSubtarget());
+  STI = &Fn.getSubtarget<ARMSubtarget>();
   TII = STI->getInstrInfo();
   TRI = STI->getRegisterInfo();
   MRI = &Fn.getRegInfo();
diff --git a/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp b/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
--- a/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
+++ b/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
@@ -1299,7 +1299,7 @@
 }
 
 bool ARMLowOverheadLoops::runOnMachineFunction(MachineFunction &mf) {
-  const ARMSubtarget &ST = static_cast<const ARMSubtarget&>(mf.getSubtarget());
+  const ARMSubtarget &ST = mf.getSubtarget<ARMSubtarget>();
   if (!ST.hasLOB())
     return false;
 
diff --git a/llvm/lib/Target/ARM/MVETPAndVPTOptimisationsPass.cpp b/llvm/lib/Target/ARM/MVETPAndVPTOptimisationsPass.cpp
--- a/llvm/lib/Target/ARM/MVETPAndVPTOptimisationsPass.cpp
+++ b/llvm/lib/Target/ARM/MVETPAndVPTOptimisationsPass.cpp
@@ -1041,8 +1041,7 @@
 }
 
 bool MVETPAndVPTOptimisations::runOnMachineFunction(MachineFunction &Fn) {
-  const ARMSubtarget &STI =
-      static_cast<const ARMSubtarget &>(Fn.getSubtarget());
+  const ARMSubtarget &STI = Fn.getSubtarget<ARMSubtarget>();
 
   if (!STI.isThumb2() || !STI.hasLOB())
     return false;
diff --git a/llvm/lib/Target/ARM/MVEVPTBlockPass.cpp b/llvm/lib/Target/ARM/MVEVPTBlockPass.cpp
--- a/llvm/lib/Target/ARM/MVEVPTBlockPass.cpp
+++ b/llvm/lib/Target/ARM/MVEVPTBlockPass.cpp
@@ -312,8 +312,7 @@
 }
 
 bool MVEVPTBlock::runOnMachineFunction(MachineFunction &Fn) {
-  const ARMSubtarget &STI =
-      static_cast<const ARMSubtarget &>(Fn.getSubtarget());
+  const ARMSubtarget &STI = Fn.getSubtarget<ARMSubtarget>();
 
   if (!STI.isThumb2() || !STI.hasMVEIntegerOps())
     return false;
diff --git a/llvm/lib/Target/ARM/Thumb2ITBlockPass.cpp b/llvm/lib/Target/ARM/Thumb2ITBlockPass.cpp
--- a/llvm/lib/Target/ARM/Thumb2ITBlockPass.cpp
+++ b/llvm/lib/Target/ARM/Thumb2ITBlockPass.cpp
@@ -284,8 +284,7 @@
 }
 
 bool Thumb2ITBlock::runOnMachineFunction(MachineFunction &Fn) {
-  const ARMSubtarget &STI =
-      static_cast<const ARMSubtarget &>(Fn.getSubtarget());
+  const ARMSubtarget &STI = Fn.getSubtarget<ARMSubtarget>();
   if (!STI.isThumb2())
     return false;
   AFI = Fn.getInfo<ARMFunctionInfo>();
diff --git a/llvm/lib/Target/ARM/Thumb2SizeReduction.cpp b/llvm/lib/Target/ARM/Thumb2SizeReduction.cpp
--- a/llvm/lib/Target/ARM/Thumb2SizeReduction.cpp
+++ b/llvm/lib/Target/ARM/Thumb2SizeReduction.cpp
@@ -1130,7 +1130,7 @@
   if (PredicateFtor && !PredicateFtor(MF.getFunction()))
     return false;
 
-  STI = &static_cast<const ARMSubtarget &>(MF.getSubtarget());
+  STI = &MF.getSubtarget<ARMSubtarget>();
   if (STI->isThumb1Only() || STI->prefers32BitThumb())
     return false;
 
diff --git a/llvm/lib/Target/CSKY/CSKYConstantIslandPass.cpp b/llvm/lib/Target/CSKY/CSKYConstantIslandPass.cpp
--- a/llvm/lib/Target/CSKY/CSKYConstantIslandPass.cpp
+++ b/llvm/lib/Target/CSKY/CSKYConstantIslandPass.cpp
@@ -287,7 +287,7 @@
 bool CSKYConstantIslands::runOnMachineFunction(MachineFunction &Mf) {
   MF = &Mf;
   MCP = Mf.getConstantPool();
-  STI = &static_cast<const CSKYSubtarget &>(Mf.getSubtarget());
+  STI = &Mf.getSubtarget<CSKYSubtarget>();
 
   LLVM_DEBUG(dbgs() << "***** CSKYConstantIslands: "
                     << MCP->getConstants().size() << " CP entries, aligned to "
diff --git a/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp b/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp
--- a/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelDAGToDAGHVX.cpp
@@ -801,7 +801,7 @@
   return static_cast<const HexagonTargetLowering&>(G.getTargetLoweringInfo());
 }
 static const HexagonSubtarget &getHexagonSubtarget(SelectionDAG &G) {
-  return static_cast<const HexagonSubtarget&>(G.getSubtarget());
+  return G.getSubtarget<HexagonSubtarget>();
 }
 
 namespace llvm {
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -1396,10 +1396,9 @@
   Chain = DAG.getCopyToReg(DAG.getEntryNode(), dl, Hexagon::R0, Chain, InFlag);
   InFlag = Chain.getValue(1);
 
-  unsigned Flags =
-      static_cast<const HexagonSubtarget &>(DAG.getSubtarget()).useLongCalls()
-          ? HexagonII::MO_GDPLT | HexagonII::HMOTF_ConstExtended
-          : HexagonII::MO_GDPLT;
+  unsigned Flags = DAG.getSubtarget<HexagonSubtarget>().useLongCalls()
+                       ? HexagonII::MO_GDPLT | HexagonII::HMOTF_ConstExtended
+                       : HexagonII::MO_GDPLT;
 
   return GetDynamicTLSAddr(DAG, Chain, GA, InFlag, PtrVT,
                            Hexagon::R0, Flags);
diff --git a/llvm/lib/Target/Hexagon/HexagonPatternsHVX.td b/llvm/lib/Target/Hexagon/HexagonPatternsHVX.td
--- a/llvm/lib/Target/Hexagon/HexagonPatternsHVX.td
+++ b/llvm/lib/Target/Hexagon/HexagonPatternsHVX.td
@@ -37,7 +37,7 @@
 def HexagonVINSERTW0: SDNode<"HexagonISD::VINSERTW0", SDTHexagonVINSERTW0>;
 
 def HwLen2: SDNodeXForm<imm, [{
-  const auto &ST = static_cast<const HexagonSubtarget&>(CurDAG->getSubtarget());
+  const auto &ST = CurDAG->getSubtarget<HexagonSubtarget>();
   return CurDAG->getTargetConstant(ST.getVectorLength()/2, SDLoc(N), MVT::i32);
 }]>;
 
diff --git a/llvm/lib/Target/M68k/M68kCollapseMOVEMPass.cpp b/llvm/lib/Target/M68k/M68kCollapseMOVEMPass.cpp
--- a/llvm/lib/Target/M68k/M68kCollapseMOVEMPass.cpp
+++ b/llvm/lib/Target/M68k/M68kCollapseMOVEMPass.cpp
@@ -231,7 +231,7 @@
   }
 
   bool runOnMachineFunction(MachineFunction &MF) override {
-    STI = &static_cast<const M68kSubtarget &>(MF.getSubtarget());
+    STI = &MF.getSubtarget<M68kSubtarget>();
     TII = STI->getInstrInfo();
     TRI = STI->getRegisterInfo();
     MFI = MF.getInfo<M68kMachineFunctionInfo>();
diff --git a/llvm/lib/Target/M68k/M68kExpandPseudo.cpp b/llvm/lib/Target/M68k/M68kExpandPseudo.cpp
--- a/llvm/lib/Target/M68k/M68kExpandPseudo.cpp
+++ b/llvm/lib/Target/M68k/M68kExpandPseudo.cpp
@@ -302,7 +302,7 @@
 }
 
 bool M68kExpandPseudo::runOnMachineFunction(MachineFunction &MF) {
-  STI = &static_cast<const M68kSubtarget &>(MF.getSubtarget());
+  STI = &MF.getSubtarget<M68kSubtarget>();
   TII = STI->getInstrInfo();
   TRI = STI->getRegisterInfo();
   MFI = MF.getInfo<M68kMachineFunctionInfo>();
diff --git a/llvm/lib/Target/M68k/M68kISelDAGToDAG.cpp b/llvm/lib/Target/M68k/M68kISelDAGToDAG.cpp
--- a/llvm/lib/Target/M68k/M68kISelDAGToDAG.cpp
+++ b/llvm/lib/Target/M68k/M68kISelDAGToDAG.cpp
@@ -312,7 +312,7 @@
 } // namespace
 
 bool M68kDAGToDAGISel::runOnMachineFunction(MachineFunction &MF) {
-  Subtarget = &static_cast<const M68kSubtarget &>(MF.getSubtarget());
+  Subtarget = &MF.getSubtarget<M68kSubtarget>();
   return SelectionDAGISel::runOnMachineFunction(MF);
 }
 
diff --git a/llvm/lib/Target/Mips/MicroMipsSizeReduction.cpp b/llvm/lib/Target/Mips/MicroMipsSizeReduction.cpp
--- a/llvm/lib/Target/Mips/MicroMipsSizeReduction.cpp
+++ b/llvm/lib/Target/Mips/MicroMipsSizeReduction.cpp
@@ -774,7 +774,7 @@
 
 bool MicroMipsSizeReduce::runOnMachineFunction(MachineFunction &MF) {
 
-  Subtarget = &static_cast<const MipsSubtarget &>(MF.getSubtarget());
+  Subtarget = &MF.getSubtarget<MipsSubtarget>();
 
   // TODO: Add support for the subtarget microMIPS32R6.
   if (!Subtarget->inMicroMipsMode() || !Subtarget->hasMips32r2() ||
diff --git a/llvm/lib/Target/Mips/Mips16ISelDAGToDAG.cpp b/llvm/lib/Target/Mips/Mips16ISelDAGToDAG.cpp
--- a/llvm/lib/Target/Mips/Mips16ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/Mips/Mips16ISelDAGToDAG.cpp
@@ -35,7 +35,7 @@
 #define DEBUG_TYPE "mips-isel"
 
 bool Mips16DAGToDAGISel::runOnMachineFunction(MachineFunction &MF) {
-  Subtarget = &static_cast<const MipsSubtarget &>(MF.getSubtarget());
+  Subtarget = &MF.getSubtarget<MipsSubtarget>();
   if (!Subtarget->inMips16Mode())
     return false;
   return MipsDAGToDAGISel::runOnMachineFunction(MF);
diff --git a/llvm/lib/Target/Mips/MipsBranchExpansion.cpp b/llvm/lib/Target/Mips/MipsBranchExpansion.cpp
--- a/llvm/lib/Target/Mips/MipsBranchExpansion.cpp
+++ b/llvm/lib/Target/Mips/MipsBranchExpansion.cpp
@@ -880,7 +880,7 @@
   const TargetMachine &TM = MF.getTarget();
   IsPIC = TM.isPositionIndependent();
   ABI = static_cast<const MipsTargetMachine &>(TM).getABI();
-  STI = &static_cast<const MipsSubtarget &>(MF.getSubtarget());
+  STI = &MF.getSubtarget<MipsSubtarget>();
   TII = static_cast<const MipsInstrInfo *>(STI->getInstrInfo());
 
   if (IsPIC && ABI.IsO32() &&
diff --git a/llvm/lib/Target/Mips/MipsCallLowering.cpp b/llvm/lib/Target/Mips/MipsCallLowering.cpp
--- a/llvm/lib/Target/Mips/MipsCallLowering.cpp
+++ b/llvm/lib/Target/Mips/MipsCallLowering.cpp
@@ -541,8 +541,7 @@
   }
   MIRBuilder.insertInstr(MIB);
   if (MIB->getOpcode() == Mips::JALRPseudo) {
-    const MipsSubtarget &STI =
-        static_cast<const MipsSubtarget &>(MIRBuilder.getMF().getSubtarget());
+    const MipsSubtarget &STI = MIRBuilder.getMF().getSubtarget<MipsSubtarget>();
     MIB.constrainAllUses(MIRBuilder.getTII(), *STI.getRegisterInfo(),
                          *STI.getRegBankInfo());
   }
diff --git a/llvm/lib/Target/Mips/MipsConstantIslandPass.cpp b/llvm/lib/Target/Mips/MipsConstantIslandPass.cpp
--- a/llvm/lib/Target/Mips/MipsConstantIslandPass.cpp
+++ b/llvm/lib/Target/Mips/MipsConstantIslandPass.cpp
@@ -436,7 +436,7 @@
   // FIXME:
   MF = &mf;
   MCP = mf.getConstantPool();
-  STI = &static_cast<const MipsSubtarget &>(mf.getSubtarget());
+  STI = &mf.getSubtarget<MipsSubtarget>();
   LLVM_DEBUG(dbgs() << "constant island machine function "
                     << "\n");
   if (!STI->inMips16Mode() || !MipsSubtarget::useConstantIslands()) {
diff --git a/llvm/lib/Target/Mips/MipsExpandPseudo.cpp b/llvm/lib/Target/Mips/MipsExpandPseudo.cpp
--- a/llvm/lib/Target/Mips/MipsExpandPseudo.cpp
+++ b/llvm/lib/Target/Mips/MipsExpandPseudo.cpp
@@ -892,7 +892,7 @@
 }
 
 bool MipsExpandPseudo::runOnMachineFunction(MachineFunction &MF) {
-  STI = &static_cast<const MipsSubtarget &>(MF.getSubtarget());
+  STI = &MF.getSubtarget<MipsSubtarget>();
   TII = STI->getInstrInfo();
 
   bool Modified = false;
diff --git a/llvm/lib/Target/Mips/MipsISelDAGToDAG.cpp b/llvm/lib/Target/Mips/MipsISelDAGToDAG.cpp
--- a/llvm/lib/Target/Mips/MipsISelDAGToDAG.cpp
+++ b/llvm/lib/Target/Mips/MipsISelDAGToDAG.cpp
@@ -54,7 +54,7 @@
 }
 
 bool MipsDAGToDAGISel::runOnMachineFunction(MachineFunction &MF) {
-  Subtarget = &static_cast<const MipsSubtarget &>(MF.getSubtarget());
+  Subtarget = &MF.getSubtarget<MipsSubtarget>();
   bool Ret = SelectionDAGISel::runOnMachineFunction(MF);
 
   processFunctionAfterISel(MF);
diff --git a/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp b/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp
--- a/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp
+++ b/llvm/lib/Target/Mips/MipsLegalizerInfo.cpp
@@ -503,8 +503,7 @@
 bool MipsLegalizerInfo::legalizeIntrinsic(LegalizerHelper &Helper,
                                           MachineInstr &MI) const {
   MachineIRBuilder &MIRBuilder = Helper.MIRBuilder;
-  const MipsSubtarget &ST =
-      static_cast<const MipsSubtarget &>(MI.getMF()->getSubtarget());
+  const MipsSubtarget &ST = MI.getMF()->getSubtarget<MipsSubtarget>();
   const MipsInstrInfo &TII = *ST.getInstrInfo();
   const MipsRegisterInfo &TRI = *ST.getRegisterInfo();
   const RegisterBankInfo &RBI = *ST.getRegBankInfo();
diff --git a/llvm/lib/Target/Mips/MipsMachineFunction.cpp b/llvm/lib/Target/Mips/MipsMachineFunction.cpp
--- a/llvm/lib/Target/Mips/MipsMachineFunction.cpp
+++ b/llvm/lib/Target/Mips/MipsMachineFunction.cpp
@@ -29,7 +29,7 @@
 }
 
 static const TargetRegisterClass &getGlobalBaseRegClass(MachineFunction &MF) {
-  auto &STI = static_cast<const MipsSubtarget &>(MF.getSubtarget());
+  auto &STI = MF.getSubtarget<MipsSubtarget>();
   auto &TM = static_cast<const MipsTargetMachine &>(MF.getTarget());
 
   if (STI.inMips16Mode())
diff --git a/llvm/lib/Target/Mips/MipsOptimizePICCall.cpp b/llvm/lib/Target/Mips/MipsOptimizePICCall.cpp
--- a/llvm/lib/Target/Mips/MipsOptimizePICCall.cpp
+++ b/llvm/lib/Target/Mips/MipsOptimizePICCall.cpp
@@ -194,7 +194,7 @@
 
 // OptimizePICCall methods.
 bool OptimizePICCall::runOnMachineFunction(MachineFunction &F) {
-  if (static_cast<const MipsSubtarget &>(F.getSubtarget()).inMips16Mode())
+  if (F.getSubtarget<MipsSubtarget>().inMips16Mode())
     return false;
 
   // Do a pre-order traversal of the dominator tree.
diff --git a/llvm/lib/Target/Mips/MipsPreLegalizerCombiner.cpp b/llvm/lib/Target/Mips/MipsPreLegalizerCombiner.cpp
--- a/llvm/lib/Target/Mips/MipsPreLegalizerCombiner.cpp
+++ b/llvm/lib/Target/Mips/MipsPreLegalizerCombiner.cpp
@@ -51,8 +51,7 @@
     // Don't attempt to combine non power of 2 loads or unaligned loads when
     // subtarget doesn't support them.
     auto MMO = *MI.memoperands_begin();
-    const MipsSubtarget &STI =
-        static_cast<const MipsSubtarget &>(MI.getMF()->getSubtarget());
+    const MipsSubtarget &STI = MI.getMF()->getSubtarget<MipsSubtarget>();
     if (!isPowerOf2_64(MMO->getSize()))
       return false;
     bool isUnaligned = MMO->getAlign() < MMO->getSize();
diff --git a/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp b/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp
--- a/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp
+++ b/llvm/lib/Target/Mips/MipsRegisterBankInfo.cpp
@@ -153,8 +153,7 @@
   if (MI->getOpcode() == TargetOpcode::G_LOAD ||
       MI->getOpcode() == TargetOpcode::G_STORE) {
     auto MMO = *MI->memoperands_begin();
-    const MipsSubtarget &STI =
-        static_cast<const MipsSubtarget &>(MI->getMF()->getSubtarget());
+    const MipsSubtarget &STI = MI->getMF()->getSubtarget<MipsSubtarget>();
     if (MMO->getSize() == 4 && (!STI.systemSupportsUnalignedAccess() &&
                                 MMO->getAlign() < MMO->getSize()))
       return true;
@@ -398,7 +397,7 @@
 
 static const MipsRegisterBankInfo::ValueMapping *
 getMSAMapping(const MachineFunction &MF) {
-  assert(static_cast<const MipsSubtarget &>(MF.getSubtarget()).hasMSA() &&
+  assert(MF.getSubtarget<MipsSubtarget>().hasMSA() &&
          "MSA mapping not available on target without MSA.");
   return &Mips::ValueMappings[Mips::MSAIdx];
 }
diff --git a/llvm/lib/Target/Mips/MipsSEFrameLowering.cpp b/llvm/lib/Target/Mips/MipsSEFrameLowering.cpp
--- a/llvm/lib/Target/Mips/MipsSEFrameLowering.cpp
+++ b/llvm/lib/Target/Mips/MipsSEFrameLowering.cpp
@@ -97,7 +97,7 @@
 
 ExpandPseudo::ExpandPseudo(MachineFunction &MF_)
     : MF(MF_), MRI(MF.getRegInfo()),
-      Subtarget(static_cast<const MipsSubtarget &>(MF.getSubtarget())),
+      Subtarget(MF.getSubtarget<MipsSubtarget>()),
       TII(*static_cast<const MipsSEInstrInfo *>(Subtarget.getInstrInfo())),
       RegInfo(*Subtarget.getRegisterInfo()) {}
 
diff --git a/llvm/lib/Target/Mips/MipsSEISelDAGToDAG.cpp b/llvm/lib/Target/Mips/MipsSEISelDAGToDAG.cpp
--- a/llvm/lib/Target/Mips/MipsSEISelDAGToDAG.cpp
+++ b/llvm/lib/Target/Mips/MipsSEISelDAGToDAG.cpp
@@ -38,7 +38,7 @@
 #define DEBUG_TYPE "mips-isel"
 
 bool MipsSEDAGToDAGISel::runOnMachineFunction(MachineFunction &MF) {
-  Subtarget = &static_cast<const MipsSubtarget &>(MF.getSubtarget());
+  Subtarget = &MF.getSubtarget<MipsSubtarget>();
   if (Subtarget->inMips16Mode())
     return false;
   return MipsDAGToDAGISel::runOnMachineFunction(MF);
diff --git a/llvm/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp b/llvm/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp
--- a/llvm/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp
+++ b/llvm/lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp
@@ -42,7 +42,7 @@
 }
 
 bool NVPTXDAGToDAGISel::runOnMachineFunction(MachineFunction &MF) {
-  Subtarget = &static_cast<const NVPTXSubtarget &>(MF.getSubtarget());
+  Subtarget = &MF.getSubtarget<NVPTXSubtarget>();
   return SelectionDAGISel::runOnMachineFunction(MF);
 }
 
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -1883,8 +1883,7 @@
 /// For the latter, the input operands are swapped (see PPCInstrAltivec.td).
 bool PPC::isVPKUDUMShuffleMask(ShuffleVectorSDNode *N, unsigned ShuffleKind,
                                SelectionDAG &DAG) {
-  const PPCSubtarget& Subtarget =
-      static_cast<const PPCSubtarget&>(DAG.getSubtarget());
+  const PPCSubtarget &Subtarget = DAG.getSubtarget<PPCSubtarget>();
   if (!Subtarget.hasP8Vector())
     return false;
 
@@ -6912,8 +6911,7 @@
   if (useSoftFloat())
     report_fatal_error("Soft float support is unimplemented on AIX.");
 
-  const PPCSubtarget &Subtarget =
-      static_cast<const PPCSubtarget &>(DAG.getSubtarget());
+  const PPCSubtarget &Subtarget = DAG.getSubtarget<PPCSubtarget>();
 
   const bool IsPPC64 = Subtarget.isPPC64();
   const unsigned PtrByteSize = IsPPC64 ? 8 : 4;
@@ -7218,8 +7216,7 @@
   if (CFlags.IsPatchPoint)
     report_fatal_error("This call type is unimplemented on AIX.");
 
-  const PPCSubtarget& Subtarget =
-      static_cast<const PPCSubtarget&>(DAG.getSubtarget());
+  const PPCSubtarget &Subtarget = DAG.getSubtarget<PPCSubtarget>();
 
   MachineFunction &MF = DAG.getMachineFunction();
   SmallVector<CCValAssign, 16> ArgLocs;
diff --git a/llvm/lib/Target/SystemZ/SystemZAsmPrinter.cpp b/llvm/lib/Target/SystemZ/SystemZAsmPrinter.cpp
--- a/llvm/lib/Target/SystemZ/SystemZAsmPrinter.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZAsmPrinter.cpp
@@ -820,8 +820,7 @@
 }
 
 void SystemZAsmPrinter::emitFunctionEntryLabel() {
-  const SystemZSubtarget &Subtarget =
-      static_cast<const SystemZSubtarget &>(MF->getSubtarget());
+  const SystemZSubtarget &Subtarget = MF->getSubtarget<SystemZSubtarget>();
 
   if (Subtarget.getTargetTriple().isOSzOS()) {
     MCContext &OutContext = OutStreamer->getContext();
diff --git a/llvm/lib/Target/X86/X86CallingConv.cpp b/llvm/lib/Target/X86/X86CallingConv.cpp
--- a/llvm/lib/Target/X86/X86CallingConv.cpp
+++ b/llvm/lib/Target/X86/X86CallingConv.cpp
@@ -299,7 +299,7 @@
                         ISD::ArgFlagsTy &ArgFlags, CCState &State) {
   const MachineFunction &MF = State.getMachineFunction();
   size_t ArgCount = State.getMachineFunction().getFunction().arg_size();
-  bool Is64Bit = static_cast<const X86Subtarget &>(MF.getSubtarget()).is64Bit();
+  bool Is64Bit = MF.getSubtarget<X86Subtarget>().is64Bit();
   unsigned SlotSize = Is64Bit ? 8 : 4;
   unsigned Offset;
   if (ArgCount == 1 && ValNo == 0) {
diff --git a/llvm/lib/Target/X86/X86ExpandPseudo.cpp b/llvm/lib/Target/X86/X86ExpandPseudo.cpp
--- a/llvm/lib/Target/X86/X86ExpandPseudo.cpp
+++ b/llvm/lib/Target/X86/X86ExpandPseudo.cpp
@@ -730,7 +730,7 @@
 }
 
 bool X86ExpandPseudo::runOnMachineFunction(MachineFunction &MF) {
-  STI = &static_cast<const X86Subtarget &>(MF.getSubtarget());
+  STI = &MF.getSubtarget<X86Subtarget>();
   TII = STI->getInstrInfo();
   TRI = STI->getRegisterInfo();
   X86FI = MF.getInfo<X86MachineFunctionInfo>();
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -26073,8 +26073,7 @@
 
   // Return EntryEBP + ParentFrameOffset for x64. This adjusts from RSP after
   // prologue to RBP in the parent function.
-  const X86Subtarget &Subtarget =
-      static_cast<const X86Subtarget &>(DAG.getSubtarget());
+  const X86Subtarget &Subtarget = DAG.getSubtarget<X86Subtarget>();
   if (Subtarget.is64Bit())
     return DAG.getNode(ISD::ADD, dl, PtrVT, EntryEBP, ParentFrameOffset);
 
@@ -39107,8 +39106,7 @@
     SDValue N0 = V.getOperand(0);
     SDValue N1 = V.getOperand(1);
     unsigned Imm = V.getConstantOperandVal(2);
-    const X86Subtarget &Subtarget =
-        static_cast<const X86Subtarget &>(DAG.getSubtarget());
+    const X86Subtarget &Subtarget = DAG.getSubtarget<X86Subtarget>();
     if (!X86::mayFoldLoad(peekThroughOneUseBitcasts(N0), Subtarget) ||
         X86::mayFoldLoad(peekThroughOneUseBitcasts(N1), Subtarget))
       return SDValue();