Vendor import of clang release_39 branch r276489:vendor/clang/clang-release_39-r276489

https://llvm.org/svn/llvm-project/cfe/branches/release_39@276489

68 files changed, 14120 insertions, 4952 deletions

diff --git a/lib/CodeGen/ABIInfo.h b/lib/CodeGen/ABIInfo.h
index a65f27085616..530a7ef560c5 100644
--- a/lib/CodeGen/ABIInfo.h
+++ b/lib/CodeGen/ABIInfo.h
@@ -18,20 +18,25 @@ namespace llvm {
   class Value;
   class LLVMContext;
   class DataLayout;
+  class Type;
 }
 
 namespace clang {
   class ASTContext;
   class TargetInfo;
 
-  namespace CodeGen {
-    class ABIArgInfo;
-    class Address;
-    class CGCXXABI;
-    class CGFunctionInfo;
-    class CodeGenFunction;
-    class CodeGenTypes;
-  }
+namespace CodeGen {
+  class ABIArgInfo;
+  class Address;
+  class CGCXXABI;
+  class CGFunctionInfo;
+  class CodeGenFunction;
+  class CodeGenTypes;
+  class SwiftABIInfo;
+
+namespace swiftcall {
+  class SwiftAggLowering;
+}
 
   // FIXME: All of this stuff should be part of the target interface
   // somehow. It is currently here because it is not clear how to factor
@@ -55,6 +60,8 @@ namespace clang {
 
     virtual ~ABIInfo();
 
+    virtual bool supportsSwift() const { return false; }
+
     CodeGen::CGCXXABI &getCXXABI() const;
     ASTContext &getContext() const;
     llvm::LLVMContext &getVMContext() const;
@@ -85,6 +92,8 @@ namespace clang {
                                        CodeGen::Address VAListAddr,
                                        QualType Ty) const = 0;
 
+    bool isAndroid() const;
+
     /// Emit the target dependent code to load a value of
     /// \arg Ty from the \c __builtin_ms_va_list pointed to by \arg VAListAddr.
     virtual CodeGen::Address EmitMSVAArg(CodeGen::CodeGenFunction &CGF,
@@ -110,7 +119,35 @@ namespace clang {
 
     CodeGen::ABIArgInfo
     getNaturalAlignIndirectInReg(QualType Ty, bool Realign = false) const;
+
+
+  };
+
+  /// A refining implementation of ABIInfo for targets that support swiftcall.
+  ///
+  /// If we find ourselves wanting multiple such refinements, they'll probably
+  /// be independent refinements, and we should probably find another way
+  /// to do it than simple inheritance.
+  class SwiftABIInfo : public ABIInfo {
+  public:
+    SwiftABIInfo(CodeGen::CodeGenTypes &cgt) : ABIInfo(cgt) {}
+
+    bool supportsSwift() const final override { return true; }
+
+    virtual bool shouldPassIndirectlyForSwift(CharUnits totalSize,
+                                              ArrayRef<llvm::Type*> types,
+                                              bool asReturnValue) const = 0;
+
+    virtual bool isLegalVectorTypeForSwift(CharUnits totalSize,
+                                           llvm::Type *eltTy,
+                                           unsigned elts) const;
+
+    static bool classof(const ABIInfo *info) {
+      return info->supportsSwift();
+    }
   };
+
+}  // end namespace CodeGen
 }  // end namespace clang
 
 #endif
diff --git a/lib/CodeGen/BackendUtil.cpp b/lib/CodeGen/BackendUtil.cpp
index 6d746c25eed1..165b6dd55c9b 100644
--- a/lib/CodeGen/BackendUtil.cpp
+++ b/lib/CodeGen/BackendUtil.cpp
@@ -16,19 +16,21 @@
 #include "clang/Frontend/Utils.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/StringSwitch.h"
+#include "llvm/ADT/Triple.h"
 #include "llvm/Analysis/TargetLibraryInfo.h"
 #include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/Bitcode/BitcodeWriterPass.h"
+#include "llvm/Bitcode/ReaderWriter.h"
 #include "llvm/CodeGen/RegAllocRegistry.h"
 #include "llvm/CodeGen/SchedulerRegistry.h"
 #include "llvm/IR/DataLayout.h"
-#include "llvm/IR/FunctionInfo.h"
+#include "llvm/IR/ModuleSummaryIndex.h"
 #include "llvm/IR/IRPrintingPasses.h"
 #include "llvm/IR/LegacyPassManager.h"
 #include "llvm/IR/Module.h"
 #include "llvm/IR/Verifier.h"
 #include "llvm/MC/SubtargetFeature.h"
-#include "llvm/Object/FunctionIndexObjectFile.h"
+#include "llvm/Object/ModuleSummaryIndexObjectFile.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/PrettyStackTrace.h"
 #include "llvm/Support/TargetRegistry.h"
@@ -42,6 +44,7 @@
 #include "llvm/Transforms/Instrumentation.h"
 #include "llvm/Transforms/ObjCARC.h"
 #include "llvm/Transforms/Scalar.h"
+#include "llvm/Transforms/Scalar/GVN.h"
 #include "llvm/Transforms/Utils/SymbolRewriter.h"
 #include <memory>
 using namespace clang;
@@ -58,9 +61,7 @@ class EmitAssemblyHelper {
 
   Timer CodeGenerationTime;
 
-  mutable legacy::PassManager *CodeGenPasses;
-  mutable legacy::PassManager *PerModulePasses;
-  mutable legacy::FunctionPassManager *PerFunctionPasses;
+  std::unique_ptr<raw_pwrite_stream> OS;
 
 private:
   TargetIRAnalysis getTargetIRAnalysis() const {
@@ -70,70 +71,44 @@ private:
     return TargetIRAnalysis();
   }
 
-  legacy::PassManager *getCodeGenPasses() const {
-    if (!CodeGenPasses) {
-      CodeGenPasses = new legacy::PassManager();
-      CodeGenPasses->add(
-          createTargetTransformInfoWrapperPass(getTargetIRAnalysis()));
-    }
-    return CodeGenPasses;
-  }
-
-  legacy::PassManager *getPerModulePasses() const {
-    if (!PerModulePasses) {
-      PerModulePasses = new legacy::PassManager();
-      PerModulePasses->add(
-          createTargetTransformInfoWrapperPass(getTargetIRAnalysis()));
-    }
-    return PerModulePasses;
-  }
-
-  legacy::FunctionPassManager *getPerFunctionPasses() const {
-    if (!PerFunctionPasses) {
-      PerFunctionPasses = new legacy::FunctionPassManager(TheModule);
-      PerFunctionPasses->add(
-          createTargetTransformInfoWrapperPass(getTargetIRAnalysis()));
-    }
-    return PerFunctionPasses;
-  }
+  /// Set LLVM command line options passed through -backend-option.
+  void setCommandLineOpts();
 
-  void CreatePasses(FunctionInfoIndex *FunctionIndex);
+  void CreatePasses(legacy::PassManager &MPM, legacy::FunctionPassManager &FPM,
+                    ModuleSummaryIndex *ModuleSummary);
 
   /// Generates the TargetMachine.
-  /// Returns Null if it is unable to create the target machine.
+  /// Leaves TM unchanged if it is unable to create the target machine.
   /// Some of our clang tests specify triples which are not built
   /// into clang. This is okay because these tests check the generated
   /// IR, and they require DataLayout which depends on the triple.
   /// In this case, we allow this method to fail and not report an error.
   /// When MustCreateTM is used, we print an error if we are unable to load
   /// the requested target.
-  TargetMachine *CreateTargetMachine(bool MustCreateTM);
+  void CreateTargetMachine(bool MustCreateTM);
 
   /// Add passes necessary to emit assembly or LLVM IR.
   ///
   /// \return True on success.
-  bool AddEmitPasses(BackendAction Action, raw_pwrite_stream &OS);
+  bool AddEmitPasses(legacy::PassManager &CodeGenPasses, BackendAction Action,
+                     raw_pwrite_stream &OS);
 
 public:
   EmitAssemblyHelper(DiagnosticsEngine &_Diags, const CodeGenOptions &CGOpts,
                      const clang::TargetOptions &TOpts,
                      const LangOptions &LOpts, Module *M)
       : Diags(_Diags), CodeGenOpts(CGOpts), TargetOpts(TOpts), LangOpts(LOpts),
-        TheModule(M), CodeGenerationTime("Code Generation Time"),
-        CodeGenPasses(nullptr), PerModulePasses(nullptr),
-        PerFunctionPasses(nullptr) {}
+        TheModule(M), CodeGenerationTime("Code Generation Time") {}
 
   ~EmitAssemblyHelper() {
-    delete CodeGenPasses;
-    delete PerModulePasses;
-    delete PerFunctionPasses;
     if (CodeGenOpts.DisableFree)
       BuryPointer(std::move(TM));
   }
 
   std::unique_ptr<TargetMachine> TM;
 
-  void EmitAssembly(BackendAction Action, raw_pwrite_stream *OS);
+  void EmitAssembly(BackendAction Action,
+                    std::unique_ptr<raw_pwrite_stream> OS);
 };
 
 // We need this wrapper to access LangOpts and CGOpts from extension functions
@@ -172,8 +147,19 @@ static void addAddDiscriminatorsPass(const PassManagerBuilder &Builder,
   PM.add(createAddDiscriminatorsPass());
 }
 
+static void addCleanupPassesForSampleProfiler(
+    const PassManagerBuilder &Builder, legacy::PassManagerBase &PM) {
+  // instcombine is needed before sample profile annotation because it converts
+  // certain function calls to be inlinable. simplifycfg and sroa are needed
+  // before instcombine for necessary preparation. E.g. load store is eliminated
+  // properly so that instcombine will not introduce unecessary liverange.
+  PM.add(createCFGSimplificationPass());
+  PM.add(createSROAPass());
+  PM.add(createInstructionCombiningPass());
+}
+
 static void addBoundsCheckingPass(const PassManagerBuilder &Builder,
-                                    legacy::PassManagerBase &PM) {
+                                  legacy::PassManagerBase &PM) {
   PM.add(createBoundsCheckingPass());
 }
 
@@ -189,6 +175,7 @@ static void addSanitizerCoveragePass(const PassManagerBuilder &Builder,
   Opts.TraceBB = CGOpts.SanitizeCoverageTraceBB;
   Opts.TraceCmp = CGOpts.SanitizeCoverageTraceCmp;
   Opts.Use8bitCounters = CGOpts.SanitizeCoverage8bitCounters;
+  Opts.TracePC = CGOpts.SanitizeCoverageTracePC;
   PM.add(createSanitizerCoverageModulePass(Opts));
 }
 
@@ -198,14 +185,17 @@ static void addAddressSanitizerPasses(const PassManagerBuilder &Builder,
       static_cast<const PassManagerBuilderWrapper&>(Builder);
   const CodeGenOptions &CGOpts = BuilderWrapper.getCGOpts();
   bool Recover = CGOpts.SanitizeRecover.has(SanitizerKind::Address);
-  PM.add(createAddressSanitizerFunctionPass(/*CompileKernel*/false, Recover));
+  bool UseAfterScope = CGOpts.SanitizeAddressUseAfterScope;
+  PM.add(createAddressSanitizerFunctionPass(/*CompileKernel*/ false, Recover,
+                                            UseAfterScope));
   PM.add(createAddressSanitizerModulePass(/*CompileKernel*/false, Recover));
 }
 
 static void addKernelAddressSanitizerPasses(const PassManagerBuilder &Builder,
                                             legacy::PassManagerBase &PM) {
-  PM.add(createAddressSanitizerFunctionPass(/*CompileKernel*/true,
-                                            /*Recover*/true));
+  PM.add(createAddressSanitizerFunctionPass(
+      /*CompileKernel*/ true,
+      /*Recover*/ true, /*UseAfterScope*/ false));
   PM.add(createAddressSanitizerModulePass(/*CompileKernel*/true,
                                           /*Recover*/true));
 }
@@ -243,6 +233,19 @@ static void addDataFlowSanitizerPass(const PassManagerBuilder &Builder,
   PM.add(createDataFlowSanitizerPass(LangOpts.SanitizerBlacklistFiles));
 }
 
+static void addEfficiencySanitizerPass(const PassManagerBuilder &Builder,
+                                       legacy::PassManagerBase &PM) {
+  const PassManagerBuilderWrapper &BuilderWrapper =
+      static_cast<const PassManagerBuilderWrapper&>(Builder);
+  const LangOptions &LangOpts = BuilderWrapper.getLangOpts();
+  EfficiencySanitizerOptions Opts;
+  if (LangOpts.Sanitize.has(SanitizerKind::EfficiencyCacheFrag))
+    Opts.ToolType = EfficiencySanitizerOptions::ESAN_CacheFrag;
+  else if (LangOpts.Sanitize.has(SanitizerKind::EfficiencyWorkingSet))
+    Opts.ToolType = EfficiencySanitizerOptions::ESAN_WorkingSet;
+  PM.add(createEfficiencySanitizerPass(Opts));
+}
+
 static TargetLibraryInfoImpl *createTLII(llvm::Triple &TargetTriple,
                                          const CodeGenOptions &CodeGenOpts) {
   TargetLibraryInfoImpl *TLII = new TargetLibraryInfoImpl(TargetTriple);
@@ -277,7 +280,9 @@ static void addSymbolRewriterPass(const CodeGenOptions &Opts,
   MPM->add(createRewriteSymbolsPass(DL));
 }
 
-void EmitAssemblyHelper::CreatePasses(FunctionInfoIndex *FunctionIndex) {
+void EmitAssemblyHelper::CreatePasses(legacy::PassManager &MPM,
+                                      legacy::FunctionPassManager &FPM,
+                                      ModuleSummaryIndex *ModuleSummary) {
   if (CodeGenOpts.DisableLLVMPasses)
     return;
 
@@ -300,7 +305,8 @@ void EmitAssemblyHelper::CreatePasses(FunctionInfoIndex *FunctionIndex) {
   switch (Inlining) {
   case CodeGenOptions::NoInlining:
     break;
-  case CodeGenOptions::NormalInlining: {
+  case CodeGenOptions::NormalInlining:
+  case CodeGenOptions::OnlyHintInlining: {
     PMBuilder.Inliner =
         createFunctionInliningPass(OptLevel, CodeGenOpts.OptimizeSize);
     break;
@@ -321,22 +327,28 @@ void EmitAssemblyHelper::CreatePasses(FunctionInfoIndex *FunctionIndex) {
   PMBuilder.SLPVectorize = CodeGenOpts.VectorizeSLP;
   PMBuilder.LoopVectorize = CodeGenOpts.VectorizeLoop;
 
-  PMBuilder.DisableUnitAtATime = !CodeGenOpts.UnitAtATime;
   PMBuilder.DisableUnrollLoops = !CodeGenOpts.UnrollLoops;
   PMBuilder.MergeFunctions = CodeGenOpts.MergeFunctions;
+  PMBuilder.PrepareForThinLTO = CodeGenOpts.EmitSummaryIndex;
   PMBuilder.PrepareForLTO = CodeGenOpts.PrepareForLTO;
   PMBuilder.RerollLoops = CodeGenOpts.RerollLoops;
 
-  legacy::PassManager *MPM = getPerModulePasses();
-
   // If we are performing a ThinLTO importing compile, invoke the LTO
-  // pipeline and pass down the in-memory function index.
-  if (FunctionIndex) {
-    PMBuilder.FunctionIndex = FunctionIndex;
-    PMBuilder.populateLTOPassManager(*MPM);
+  // pipeline and pass down the in-memory module summary index.
+  if (ModuleSummary) {
+    PMBuilder.ModuleSummary = ModuleSummary;
+    PMBuilder.populateThinLTOPassManager(MPM);
     return;
   }
 
+  // Add target-specific passes that need to run as early as possible.
+  if (TM)
+    PMBuilder.addExtension(
+        PassManagerBuilder::EP_EarlyAsPossible,
+        [&](const PassManagerBuilder &, legacy::PassManagerBase &PM) {
+          TM->addEarlyAsPossiblePasses(PM);
+        });
+
   PMBuilder.addExtension(PassManagerBuilder::EP_EarlyAsPossible,
                          addAddDiscriminatorsPass);
 
@@ -401,15 +413,20 @@ void EmitAssemblyHelper::CreatePasses(FunctionInfoIndex *FunctionIndex) {
                            addDataFlowSanitizerPass);
   }
 
+  if (LangOpts.Sanitize.hasOneOf(SanitizerKind::Efficiency)) {
+    PMBuilder.addExtension(PassManagerBuilder::EP_OptimizerLast,
+                           addEfficiencySanitizerPass);
+    PMBuilder.addExtension(PassManagerBuilder::EP_EnabledOnOptLevel0,
+                           addEfficiencySanitizerPass);
+  }
+
   // Set up the per-function pass manager.
-  legacy::FunctionPassManager *FPM = getPerFunctionPasses();
   if (CodeGenOpts.VerifyModule)
-    FPM->add(createVerifierPass());
-  PMBuilder.populateFunctionPassManager(*FPM);
+    FPM.add(createVerifierPass());
 
   // Set up the per-module pass manager.
   if (!CodeGenOpts.RewriteMapFiles.empty())
-    addSymbolRewriterPass(CodeGenOpts, MPM);
+    addSymbolRewriterPass(CodeGenOpts, &MPM);
 
   if (!CodeGenOpts.DisableGCov &&
       (CodeGenOpts.EmitGcovArcs || CodeGenOpts.EmitGcovNotes)) {
@@ -424,25 +441,56 @@ void EmitAssemblyHelper::CreatePasses(FunctionInfoIndex *FunctionIndex) {
     Options.FunctionNamesInData =
         !CodeGenOpts.CoverageNoFunctionNamesInData;
     Options.ExitBlockBeforeBody = CodeGenOpts.CoverageExitBlockBeforeBody;
-    MPM->add(createGCOVProfilerPass(Options));
-    if (CodeGenOpts.getDebugInfo() == CodeGenOptions::NoDebugInfo)
-      MPM->add(createStripSymbolsPass(true));
+    MPM.add(createGCOVProfilerPass(Options));
+    if (CodeGenOpts.getDebugInfo() == codegenoptions::NoDebugInfo)
+      MPM.add(createStripSymbolsPass(true));
   }
 
-  if (CodeGenOpts.ProfileInstrGenerate) {
+  if (CodeGenOpts.hasProfileClangInstr()) {
     InstrProfOptions Options;
     Options.NoRedZone = CodeGenOpts.DisableRedZone;
     Options.InstrProfileOutput = CodeGenOpts.InstrProfileOutput;
-    MPM->add(createInstrProfilingPass(Options));
+    MPM.add(createInstrProfilingLegacyPass(Options));
+  }
+  if (CodeGenOpts.hasProfileIRInstr()) {
+    if (!CodeGenOpts.InstrProfileOutput.empty())
+      PMBuilder.PGOInstrGen = CodeGenOpts.InstrProfileOutput;
+    else
+      PMBuilder.PGOInstrGen = "default.profraw";
+  }
+  if (CodeGenOpts.hasProfileIRUse())
+    PMBuilder.PGOInstrUse = CodeGenOpts.ProfileInstrumentUsePath;
+
+  if (!CodeGenOpts.SampleProfileFile.empty()) {
+    MPM.add(createPruneEHPass());
+    MPM.add(createSampleProfileLoaderPass(CodeGenOpts.SampleProfileFile));
+    PMBuilder.addExtension(PassManagerBuilder::EP_EarlyAsPossible,
+                           addCleanupPassesForSampleProfiler);
   }
 
-  if (!CodeGenOpts.SampleProfileFile.empty())
-    MPM->add(createSampleProfileLoaderPass(CodeGenOpts.SampleProfileFile));
+  PMBuilder.populateFunctionPassManager(FPM);
+  PMBuilder.populateModulePassManager(MPM);
+}
 
-  PMBuilder.populateModulePassManager(*MPM);
+void EmitAssemblyHelper::setCommandLineOpts() {
+  SmallVector<const char *, 16> BackendArgs;
+  BackendArgs.push_back("clang"); // Fake program name.
+  if (!CodeGenOpts.DebugPass.empty()) {
+    BackendArgs.push_back("-debug-pass");
+    BackendArgs.push_back(CodeGenOpts.DebugPass.c_str());
+  }
+  if (!CodeGenOpts.LimitFloatPrecision.empty()) {
+    BackendArgs.push_back("-limit-float-precision");
+    BackendArgs.push_back(CodeGenOpts.LimitFloatPrecision.c_str());
+  }
+  for (const std::string &BackendOption : CodeGenOpts.BackendOptions)
+    BackendArgs.push_back(BackendOption.c_str());
+  BackendArgs.push_back(nullptr);
+  llvm::cl::ParseCommandLineOptions(BackendArgs.size() - 1,
+                                    BackendArgs.data());
 }
 
-TargetMachine *EmitAssemblyHelper::CreateTargetMachine(bool MustCreateTM) {
+void EmitAssemblyHelper::CreateTargetMachine(bool MustCreateTM) {
   // Create the TargetMachine for generating code.
   std::string Error;
   std::string Triple = TheModule->getTargetTriple();
@@ -450,7 +498,7 @@ TargetMachine *EmitAssemblyHelper::CreateTargetMachine(bool MustCreateTM) {
   if (!TheTarget) {
     if (MustCreateTM)
       Diags.Report(diag::err_fe_unable_to_create_target) << Error;
-    return nullptr;
+    return;
   }
 
   unsigned CodeModel =
@@ -464,27 +512,11 @@ TargetMachine *EmitAssemblyHelper::CreateTargetMachine(bool MustCreateTM) {
   assert(CodeModel != ~0u && "invalid code model!");
   llvm::CodeModel::Model CM = static_cast<llvm::CodeModel::Model>(CodeModel);
 
-  SmallVector<const char *, 16> BackendArgs;
-  BackendArgs.push_back("clang"); // Fake program name.
-  if (!CodeGenOpts.DebugPass.empty()) {
-    BackendArgs.push_back("-debug-pass");
-    BackendArgs.push_back(CodeGenOpts.DebugPass.c_str());
-  }
-  if (!CodeGenOpts.LimitFloatPrecision.empty()) {
-    BackendArgs.push_back("-limit-float-precision");
-    BackendArgs.push_back(CodeGenOpts.LimitFloatPrecision.c_str());
-  }
-  for (const std::string &BackendOption : CodeGenOpts.BackendOptions)
-    BackendArgs.push_back(BackendOption.c_str());
-  BackendArgs.push_back(nullptr);
-  llvm::cl::ParseCommandLineOptions(BackendArgs.size() - 1,
-                                    BackendArgs.data());
-
   std::string FeaturesStr =
       llvm::join(TargetOpts.Features.begin(), TargetOpts.Features.end(), ",");
 
   // Keep this synced with the equivalent code in tools/driver/cc1as_main.cpp.
-  llvm::Reloc::Model RM = llvm::Reloc::Default;
+  llvm::Optional<llvm::Reloc::Model> RM;
   if (CodeGenOpts.RelocationModel == "static") {
     RM = llvm::Reloc::Static;
   } else if (CodeGenOpts.RelocationModel == "pic") {
@@ -539,38 +571,29 @@ TargetMachine *EmitAssemblyHelper::CreateTargetMachine(bool MustCreateTM) {
   Options.UseInitArray = CodeGenOpts.UseInitArray;
   Options.DisableIntegratedAS = CodeGenOpts.DisableIntegratedAS;
   Options.CompressDebugSections = CodeGenOpts.CompressDebugSections;
+  Options.RelaxELFRelocations = CodeGenOpts.RelaxELFRelocations;
 
   // Set EABI version.
-  Options.EABIVersion = llvm::StringSwitch<llvm::EABI>(CodeGenOpts.EABIVersion)
+  Options.EABIVersion = llvm::StringSwitch<llvm::EABI>(TargetOpts.EABIVersion)
                             .Case("4", llvm::EABI::EABI4)
                             .Case("5", llvm::EABI::EABI5)
                             .Case("gnu", llvm::EABI::GNU)
                             .Default(llvm::EABI::Default);
 
+  if (LangOpts.SjLjExceptions)
+    Options.ExceptionModel = llvm::ExceptionHandling::SjLj;
+
   Options.LessPreciseFPMADOption = CodeGenOpts.LessPreciseFPMAD;
   Options.NoInfsFPMath = CodeGenOpts.NoInfsFPMath;
   Options.NoNaNsFPMath = CodeGenOpts.NoNaNsFPMath;
   Options.NoZerosInBSS = CodeGenOpts.NoZeroInitializedInBSS;
   Options.UnsafeFPMath = CodeGenOpts.UnsafeFPMath;
   Options.StackAlignmentOverride = CodeGenOpts.StackAlignment;
-  Options.PositionIndependentExecutable = LangOpts.PIELevel != 0;
   Options.FunctionSections = CodeGenOpts.FunctionSections;
   Options.DataSections = CodeGenOpts.DataSections;
   Options.UniqueSectionNames = CodeGenOpts.UniqueSectionNames;
   Options.EmulatedTLS = CodeGenOpts.EmulatedTLS;
-  switch (CodeGenOpts.getDebuggerTuning()) {
-  case CodeGenOptions::DebuggerKindGDB:
-    Options.DebuggerTuning = llvm::DebuggerKind::GDB;
-    break;
-  case CodeGenOptions::DebuggerKindLLDB:
-    Options.DebuggerTuning = llvm::DebuggerKind::LLDB;
-    break;
-  case CodeGenOptions::DebuggerKindSCE:
-    Options.DebuggerTuning = llvm::DebuggerKind::SCE;
-    break;
-  default:
-    break;
-  }
+  Options.DebuggerTuning = CodeGenOpts.getDebuggerTuning();
 
   Options.MCOptions.MCRelaxAll = CodeGenOpts.RelaxAll;
   Options.MCOptions.MCSaveTempLabels = CodeGenOpts.SaveTempLabels;
@@ -582,24 +605,18 @@ TargetMachine *EmitAssemblyHelper::CreateTargetMachine(bool MustCreateTM) {
   Options.MCOptions.AsmVerbose = CodeGenOpts.AsmVerbose;
   Options.MCOptions.ABIName = TargetOpts.ABI;
 
-  TargetMachine *TM = TheTarget->createTargetMachine(Triple, TargetOpts.CPU,
-                                                     FeaturesStr, Options,
-                                                     RM, CM, OptLevel);
-
-  return TM;
+  TM.reset(TheTarget->createTargetMachine(Triple, TargetOpts.CPU, FeaturesStr,
+                                          Options, RM, CM, OptLevel));
 }
 
-bool EmitAssemblyHelper::AddEmitPasses(BackendAction Action,
+bool EmitAssemblyHelper::AddEmitPasses(legacy::PassManager &CodeGenPasses,
+                                       BackendAction Action,
                                        raw_pwrite_stream &OS) {
-
-  // Create the code generator passes.
-  legacy::PassManager *PM = getCodeGenPasses();
-
   // Add LibraryInfo.
   llvm::Triple TargetTriple(TheModule->getTargetTriple());
   std::unique_ptr<TargetLibraryInfoImpl> TLII(
       createTLII(TargetTriple, CodeGenOpts));
-  PM->add(new TargetLibraryInfoWrapperPass(*TLII));
+  CodeGenPasses.add(new TargetLibraryInfoWrapperPass(*TLII));
 
   // Normal mode, emit a .s or .o file by running the code generator. Note,
   // this also adds codegenerator level optimization passes.
@@ -615,9 +632,9 @@ bool EmitAssemblyHelper::AddEmitPasses(BackendAction Action,
   // "codegen" passes so that it isn't run multiple times when there is
   // inlining happening.
   if (CodeGenOpts.OptimizationLevel > 0)
-    PM->add(createObjCARCContractPass());
+    CodeGenPasses.add(createObjCARCContractPass());
 
-  if (TM->addPassesToEmitFile(*PM, OS, CGFT,
+  if (TM->addPassesToEmitFile(CodeGenPasses, OS, CGFT,
                               /*DisableVerify=*/!CodeGenOpts.VerifyModule)) {
     Diags.Report(diag::err_fe_unable_to_interface_with_target);
     return false;
@@ -627,14 +644,15 @@ bool EmitAssemblyHelper::AddEmitPasses(BackendAction Action,
 }
 
 void EmitAssemblyHelper::EmitAssembly(BackendAction Action,
-                                      raw_pwrite_stream *OS) {
+                                      std::unique_ptr<raw_pwrite_stream> OS) {
   TimeRegion Region(llvm::TimePassesIsEnabled ? &CodeGenerationTime : nullptr);
 
+  setCommandLineOpts();
+
   bool UsesCodeGen = (Action != Backend_EmitNothing &&
                       Action != Backend_EmitBC &&
                       Action != Backend_EmitLL);
-  if (!TM)
-    TM.reset(CreateTargetMachine(UsesCodeGen));
+  CreateTargetMachine(UsesCodeGen);
 
   if (UsesCodeGen && !TM)
     return;
@@ -644,41 +662,54 @@ void EmitAssemblyHelper::EmitAssembly(BackendAction Action,
   // If we are performing a ThinLTO importing compile, load the function
   // index into memory and pass it into CreatePasses, which will add it
   // to the PassManagerBuilder and invoke LTO passes.
-  std::unique_ptr<FunctionInfoIndex> FunctionIndex;
+  std::unique_ptr<ModuleSummaryIndex> ModuleSummary;
   if (!CodeGenOpts.ThinLTOIndexFile.empty()) {
-    ErrorOr<std::unique_ptr<FunctionInfoIndex>> IndexOrErr =
-        llvm::getFunctionIndexForFile(CodeGenOpts.ThinLTOIndexFile,
-                                      [&](const DiagnosticInfo &DI) {
-                                        TheModule->getContext().diagnose(DI);
-                                      });
+    ErrorOr<std::unique_ptr<ModuleSummaryIndex>> IndexOrErr =
+        llvm::getModuleSummaryIndexForFile(
+            CodeGenOpts.ThinLTOIndexFile, [&](const DiagnosticInfo &DI) {
+              TheModule->getContext().diagnose(DI);
+            });
     if (std::error_code EC = IndexOrErr.getError()) {
       std::string Error = EC.message();
       errs() << "Error loading index file '" << CodeGenOpts.ThinLTOIndexFile
              << "': " << Error << "\n";
       return;
     }
-    FunctionIndex = std::move(IndexOrErr.get());
-    assert(FunctionIndex && "Expected non-empty function index");
+    ModuleSummary = std::move(IndexOrErr.get());
+    assert(ModuleSummary && "Expected non-empty module summary index");
   }
 
-  CreatePasses(FunctionIndex.get());
+  legacy::PassManager PerModulePasses;
+  PerModulePasses.add(
+      createTargetTransformInfoWrapperPass(getTargetIRAnalysis()));
+
+  legacy::FunctionPassManager PerFunctionPasses(TheModule);
+  PerFunctionPasses.add(
+      createTargetTransformInfoWrapperPass(getTargetIRAnalysis()));
+
+  CreatePasses(PerModulePasses, PerFunctionPasses, ModuleSummary.get());
+
+  legacy::PassManager CodeGenPasses;
+  CodeGenPasses.add(
+      createTargetTransformInfoWrapperPass(getTargetIRAnalysis()));
 
   switch (Action) {
   case Backend_EmitNothing:
     break;
 
   case Backend_EmitBC:
-    getPerModulePasses()->add(createBitcodeWriterPass(
-        *OS, CodeGenOpts.EmitLLVMUseLists, CodeGenOpts.EmitFunctionSummary));
+    PerModulePasses.add(createBitcodeWriterPass(
+        *OS, CodeGenOpts.EmitLLVMUseLists, CodeGenOpts.EmitSummaryIndex,
+        CodeGenOpts.EmitSummaryIndex));
     break;
 
   case Backend_EmitLL:
-    getPerModulePasses()->add(
+    PerModulePasses.add(
         createPrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists));
     break;
 
   default:
-    if (!AddEmitPasses(Action, *OS))
+    if (!AddEmitPasses(CodeGenPasses, Action, *OS))
       return;
   }
 
@@ -688,46 +719,165 @@ void EmitAssemblyHelper::EmitAssembly(BackendAction Action,
   // Run passes. For now we do all passes at once, but eventually we
   // would like to have the option of streaming code generation.
 
-  if (PerFunctionPasses) {
+  {
     PrettyStackTraceString CrashInfo("Per-function optimization");
 
-    PerFunctionPasses->doInitialization();
+    PerFunctionPasses.doInitialization();
     for (Function &F : *TheModule)
       if (!F.isDeclaration())
-        PerFunctionPasses->run(F);
-    PerFunctionPasses->doFinalization();
+        PerFunctionPasses.run(F);
+    PerFunctionPasses.doFinalization();
   }
 
-  if (PerModulePasses) {
+  {
     PrettyStackTraceString CrashInfo("Per-module optimization passes");
-    PerModulePasses->run(*TheModule);
+    PerModulePasses.run(*TheModule);
   }
 
-  if (CodeGenPasses) {
+  {
     PrettyStackTraceString CrashInfo("Code generation");
-    CodeGenPasses->run(*TheModule);
+    CodeGenPasses.run(*TheModule);
   }
 }
 
 void clang::EmitBackendOutput(DiagnosticsEngine &Diags,
                               const CodeGenOptions &CGOpts,
                               const clang::TargetOptions &TOpts,
-                              const LangOptions &LOpts, StringRef TDesc,
+                              const LangOptions &LOpts, const llvm::DataLayout &TDesc,
                               Module *M, BackendAction Action,
-                              raw_pwrite_stream *OS) {
+                              std::unique_ptr<raw_pwrite_stream> OS) {
   EmitAssemblyHelper AsmHelper(Diags, CGOpts, TOpts, LOpts, M);
 
-  AsmHelper.EmitAssembly(Action, OS);
+  AsmHelper.EmitAssembly(Action, std::move(OS));
 
-  // If an optional clang TargetInfo description string was passed in, use it to
-  // verify the LLVM TargetMachine's DataLayout.
-  if (AsmHelper.TM && !TDesc.empty()) {
+  // Verify clang's TargetInfo DataLayout against the LLVM TargetMachine's
+  // DataLayout.
+  if (AsmHelper.TM) {
     std::string DLDesc = M->getDataLayout().getStringRepresentation();
-    if (DLDesc != TDesc) {
+    if (DLDesc != TDesc.getStringRepresentation()) {
       unsigned DiagID = Diags.getCustomDiagID(
           DiagnosticsEngine::Error, "backend data layout '%0' does not match "
                                     "expected target description '%1'");
-      Diags.Report(DiagID) << DLDesc << TDesc;
+      Diags.Report(DiagID) << DLDesc << TDesc.getStringRepresentation();
+    }
+  }
+}
+
+static const char* getSectionNameForBitcode(const Triple &T) {
+  switch (T.getObjectFormat()) {
+  case Triple::MachO:
+    return "__LLVM,__bitcode";
+  case Triple::COFF:
+  case Triple::ELF:
+  case Triple::UnknownObjectFormat:
+    return ".llvmbc";
+  }
+  llvm_unreachable("Unimplemented ObjectFormatType");
+}
+
+static const char* getSectionNameForCommandline(const Triple &T) {
+  switch (T.getObjectFormat()) {
+  case Triple::MachO:
+    return "__LLVM,__cmdline";
+  case Triple::COFF:
+  case Triple::ELF:
+  case Triple::UnknownObjectFormat:
+    return ".llvmcmd";
+  }
+  llvm_unreachable("Unimplemented ObjectFormatType");
+}
+
+// With -fembed-bitcode, save a copy of the llvm IR as data in the
+// __LLVM,__bitcode section.
+void clang::EmbedBitcode(llvm::Module *M, const CodeGenOptions &CGOpts,
+                         llvm::MemoryBufferRef Buf) {
+  if (CGOpts.getEmbedBitcode() == CodeGenOptions::Embed_Off)
+    return;
+
+  // Save llvm.compiler.used and remote it.
+  SmallVector<Constant*, 2> UsedArray;
+  SmallSet<GlobalValue*, 4> UsedGlobals;
+  Type *UsedElementType = Type::getInt8Ty(M->getContext())->getPointerTo(0);
+  GlobalVariable *Used = collectUsedGlobalVariables(*M, UsedGlobals, true);
+  for (auto *GV : UsedGlobals) {
+    if (GV->getName() != "llvm.embedded.module" &&
+        GV->getName() != "llvm.cmdline")
+      UsedArray.push_back(
+          ConstantExpr::getPointerBitCastOrAddrSpaceCast(GV, UsedElementType));
+  }
+  if (Used)
+    Used->eraseFromParent();
+
+  // Embed the bitcode for the llvm module.
+  std::string Data;
+  ArrayRef<uint8_t> ModuleData;
+  Triple T(M->getTargetTriple());
+  // Create a constant that contains the bitcode.
+  // In case of embedding a marker, ignore the input Buf and use the empty
+  // ArrayRef. It is also legal to create a bitcode marker even Buf is empty.
+  if (CGOpts.getEmbedBitcode() != CodeGenOptions::Embed_Marker) {
+    if (!isBitcode((const unsigned char *)Buf.getBufferStart(),
+                   (const unsigned char *)Buf.getBufferEnd())) {
+      // If the input is LLVM Assembly, bitcode is produced by serializing
+      // the module. Use-lists order need to be perserved in this case.
+      llvm::raw_string_ostream OS(Data);
+      llvm::WriteBitcodeToFile(M, OS, /* ShouldPreserveUseListOrder */ true);
+      ModuleData =
+          ArrayRef<uint8_t>((const uint8_t *)OS.str().data(), OS.str().size());
+    } else
+      // If the input is LLVM bitcode, write the input byte stream directly.
+      ModuleData = ArrayRef<uint8_t>((const uint8_t *)Buf.getBufferStart(),
+                                     Buf.getBufferSize());
+  }
+  llvm::Constant *ModuleConstant =
+      llvm::ConstantDataArray::get(M->getContext(), ModuleData);
+  llvm::GlobalVariable *GV = new llvm::GlobalVariable(
+      *M, ModuleConstant->getType(), true, llvm::GlobalValue::PrivateLinkage,
+      ModuleConstant);
+  GV->setSection(getSectionNameForBitcode(T));
+  UsedArray.push_back(
+      ConstantExpr::getPointerBitCastOrAddrSpaceCast(GV, UsedElementType));
+  if (llvm::GlobalVariable *Old =
+          M->getGlobalVariable("llvm.embedded.module", true)) {
+    assert(Old->hasOneUse() &&
+           "llvm.embedded.module can only be used once in llvm.compiler.used");
+    GV->takeName(Old);
+    Old->eraseFromParent();
+  } else {
+    GV->setName("llvm.embedded.module");
+  }
+
+  // Skip if only bitcode needs to be embedded.
+  if (CGOpts.getEmbedBitcode() != CodeGenOptions::Embed_Bitcode) {
+    // Embed command-line options.
+    ArrayRef<uint8_t> CmdData(const_cast<uint8_t *>(CGOpts.CmdArgs.data()),
+                              CGOpts.CmdArgs.size());
+    llvm::Constant *CmdConstant =
+      llvm::ConstantDataArray::get(M->getContext(), CmdData);
+    GV = new llvm::GlobalVariable(*M, CmdConstant->getType(), true,
+                                  llvm::GlobalValue::PrivateLinkage,
+                                  CmdConstant);
+    GV->setSection(getSectionNameForCommandline(T));
+    UsedArray.push_back(
+        ConstantExpr::getPointerBitCastOrAddrSpaceCast(GV, UsedElementType));
+    if (llvm::GlobalVariable *Old =
+            M->getGlobalVariable("llvm.cmdline", true)) {
+      assert(Old->hasOneUse() &&
+             "llvm.cmdline can only be used once in llvm.compiler.used");
+      GV->takeName(Old);
+      Old->eraseFromParent();
+    } else {
+      GV->setName("llvm.cmdline");
     }
   }
+
+  if (UsedArray.empty())
+    return;
+
+  // Recreate llvm.compiler.used.
+  ArrayType *ATy = ArrayType::get(UsedElementType, UsedArray.size());
+  auto *NewUsed = new GlobalVariable(
+      *M, ATy, false, llvm::GlobalValue::AppendingLinkage,
+      llvm::ConstantArray::get(ATy, UsedArray), "llvm.compiler.used");
+  NewUsed->setSection("llvm.metadata");
 }
diff --git a/lib/CodeGen/CGAtomic.cpp b/lib/CodeGen/CGAtomic.cpp
index 24de30b0b862..7b747c138303 100644
--- a/lib/CodeGen/CGAtomic.cpp
+++ b/lib/CodeGen/CGAtomic.cpp
@@ -79,7 +79,7 @@ namespace {
         auto Offset = OrigBFI.Offset % C.toBits(lvalue.getAlignment());
         AtomicSizeInBits = C.toBits(
             C.toCharUnitsFromBits(Offset + OrigBFI.Size + C.getCharWidth() - 1)
-                .RoundUpToAlignment(lvalue.getAlignment()));
+                .alignTo(lvalue.getAlignment()));
         auto VoidPtrAddr = CGF.EmitCastToVoidPtr(lvalue.getBitFieldPointer());
         auto OffsetInChars =
             (C.toCharUnitsFromBits(OrigBFI.Offset) / lvalue.getAlignment()) *
@@ -221,11 +221,13 @@ namespace {
     /// \param IsWeak true if atomic operation is weak, false otherwise.
     /// \returns Pair of values: previous value from storage (value type) and
     /// boolean flag (i1 type) with true if success and false otherwise.
-    std::pair<RValue, llvm::Value *> EmitAtomicCompareExchange(
-        RValue Expected, RValue Desired,
-        llvm::AtomicOrdering Success = llvm::SequentiallyConsistent,
-        llvm::AtomicOrdering Failure = llvm::SequentiallyConsistent,
-        bool IsWeak = false);
+    std::pair<RValue, llvm::Value *>
+    EmitAtomicCompareExchange(RValue Expected, RValue Desired,
+                              llvm::AtomicOrdering Success =
+                                  llvm::AtomicOrdering::SequentiallyConsistent,
+                              llvm::AtomicOrdering Failure =
+                                  llvm::AtomicOrdering::SequentiallyConsistent,
+                              bool IsWeak = false);
 
     /// \brief Emits atomic update.
     /// \param AO Atomic ordering.
@@ -241,11 +243,6 @@ namespace {
     /// Materialize an atomic r-value in atomic-layout memory.
     Address materializeRValue(RValue rvalue) const;
 
-    /// \brief Translates LLVM atomic ordering to GNU atomic ordering for
-    /// libcalls.
-    static AtomicExpr::AtomicOrderingKind
-    translateAtomicOrdering(const llvm::AtomicOrdering AO);
-
     /// \brief Creates temp alloca for intermediate operations on atomic value.
     Address CreateTempAlloca() const;
   private:
@@ -260,13 +257,17 @@ namespace {
     /// \brief Emits atomic compare-and-exchange op as a libcall.
     llvm::Value *EmitAtomicCompareExchangeLibcall(
         llvm::Value *ExpectedAddr, llvm::Value *DesiredAddr,
-        llvm::AtomicOrdering Success = llvm::SequentiallyConsistent,
-        llvm::AtomicOrdering Failure = llvm::SequentiallyConsistent);
+        llvm::AtomicOrdering Success =
+            llvm::AtomicOrdering::SequentiallyConsistent,
+        llvm::AtomicOrdering Failure =
+            llvm::AtomicOrdering::SequentiallyConsistent);
     /// \brief Emits atomic compare-and-exchange op as LLVM instruction.
     std::pair<llvm::Value *, llvm::Value *> EmitAtomicCompareExchangeOp(
         llvm::Value *ExpectedVal, llvm::Value *DesiredVal,
-        llvm::AtomicOrdering Success = llvm::SequentiallyConsistent,
-        llvm::AtomicOrdering Failure = llvm::SequentiallyConsistent,
+        llvm::AtomicOrdering Success =
+            llvm::AtomicOrdering::SequentiallyConsistent,
+        llvm::AtomicOrdering Failure =
+            llvm::AtomicOrdering::SequentiallyConsistent,
         bool IsWeak = false);
     /// \brief Emit atomic update as libcalls.
     void
@@ -286,25 +287,6 @@ namespace {
   };
 }
 
-AtomicExpr::AtomicOrderingKind
-AtomicInfo::translateAtomicOrdering(const llvm::AtomicOrdering AO) {
-  switch (AO) {
-  case llvm::Unordered:
-  case llvm::NotAtomic:
-  case llvm::Monotonic:
-    return AtomicExpr::AO_ABI_memory_order_relaxed;
-  case llvm::Acquire:
-    return AtomicExpr::AO_ABI_memory_order_acquire;
-  case llvm::Release:
-    return AtomicExpr::AO_ABI_memory_order_release;
-  case llvm::AcquireRelease:
-    return AtomicExpr::AO_ABI_memory_order_acq_rel;
-  case llvm::SequentiallyConsistent:
-    return AtomicExpr::AO_ABI_memory_order_seq_cst;
-  }
-  llvm_unreachable("Unhandled AtomicOrdering");
-}
-
 Address AtomicInfo::CreateTempAlloca() const {
   Address TempAlloca = CGF.CreateMemTemp(
       (LVal.isBitField() && ValueSizeInBits > AtomicSizeInBits) ? ValueTy
@@ -323,8 +305,7 @@ static RValue emitAtomicLibcall(CodeGenFunction &CGF,
                                 QualType resultType,
                                 CallArgList &args) {
   const CGFunctionInfo &fnInfo =
-    CGF.CGM.getTypes().arrangeFreeFunctionCall(resultType, args,
-            FunctionType::ExtInfo(), RequiredArgs::All);
+    CGF.CGM.getTypes().arrangeBuiltinFunctionCall(resultType, args);
   llvm::FunctionType *fnTy = CGF.CGM.getTypes().GetFunctionType(fnInfo);
   llvm::Constant *fn = CGF.CGM.CreateRuntimeFunction(fnTy, fnName);
   return CGF.EmitCall(fnInfo, fn, ReturnValueSlot(), args);
@@ -422,33 +403,39 @@ static void emitAtomicCmpXchg(CodeGenFunction &CGF, AtomicExpr *E, bool IsWeak,
 /// instructions to cope with the provided (but possibly only dynamically known)
 /// FailureOrder.
 static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
-                                        bool IsWeak, Address Dest,
-                                        Address Ptr, Address Val1,
-                                        Address Val2,
+                                        bool IsWeak, Address Dest, Address Ptr,
+                                        Address Val1, Address Val2,
                                         llvm::Value *FailureOrderVal,
                                         uint64_t Size,
                                         llvm::AtomicOrdering SuccessOrder) {
   llvm::AtomicOrdering FailureOrder;
   if (llvm::ConstantInt *FO = dyn_cast<llvm::ConstantInt>(FailureOrderVal)) {
-    switch (FO->getSExtValue()) {
-    default:
-      FailureOrder = llvm::Monotonic;
-      break;
-    case AtomicExpr::AO_ABI_memory_order_consume:
-    case AtomicExpr::AO_ABI_memory_order_acquire:
-      FailureOrder = llvm::Acquire;
-      break;
-    case AtomicExpr::AO_ABI_memory_order_seq_cst:
-      FailureOrder = llvm::SequentiallyConsistent;
-      break;
-    }
-    if (FailureOrder >= SuccessOrder) {
-      // Don't assert on undefined behaviour.
+    auto FOS = FO->getSExtValue();
+    if (!llvm::isValidAtomicOrderingCABI(FOS))
+      FailureOrder = llvm::AtomicOrdering::Monotonic;
+    else
+      switch ((llvm::AtomicOrderingCABI)FOS) {
+      case llvm::AtomicOrderingCABI::relaxed:
+      case llvm::AtomicOrderingCABI::release:
+      case llvm::AtomicOrderingCABI::acq_rel:
+        FailureOrder = llvm::AtomicOrdering::Monotonic;
+        break;
+      case llvm::AtomicOrderingCABI::consume:
+      case llvm::AtomicOrderingCABI::acquire:
+        FailureOrder = llvm::AtomicOrdering::Acquire;
+        break;
+      case llvm::AtomicOrderingCABI::seq_cst:
+        FailureOrder = llvm::AtomicOrdering::SequentiallyConsistent;
+        break;
+      }
+    if (isStrongerThan(FailureOrder, SuccessOrder)) {
+      // Don't assert on undefined behavior "failure argument shall be no
+      // stronger than the success argument".
       FailureOrder =
-        llvm::AtomicCmpXchgInst::getStrongestFailureOrdering(SuccessOrder);
+          llvm::AtomicCmpXchgInst::getStrongestFailureOrdering(SuccessOrder);
     }
-    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size,
-                      SuccessOrder, FailureOrder);
+    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+                      FailureOrder);
     return;
   }
 
@@ -456,9 +443,10 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
   llvm::BasicBlock *MonotonicBB = nullptr, *AcquireBB = nullptr,
                    *SeqCstBB = nullptr;
   MonotonicBB = CGF.createBasicBlock("monotonic_fail", CGF.CurFn);
-  if (SuccessOrder != llvm::Monotonic && SuccessOrder != llvm::Release)
+  if (SuccessOrder != llvm::AtomicOrdering::Monotonic &&
+      SuccessOrder != llvm::AtomicOrdering::Release)
     AcquireBB = CGF.createBasicBlock("acquire_fail", CGF.CurFn);
-  if (SuccessOrder == llvm::SequentiallyConsistent)
+  if (SuccessOrder == llvm::AtomicOrdering::SequentiallyConsistent)
     SeqCstBB = CGF.createBasicBlock("seqcst_fail", CGF.CurFn);
 
   llvm::BasicBlock *ContBB = CGF.createBasicBlock("atomic.continue", CGF.CurFn);
@@ -472,25 +460,25 @@ static void emitAtomicCmpXchgFailureSet(CodeGenFunction &CGF, AtomicExpr *E,
   // doesn't fold to a constant for the ordering.
   CGF.Builder.SetInsertPoint(MonotonicBB);
   emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2,
-                    Size, SuccessOrder, llvm::Monotonic);
+                    Size, SuccessOrder, llvm::AtomicOrdering::Monotonic);
   CGF.Builder.CreateBr(ContBB);
 
   if (AcquireBB) {
     CGF.Builder.SetInsertPoint(AcquireBB);
     emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2,
-                      Size, SuccessOrder, llvm::Acquire);
+                      Size, SuccessOrder, llvm::AtomicOrdering::Acquire);
     CGF.Builder.CreateBr(ContBB);
-    SI->addCase(CGF.Builder.getInt32(AtomicExpr::AO_ABI_memory_order_consume),
+    SI->addCase(CGF.Builder.getInt32((int)llvm::AtomicOrderingCABI::consume),
                 AcquireBB);
-    SI->addCase(CGF.Builder.getInt32(AtomicExpr::AO_ABI_memory_order_acquire),
+    SI->addCase(CGF.Builder.getInt32((int)llvm::AtomicOrderingCABI::acquire),
                 AcquireBB);
   }
   if (SeqCstBB) {
     CGF.Builder.SetInsertPoint(SeqCstBB);
-    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2,
-                      Size, SuccessOrder, llvm::SequentiallyConsistent);
+    emitAtomicCmpXchg(CGF, E, IsWeak, Dest, Ptr, Val1, Val2, Size, SuccessOrder,
+                      llvm::AtomicOrdering::SequentiallyConsistent);
     CGF.Builder.CreateBr(ContBB);
-    SI->addCase(CGF.Builder.getInt32(AtomicExpr::AO_ABI_memory_order_seq_cst),
+    SI->addCase(CGF.Builder.getInt32((int)llvm::AtomicOrderingCABI::seq_cst),
                 SeqCstBB);
   }
 
@@ -1037,40 +1025,39 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
                 E->getOp() == AtomicExpr::AO__atomic_load_n;
 
   if (isa<llvm::ConstantInt>(Order)) {
-    int ord = cast<llvm::ConstantInt>(Order)->getZExtValue();
-    switch (ord) {
-    case AtomicExpr::AO_ABI_memory_order_relaxed:
-      EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                   Size, llvm::Monotonic);
-      break;
-    case AtomicExpr::AO_ABI_memory_order_consume:
-    case AtomicExpr::AO_ABI_memory_order_acquire:
-      if (IsStore)
-        break; // Avoid crashing on code with undefined behavior
-      EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                   Size, llvm::Acquire);
-      break;
-    case AtomicExpr::AO_ABI_memory_order_release:
-      if (IsLoad)
-        break; // Avoid crashing on code with undefined behavior
-      EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                   Size, llvm::Release);
-      break;
-    case AtomicExpr::AO_ABI_memory_order_acq_rel:
-      if (IsLoad || IsStore)
-        break; // Avoid crashing on code with undefined behavior
-      EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                   Size, llvm::AcquireRelease);
-      break;
-    case AtomicExpr::AO_ABI_memory_order_seq_cst:
-      EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                   Size, llvm::SequentiallyConsistent);
-      break;
-    default: // invalid order
-      // We should not ever get here normally, but it's hard to
-      // enforce that in general.
-      break;
-    }
+    auto ord = cast<llvm::ConstantInt>(Order)->getZExtValue();
+    // We should not ever get to a case where the ordering isn't a valid C ABI
+    // value, but it's hard to enforce that in general.
+    if (llvm::isValidAtomicOrderingCABI(ord))
+      switch ((llvm::AtomicOrderingCABI)ord) {
+      case llvm::AtomicOrderingCABI::relaxed:
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+                     llvm::AtomicOrdering::Monotonic);
+        break;
+      case llvm::AtomicOrderingCABI::consume:
+      case llvm::AtomicOrderingCABI::acquire:
+        if (IsStore)
+          break; // Avoid crashing on code with undefined behavior
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+                     llvm::AtomicOrdering::Acquire);
+        break;
+      case llvm::AtomicOrderingCABI::release:
+        if (IsLoad)
+          break; // Avoid crashing on code with undefined behavior
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+                     llvm::AtomicOrdering::Release);
+        break;
+      case llvm::AtomicOrderingCABI::acq_rel:
+        if (IsLoad || IsStore)
+          break; // Avoid crashing on code with undefined behavior
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+                     llvm::AtomicOrdering::AcquireRelease);
+        break;
+      case llvm::AtomicOrderingCABI::seq_cst:
+        EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail, Size,
+                     llvm::AtomicOrdering::SequentiallyConsistent);
+        break;
+      }
     if (RValTy->isVoidType())
       return RValue::get(nullptr);
 
@@ -1105,39 +1092,39 @@ RValue CodeGenFunction::EmitAtomicExpr(AtomicExpr *E) {
   // Emit all the different atomics
   Builder.SetInsertPoint(MonotonicBB);
   EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-               Size, llvm::Monotonic);
+               Size, llvm::AtomicOrdering::Monotonic);
   Builder.CreateBr(ContBB);
   if (!IsStore) {
     Builder.SetInsertPoint(AcquireBB);
     EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                 Size, llvm::Acquire);
+                 Size, llvm::AtomicOrdering::Acquire);
     Builder.CreateBr(ContBB);
-    SI->addCase(Builder.getInt32(AtomicExpr::AO_ABI_memory_order_consume),
+    SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::consume),
                 AcquireBB);
-    SI->addCase(Builder.getInt32(AtomicExpr::AO_ABI_memory_order_acquire),
+    SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::acquire),
                 AcquireBB);
   }
   if (!IsLoad) {
     Builder.SetInsertPoint(ReleaseBB);
     EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                 Size, llvm::Release);
+                 Size, llvm::AtomicOrdering::Release);
     Builder.CreateBr(ContBB);
-    SI->addCase(Builder.getInt32(AtomicExpr::AO_ABI_memory_order_release),
+    SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::release),
                 ReleaseBB);
   }
   if (!IsLoad && !IsStore) {
     Builder.SetInsertPoint(AcqRelBB);
     EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-                 Size, llvm::AcquireRelease);
+                 Size, llvm::AtomicOrdering::AcquireRelease);
     Builder.CreateBr(ContBB);
-    SI->addCase(Builder.getInt32(AtomicExpr::AO_ABI_memory_order_acq_rel),
+    SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::acq_rel),
                 AcqRelBB);
   }
   Builder.SetInsertPoint(SeqCstBB);
   EmitAtomicOp(*this, E, Dest, Ptr, Val1, Val2, IsWeak, OrderFail,
-               Size, llvm::SequentiallyConsistent);
+               Size, llvm::AtomicOrdering::SequentiallyConsistent);
   Builder.CreateBr(ContBB);
-  SI->addCase(Builder.getInt32(AtomicExpr::AO_ABI_memory_order_seq_cst),
+  SI->addCase(Builder.getInt32((int)llvm::AtomicOrderingCABI::seq_cst),
               SeqCstBB);
 
   // Cleanup and return
@@ -1257,9 +1244,9 @@ void AtomicInfo::EmitAtomicLoadLibcall(llvm::Value *AddForLoaded,
            CGF.getContext().VoidPtrTy);
   Args.add(RValue::get(CGF.EmitCastToVoidPtr(AddForLoaded)),
            CGF.getContext().VoidPtrTy);
-  Args.add(RValue::get(
-               llvm::ConstantInt::get(CGF.IntTy, translateAtomicOrdering(AO))),
-           CGF.getContext().IntTy);
+  Args.add(
+      RValue::get(llvm::ConstantInt::get(CGF.IntTy, (int)llvm::toCABI(AO))),
+      CGF.getContext().IntTy);
   emitAtomicLibcall(CGF, "__atomic_load", CGF.getContext().VoidTy, Args);
 }
 
@@ -1287,28 +1274,21 @@ bool CodeGenFunction::LValueIsSuitableForInlineAtomic(LValue LV) {
   bool IsVolatile = LV.isVolatile() || hasVolatileMember(LV.getType());
   // An atomic is inline if we don't need to use a libcall.
   bool AtomicIsInline = !AI.shouldUseLibcall();
+  // MSVC doesn't seem to do this for types wider than a pointer.
+  if (getContext().getTypeSize(LV.getType()) >
+      getContext().getTypeSize(getContext().getIntPtrType()))
+    return false;
   return IsVolatile && AtomicIsInline;
 }
 
-/// An type is a candidate for having its loads and stores be made atomic if
-/// we are operating under /volatile:ms *and* we know the access is volatile and
-/// performing such an operation can be performed without a libcall.
-bool CodeGenFunction::typeIsSuitableForInlineAtomic(QualType Ty,
-                                                    bool IsVolatile) const {
-  // An atomic is inline if we don't need to use a libcall (e.g. it is builtin).
-  bool AtomicIsInline = getContext().getTargetInfo().hasBuiltinAtomic(
-      getContext().getTypeSize(Ty), getContext().getTypeAlign(Ty));
-  return CGM.getCodeGenOpts().MSVolatile && IsVolatile && AtomicIsInline;
-}
-
 RValue CodeGenFunction::EmitAtomicLoad(LValue LV, SourceLocation SL,
                                        AggValueSlot Slot) {
   llvm::AtomicOrdering AO;
   bool IsVolatile = LV.isVolatileQualified();
   if (LV.getType()->isAtomicType()) {
-    AO = llvm::SequentiallyConsistent;
+    AO = llvm::AtomicOrdering::SequentiallyConsistent;
   } else {
-    AO = llvm::Acquire;
+    AO = llvm::AtomicOrdering::Acquire;
     IsVolatile = true;
   }
   return EmitAtomicLoad(LV, SL, AO, IsVolatile, Slot);
@@ -1462,11 +1442,11 @@ AtomicInfo::EmitAtomicCompareExchangeLibcall(llvm::Value *ExpectedAddr,
            CGF.getContext().VoidPtrTy);
   Args.add(RValue::get(CGF.EmitCastToVoidPtr(DesiredAddr)),
            CGF.getContext().VoidPtrTy);
-  Args.add(RValue::get(llvm::ConstantInt::get(
-               CGF.IntTy, translateAtomicOrdering(Success))),
+  Args.add(RValue::get(
+               llvm::ConstantInt::get(CGF.IntTy, (int)llvm::toCABI(Success))),
            CGF.getContext().IntTy);
-  Args.add(RValue::get(llvm::ConstantInt::get(
-               CGF.IntTy, translateAtomicOrdering(Failure))),
+  Args.add(RValue::get(
+               llvm::ConstantInt::get(CGF.IntTy, (int)llvm::toCABI(Failure))),
            CGF.getContext().IntTy);
   auto SuccessFailureRVal = emitAtomicLibcall(CGF, "__atomic_compare_exchange",
                                               CGF.getContext().BoolTy, Args);
@@ -1477,8 +1457,9 @@ AtomicInfo::EmitAtomicCompareExchangeLibcall(llvm::Value *ExpectedAddr,
 std::pair<RValue, llvm::Value *> AtomicInfo::EmitAtomicCompareExchange(
     RValue Expected, RValue Desired, llvm::AtomicOrdering Success,
     llvm::AtomicOrdering Failure, bool IsWeak) {
-  if (Failure >= Success)
-    // Don't assert on undefined behavior.
+  if (isStrongerThan(Failure, Success))
+    // Don't assert on undefined behavior "failure argument shall be no stronger
+    // than the success argument".
     Failure = llvm::AtomicCmpXchgInst::getStrongestFailureOrdering(Success);
 
   // Check whether we should use a library call.
@@ -1727,9 +1708,9 @@ void CodeGenFunction::EmitAtomicStore(RValue rvalue, LValue lvalue,
   bool IsVolatile = lvalue.isVolatileQualified();
   llvm::AtomicOrdering AO;
   if (lvalue.getType()->isAtomicType()) {
-    AO = llvm::SequentiallyConsistent;
+    AO = llvm::AtomicOrdering::SequentiallyConsistent;
   } else {
-    AO = llvm::Release;
+    AO = llvm::AtomicOrdering::Release;
     IsVolatile = true;
   }
   return EmitAtomicStore(rvalue, lvalue, AO, IsVolatile, isInit);
@@ -1772,9 +1753,9 @@ void CodeGenFunction::EmitAtomicStore(RValue rvalue, LValue dest,
                getContext().VoidPtrTy);
       args.add(RValue::get(EmitCastToVoidPtr(srcAddr.getPointer())),
                getContext().VoidPtrTy);
-      args.add(RValue::get(llvm::ConstantInt::get(
-                   IntTy, AtomicInfo::translateAtomicOrdering(AO))),
-               getContext().IntTy);
+      args.add(
+          RValue::get(llvm::ConstantInt::get(IntTy, (int)llvm::toCABI(AO))),
+          getContext().IntTy);
       emitAtomicLibcall(*this, "__atomic_store", getContext().VoidTy, args);
       return;
     }
diff --git a/lib/CodeGen/CGBlocks.cpp b/lib/CodeGen/CGBlocks.cpp
index ba2941e9df4a..5a41f361d9ac 100644
--- a/lib/CodeGen/CGBlocks.cpp
+++ b/lib/CodeGen/CGBlocks.cpp
@@ -262,6 +262,11 @@ static bool isSafeForCXXConstantCapture(QualType type) {
 static llvm::Constant *tryCaptureAsConstant(CodeGenModule &CGM,
                                             CodeGenFunction *CGF,
                                             const VarDecl *var) {
+  // Return if this is a function paramter. We shouldn't try to
+  // rematerialize default arguments of function parameters.
+  if (isa<ParmVarDecl>(var))
+    return nullptr;
+
   QualType type = var->getType();
 
   // We can only do this if the variable is const.
@@ -508,7 +513,7 @@ static void computeBlockInfo(CodeGenModule &CGM, CodeGenFunction *CGF,
   // At this point, we just have to add padding if the end align still
   // isn't aligned right.
   if (endAlign < maxFieldAlign) {
-    CharUnits newBlockSize = blockSize.RoundUpToAlignment(maxFieldAlign);
+    CharUnits newBlockSize = blockSize.alignTo(maxFieldAlign);
     CharUnits padding = newBlockSize - blockSize;
 
     // If we haven't yet added any fields, remember that there was an
@@ -775,35 +780,34 @@ llvm::Value *CodeGenFunction::EmitBlockLiteral(const CGBlockInfo &blockInfo) {
     // Compute the address of the thing we're going to move into the
     // block literal.
     Address src = Address::invalid();
-    if (BlockInfo && CI.isNested()) {
-      // We need to use the capture from the enclosing block.
-      const CGBlockInfo::Capture &enclosingCapture =
-        BlockInfo->getCapture(variable);
-
-      // This is a [[type]]*, except that a byref entry wil just be an i8**.
-      src = Builder.CreateStructGEP(LoadBlockStruct(),
-                                    enclosingCapture.getIndex(),
-                                    enclosingCapture.getOffset(),
-                                    "block.capture.addr");
-    } else if (blockDecl->isConversionFromLambda()) {
+
+    if (blockDecl->isConversionFromLambda()) {
       // The lambda capture in a lambda's conversion-to-block-pointer is
       // special; we'll simply emit it directly.
       src = Address::invalid();
-    } else {
-      // Just look it up in the locals map, which will give us back a
-      // [[type]]*.  If that doesn't work, do the more elaborate DRE
-      // emission.
-      auto it = LocalDeclMap.find(variable);
-      if (it != LocalDeclMap.end()) {
-        src = it->second;
+    } else if (CI.isByRef()) {
+      if (BlockInfo && CI.isNested()) {
+        // We need to use the capture from the enclosing block.
+        const CGBlockInfo::Capture &enclosingCapture =
+            BlockInfo->getCapture(variable);
+
+        // This is a [[type]]*, except that a byref entry wil just be an i8**.
+        src = Builder.CreateStructGEP(LoadBlockStruct(),
+                                      enclosingCapture.getIndex(),
+                                      enclosingCapture.getOffset(),
+                                      "block.capture.addr");
       } else {
-        DeclRefExpr declRef(
-            const_cast<VarDecl *>(variable),
-            /*RefersToEnclosingVariableOrCapture*/ CI.isNested(), type,
-            VK_LValue, SourceLocation());
-        src = EmitDeclRefLValue(&declRef).getAddress();
+        auto I = LocalDeclMap.find(variable);
+        assert(I != LocalDeclMap.end());
+        src = I->second;
       }
-    }
+    } else {
+      DeclRefExpr declRef(const_cast<VarDecl *>(variable),
+                          /*RefersToEnclosingVariableOrCapture*/ CI.isNested(),
+                          type.getNonReferenceType(), VK_LValue,
+                          SourceLocation());
+      src = EmitDeclRefLValue(&declRef).getAddress();
+    };
 
     // For byrefs, we just write the pointer to the byref struct into
     // the block field.  There's no need to chase the forwarding
@@ -837,8 +841,7 @@ llvm::Value *CodeGenFunction::EmitBlockLiteral(const CGBlockInfo &blockInfo) {
 
     // If it's a reference variable, copy the reference into the block field.
     } else if (type->isReferenceType()) {
-      llvm::Value *ref = Builder.CreateLoad(src, "ref.val");
-      Builder.CreateStore(ref, blockField);
+      Builder.CreateStore(src.getPointer(), blockField);
 
     // If this is an ARC __strong block-pointer variable, don't do a
     // block copy.
@@ -1109,8 +1112,8 @@ void CodeGenFunction::setBlockContextParameter(const ImplicitParamDecl *D,
   }
 
   if (CGDebugInfo *DI = getDebugInfo()) {
-    if (CGM.getCodeGenOpts().getDebugInfo()
-          >= CodeGenOptions::LimitedDebugInfo) {
+    if (CGM.getCodeGenOpts().getDebugInfo() >=
+        codegenoptions::LimitedDebugInfo) {
       DI->setLocation(D->getLocation());
       DI->EmitDeclareOfBlockLiteralArgVariable(*BlockInfo, arg, argNum,
                                                localAddr, Builder);
@@ -1174,9 +1177,8 @@ CodeGenFunction::GenerateBlockFunction(GlobalDecl GD,
 
   // Create the function declaration.
   const FunctionProtoType *fnType = blockInfo.getBlockExpr()->getFunctionType();
-  const CGFunctionInfo &fnInfo = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      fnType->getReturnType(), args, fnType->getExtInfo(),
-      fnType->isVariadic());
+  const CGFunctionInfo &fnInfo =
+    CGM.getTypes().arrangeBlockFunctionDeclaration(fnType, args);
   if (CGM.ReturnSlotInterferesWithArgs(fnInfo))
     blockInfo.UsesStret = true;
 
@@ -1260,8 +1262,8 @@ CodeGenFunction::GenerateBlockFunction(GlobalDecl GD,
       const VarDecl *variable = CI.getVariable();
       DI->EmitLocation(Builder, variable->getLocation());
 
-      if (CGM.getCodeGenOpts().getDebugInfo()
-            >= CodeGenOptions::LimitedDebugInfo) {
+      if (CGM.getCodeGenOpts().getDebugInfo() >=
+          codegenoptions::LimitedDebugInfo) {
         const CGBlockInfo::Capture &capture = blockInfo.getCapture(variable);
         if (capture.isConstant()) {
           auto addr = LocalDeclMap.find(variable)->second;
@@ -1329,8 +1331,8 @@ CodeGenFunction::GenerateCopyHelperFunction(const CGBlockInfo &blockInfo) {
                             C.VoidPtrTy);
   args.push_back(&srcDecl);
 
-  const CGFunctionInfo &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      C.VoidTy, args, FunctionType::ExtInfo(), /*variadic=*/false);
+  const CGFunctionInfo &FI =
+    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, args);
 
   // FIXME: it would be nice if these were mergeable with things with
   // identical semantics.
@@ -1505,8 +1507,8 @@ CodeGenFunction::GenerateDestroyHelperFunction(const CGBlockInfo &blockInfo) {
                             C.VoidPtrTy);
   args.push_back(&srcDecl);
 
-  const CGFunctionInfo &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      C.VoidTy, args, FunctionType::ExtInfo(), /*variadic=*/false);
+  const CGFunctionInfo &FI =
+    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, args);
 
   // FIXME: We'd like to put these into a mergable by content, with
   // internal linkage.
@@ -1791,8 +1793,8 @@ generateByrefCopyHelper(CodeGenFunction &CGF, const BlockByrefInfo &byrefInfo,
                         Context.VoidPtrTy);
   args.push_back(&src);
 
-  const CGFunctionInfo &FI = CGF.CGM.getTypes().arrangeFreeFunctionDeclaration(
-      R, args, FunctionType::ExtInfo(), /*variadic=*/false);
+  const CGFunctionInfo &FI =
+    CGF.CGM.getTypes().arrangeBuiltinFunctionDeclaration(R, args);
 
   llvm::FunctionType *LTy = CGF.CGM.getTypes().GetFunctionType(FI);
 
@@ -1864,8 +1866,8 @@ generateByrefDisposeHelper(CodeGenFunction &CGF,
                         Context.VoidPtrTy);
   args.push_back(&src);
 
-  const CGFunctionInfo &FI = CGF.CGM.getTypes().arrangeFreeFunctionDeclaration(
-      R, args, FunctionType::ExtInfo(), /*variadic=*/false);
+  const CGFunctionInfo &FI =
+    CGF.CGM.getTypes().arrangeBuiltinFunctionDeclaration(R, args);
 
   llvm::FunctionType *LTy = CGF.CGM.getTypes().GetFunctionType(FI);
 
@@ -2108,7 +2110,7 @@ const BlockByrefInfo &CodeGenFunction::getBlockByrefInfo(const VarDecl *D) {
 
   bool packed = false;
   CharUnits varAlign = getContext().getDeclAlign(D);
-  CharUnits varOffset = size.RoundUpToAlignment(varAlign);
+  CharUnits varOffset = size.alignTo(varAlign);
 
   // We may have to insert padding.
   if (varOffset != size) {
@@ -2285,9 +2287,36 @@ void CodeGenFunction::enterByrefCleanup(const AutoVarEmission &emission) {
 /// Adjust the declaration of something from the blocks API.
 static void configureBlocksRuntimeObject(CodeGenModule &CGM,
                                          llvm::Constant *C) {
-  if (!CGM.getLangOpts().BlocksRuntimeOptional) return;
-
   auto *GV = cast<llvm::GlobalValue>(C->stripPointerCasts());
+
+  if (CGM.getTarget().getTriple().isOSBinFormatCOFF()) {
+    IdentifierInfo &II = CGM.getContext().Idents.get(C->getName());
+    TranslationUnitDecl *TUDecl = CGM.getContext().getTranslationUnitDecl();
+    DeclContext *DC = TranslationUnitDecl::castToDeclContext(TUDecl);
+
+    assert((isa<llvm::Function>(C->stripPointerCasts()) ||
+            isa<llvm::GlobalVariable>(C->stripPointerCasts())) &&
+           "expected Function or GlobalVariable");
+
+    const NamedDecl *ND = nullptr;
+    for (const auto &Result : DC->lookup(&II))
+      if ((ND = dyn_cast<FunctionDecl>(Result)) ||
+          (ND = dyn_cast<VarDecl>(Result)))
+        break;
+
+    // TODO: support static blocks runtime
+    if (GV->isDeclaration() && (!ND || !ND->hasAttr<DLLExportAttr>())) {
+      GV->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
+      GV->setLinkage(llvm::GlobalValue::ExternalLinkage);
+    } else {
+      GV->setDLLStorageClass(llvm::GlobalValue::DLLExportStorageClass);
+      GV->setLinkage(llvm::GlobalValue::ExternalLinkage);
+    }
+  }
+
+  if (!CGM.getLangOpts().BlocksRuntimeOptional)
+    return;
+
   if (GV->isDeclaration() && GV->hasExternalLinkage())
     GV->setLinkage(llvm::GlobalValue::ExternalWeakLinkage);
 }
@@ -2335,5 +2364,5 @@ llvm::Constant *CodeGenModule::getNSConcreteStackBlock() {
                                                Int8PtrTy->getPointerTo(),
                                                nullptr);
   configureBlocksRuntimeObject(*this, NSConcreteStackBlock);
-  return NSConcreteStackBlock;  
+  return NSConcreteStackBlock;
 }
diff --git a/lib/CodeGen/CGBuilder.h b/lib/CodeGen/CGBuilder.h
index 489f3413d4b8..027435d7c599 100644
--- a/lib/CodeGen/CGBuilder.h
+++ b/lib/CodeGen/CGBuilder.h
@@ -10,6 +10,7 @@
 #ifndef LLVM_CLANG_LIB_CODEGEN_CGBUILDER_H
 #define LLVM_CLANG_LIB_CODEGEN_CGBUILDER_H
 
+#include "llvm/IR/DataLayout.h"
 #include "llvm/IR/IRBuilder.h"
 #include "Address.h"
 #include "CodeGenTypeCache.h"
@@ -22,9 +23,7 @@ class CodeGenFunction;
 /// \brief This is an IRBuilder insertion helper that forwards to
 /// CodeGenFunction::InsertHelper, which adds necessary metadata to
 /// instructions.
-template <bool PreserveNames>
-class CGBuilderInserter
-    : protected llvm::IRBuilderDefaultInserter<PreserveNames> {
+class CGBuilderInserter : protected llvm::IRBuilderDefaultInserter {
 public:
   CGBuilderInserter() = default;
   explicit CGBuilderInserter(CodeGenFunction *CGF) : CGF(CGF) {}
@@ -38,17 +37,10 @@ private:
   CodeGenFunction *CGF = nullptr;
 };
 
-// Don't preserve names on values in an optimized build.
-#ifdef NDEBUG
-#define PreserveNames false
-#else
-#define PreserveNames true
-#endif
-
-typedef CGBuilderInserter<PreserveNames> CGBuilderInserterTy;
+typedef CGBuilderInserter CGBuilderInserterTy;
 
-typedef llvm::IRBuilder<PreserveNames, llvm::ConstantFolder,
-                        CGBuilderInserterTy> CGBuilderBaseTy;
+typedef llvm::IRBuilder<llvm::ConstantFolder, CGBuilderInserterTy>
+    CGBuilderBaseTy;
 
 class CGBuilderTy : public CGBuilderBaseTy {
   /// Storing a reference to the type cache here makes it a lot easier
@@ -194,6 +186,12 @@ public:
                                    Addr.getPointer(), Index, Name),
                    Addr.getAlignment().alignmentAtOffset(Offset));
   }
+  Address CreateStructGEP(Address Addr, unsigned Index,
+                          const llvm::StructLayout *Layout,
+                          const llvm::Twine &Name = "") {
+    auto Offset = CharUnits::fromQuantity(Layout->getElementOffset(Index));
+    return CreateStructGEP(Addr, Index, Offset, Name);
+  }
 
   /// Given
   ///   %addr = [n x T]* ...
@@ -298,8 +296,6 @@ public:
   }
 };
 
-#undef PreserveNames
-
 }  // end namespace CodeGen
 }  // end namespace clang
 
diff --git a/lib/CodeGen/CGBuiltin.cpp b/lib/CodeGen/CGBuiltin.cpp
index 787ac5361bbb..c74e53ea84e0 100644
--- a/lib/CodeGen/CGBuiltin.cpp
+++ b/lib/CodeGen/CGBuiltin.cpp
@@ -26,6 +26,7 @@
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/InlineAsm.h"
 #include "llvm/IR/Intrinsics.h"
+#include "llvm/IR/MDBuilder.h"
 #include <sstream>
 
 using namespace clang;
@@ -105,9 +106,8 @@ static Value *MakeBinaryAtomicValue(CodeGenFunction &CGF,
   llvm::Type *ValueType = Args[1]->getType();
   Args[1] = EmitToInt(CGF, Args[1], T, IntType);
 
-  llvm::Value *Result =
-      CGF.Builder.CreateAtomicRMW(Kind, Args[0], Args[1],
-                                  llvm::SequentiallyConsistent);
+  llvm::Value *Result = CGF.Builder.CreateAtomicRMW(
+      Kind, Args[0], Args[1], llvm::AtomicOrdering::SequentiallyConsistent);
   return EmitFromInt(CGF, Result, T, ValueType);
 }
 
@@ -167,9 +167,8 @@ static RValue EmitBinaryAtomicPost(CodeGenFunction &CGF,
   Args[1] = EmitToInt(CGF, Args[1], T, IntType);
   Args[0] = CGF.Builder.CreateBitCast(DestPtr, IntPtrType);
 
-  llvm::Value *Result =
-      CGF.Builder.CreateAtomicRMW(Kind, Args[0], Args[1],
-                                  llvm::SequentiallyConsistent);
+  llvm::Value *Result = CGF.Builder.CreateAtomicRMW(
+      Kind, Args[0], Args[1], llvm::AtomicOrdering::SequentiallyConsistent);
   Result = CGF.Builder.CreateBinOp(Op, Result, Args[1]);
   if (Invert)
     Result = CGF.Builder.CreateBinOp(llvm::Instruction::Xor, Result,
@@ -206,9 +205,9 @@ static Value *MakeAtomicCmpXchgValue(CodeGenFunction &CGF, const CallExpr *E,
   Args[1] = EmitToInt(CGF, Args[1], T, IntType);
   Args[2] = EmitToInt(CGF, CGF.EmitScalarExpr(E->getArg(2)), T, IntType);
 
-  Value *Pair = CGF.Builder.CreateAtomicCmpXchg(Args[0], Args[1], Args[2],
-                                                llvm::SequentiallyConsistent,
-                                                llvm::SequentiallyConsistent);
+  Value *Pair = CGF.Builder.CreateAtomicCmpXchg(
+      Args[0], Args[1], Args[2], llvm::AtomicOrdering::SequentiallyConsistent,
+      llvm::AtomicOrdering::SequentiallyConsistent);
   if (ReturnBool)
     // Extract boolean success flag and zext it to int.
     return CGF.Builder.CreateZExt(CGF.Builder.CreateExtractValue(Pair, 1),
@@ -219,6 +218,51 @@ static Value *MakeAtomicCmpXchgValue(CodeGenFunction &CGF, const CallExpr *E,
                        ValueType);
 }
 
+// Emit a simple mangled intrinsic that has 1 argument and a return type
+// matching the argument type.
+static Value *emitUnaryBuiltin(CodeGenFunction &CGF,
+                               const CallExpr *E,
+                               unsigned IntrinsicID) {
+  llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
+
+  Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
+  return CGF.Builder.CreateCall(F, Src0);
+}
+
+// Emit an intrinsic that has 2 operands of the same type as its result.
+static Value *emitBinaryBuiltin(CodeGenFunction &CGF,
+                                const CallExpr *E,
+                                unsigned IntrinsicID) {
+  llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
+  llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
+
+  Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
+  return CGF.Builder.CreateCall(F, { Src0, Src1 });
+}
+
+// Emit an intrinsic that has 3 operands of the same type as its result.
+static Value *emitTernaryBuiltin(CodeGenFunction &CGF,
+                                 const CallExpr *E,
+                                 unsigned IntrinsicID) {
+  llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
+  llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
+  llvm::Value *Src2 = CGF.EmitScalarExpr(E->getArg(2));
+
+  Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
+  return CGF.Builder.CreateCall(F, { Src0, Src1, Src2 });
+}
+
+// Emit an intrinsic that has 1 float or double operand, and 1 integer.
+static Value *emitFPIntBuiltin(CodeGenFunction &CGF,
+                               const CallExpr *E,
+                               unsigned IntrinsicID) {
+  llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
+  llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
+
+  Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
+  return CGF.Builder.CreateCall(F, {Src0, Src1});
+}
+
 /// EmitFAbs - Emit a call to @llvm.fabs().
 static Value *EmitFAbs(CodeGenFunction &CGF, Value *V) {
   Value *F = CGF.CGM.getIntrinsic(Intrinsic::fabs, V->getType());
@@ -248,8 +292,8 @@ static Value *EmitSignBit(CodeGenFunction &CGF, Value *V) {
     if (CGF.getTarget().isBigEndian()) {
       Value *ShiftCst = llvm::ConstantInt::get(IntTy, Width);
       V = CGF.Builder.CreateLShr(V, ShiftCst);
-    } 
-    // We are truncating value in order to extract the higher-order 
+    }
+    // We are truncating value in order to extract the higher-order
     // double, which we will be using to extract the sign from.
     IntTy = llvm::IntegerType::get(C, Width);
     V = CGF.Builder.CreateTrunc(V, IntTy);
@@ -288,6 +332,17 @@ static llvm::Value *EmitOverflowIntrinsic(CodeGenFunction &CGF,
   return CGF.Builder.CreateExtractValue(Tmp, 0);
 }
 
+static Value *emitRangedBuiltin(CodeGenFunction &CGF,
+                                unsigned IntrinsicID,
+                                int low, int high) {
+    llvm::MDBuilder MDHelper(CGF.getLLVMContext());
+    llvm::MDNode *RNode = MDHelper.createRange(APInt(32, low), APInt(32, high));
+    Value *F = CGF.CGM.getIntrinsic(IntrinsicID, {});
+    llvm::Instruction *Call = CGF.Builder.CreateCall(F);
+    Call->setMetadata(llvm::LLVMContext::MD_range, RNode);
+    return Call;
+}
+
 namespace {
   struct WidthAndSignedness {
     unsigned Width;
@@ -465,9 +520,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
   case Builtin::BI__builtin_fabs:
   case Builtin::BI__builtin_fabsf:
   case Builtin::BI__builtin_fabsl: {
-    Value *Arg1 = EmitScalarExpr(E->getArg(0));
-    Value *Result = EmitFAbs(*this, Arg1);
-    return RValue::get(Result);
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::fabs));
   }
   case Builtin::BI__builtin_fmod:
   case Builtin::BI__builtin_fmodf:
@@ -477,7 +530,51 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     Value *Result = Builder.CreateFRem(Arg1, Arg2, "fmod");
     return RValue::get(Result);
   }
-
+  case Builtin::BI__builtin_copysign:
+  case Builtin::BI__builtin_copysignf:
+  case Builtin::BI__builtin_copysignl: {
+    return RValue::get(emitBinaryBuiltin(*this, E, Intrinsic::copysign));
+  }
+  case Builtin::BI__builtin_ceil:
+  case Builtin::BI__builtin_ceilf:
+  case Builtin::BI__builtin_ceill: {
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::ceil));
+  }
+  case Builtin::BI__builtin_floor:
+  case Builtin::BI__builtin_floorf:
+  case Builtin::BI__builtin_floorl: {
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::floor));
+  }
+  case Builtin::BI__builtin_trunc:
+  case Builtin::BI__builtin_truncf:
+  case Builtin::BI__builtin_truncl: {
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::trunc));
+  }
+  case Builtin::BI__builtin_rint:
+  case Builtin::BI__builtin_rintf:
+  case Builtin::BI__builtin_rintl: {
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::rint));
+  }
+  case Builtin::BI__builtin_nearbyint:
+  case Builtin::BI__builtin_nearbyintf:
+  case Builtin::BI__builtin_nearbyintl: {
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::nearbyint));
+  }
+  case Builtin::BI__builtin_round:
+  case Builtin::BI__builtin_roundf:
+  case Builtin::BI__builtin_roundl: {
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::round));
+  }
+  case Builtin::BI__builtin_fmin:
+  case Builtin::BI__builtin_fminf:
+  case Builtin::BI__builtin_fminl: {
+    return RValue::get(emitBinaryBuiltin(*this, E, Intrinsic::minnum));
+  }
+  case Builtin::BI__builtin_fmax:
+  case Builtin::BI__builtin_fmaxf:
+  case Builtin::BI__builtin_fmaxl: {
+    return RValue::get(emitBinaryBuiltin(*this, E, Intrinsic::maxnum));
+  }
   case Builtin::BI__builtin_conj:
   case Builtin::BI__builtin_conjf:
   case Builtin::BI__builtin_conjl: {
@@ -645,10 +742,13 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
   case Builtin::BI__builtin_bswap16:
   case Builtin::BI__builtin_bswap32:
   case Builtin::BI__builtin_bswap64: {
-    Value *ArgValue = EmitScalarExpr(E->getArg(0));
-    llvm::Type *ArgType = ArgValue->getType();
-    Value *F = CGM.getIntrinsic(Intrinsic::bswap, ArgType);
-    return RValue::get(Builder.CreateCall(F, ArgValue));
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::bswap));
+  }
+  case Builtin::BI__builtin_bitreverse8:
+  case Builtin::BI__builtin_bitreverse16:
+  case Builtin::BI__builtin_bitreverse32:
+  case Builtin::BI__builtin_bitreverse64: {
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::bitreverse));
   }
   case Builtin::BI__builtin_object_size: {
     unsigned Type =
@@ -751,13 +851,19 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
   }
 
-  case Builtin::BI__builtin_isinf: {
-    // isinf(x) --> fabs(x) == infinity
+  case Builtin::BI__builtin_isinf:
+  case Builtin::BI__builtin_isfinite: {
+    // isinf(x)    --> fabs(x) == infinity
+    // isfinite(x) --> fabs(x) != infinity
+    // x != NaN via the ordered compare in either case.
     Value *V = EmitScalarExpr(E->getArg(0));
-    V = EmitFAbs(*this, V);
-
-    V = Builder.CreateFCmpOEQ(V, ConstantFP::getInfinity(V->getType()),"isinf");
-    return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
+    Value *Fabs = EmitFAbs(*this, V);
+    Constant *Infinity = ConstantFP::getInfinity(V->getType());
+    CmpInst::Predicate Pred = (BuiltinID == Builtin::BI__builtin_isinf)
+                                  ? CmpInst::FCMP_OEQ
+                                  : CmpInst::FCMP_ONE;
+    Value *FCmp = Builder.CreateFCmp(Pred, Fabs, Infinity, "cmpinf");
+    return RValue::get(Builder.CreateZExt(FCmp, ConvertType(E->getType())));
   }
 
   case Builtin::BI__builtin_isinf_sign: {
@@ -795,19 +901,6 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
   }
 
-  case Builtin::BI__builtin_isfinite: {
-    // isfinite(x) --> x == x && fabs(x) != infinity;
-    Value *V = EmitScalarExpr(E->getArg(0));
-    Value *Eq = Builder.CreateFCmpOEQ(V, V, "iseq");
-
-    Value *Abs = EmitFAbs(*this, V);
-    Value *IsNotInf =
-      Builder.CreateFCmpUNE(Abs, ConstantFP::getInfinity(V->getType()),"isinf");
-
-    V = Builder.CreateAnd(Eq, IsNotInf, "and");
-    return RValue::get(Builder.CreateZExt(V, ConvertType(E->getType())));
-  }
-
   case Builtin::BI__builtin_fpclassify: {
     Value *V = EmitScalarExpr(E->getArg(5));
     llvm::Type *Ty = ConvertType(E->getArg(5)->getType());
@@ -1258,7 +1351,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     llvm::StoreInst *Store =
       Builder.CreateAlignedStore(llvm::Constant::getNullValue(ITy), Ptr,
                                  StoreSize);
-    Store->setAtomic(llvm::Release);
+    Store->setAtomic(llvm::AtomicOrdering::Release);
     return RValue::get(nullptr);
   }
 
@@ -1270,7 +1363,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     // any way to safely use it... but in practice, it mostly works
     // to use it with non-atomic loads and stores to get acquire/release
     // semantics.
-    Builder.CreateFence(llvm::SequentiallyConsistent);
+    Builder.CreateFence(llvm::AtomicOrdering::SequentiallyConsistent);
     return RValue::get(nullptr);
   }
 
@@ -1294,9 +1387,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
       Args.add(RValue::get(llvm::Constant::getNullValue(VoidPtrTy)),
                getContext().VoidPtrTy);
     const CGFunctionInfo &FuncInfo =
-        CGM.getTypes().arrangeFreeFunctionCall(E->getType(), Args,
-                                               FunctionType::ExtInfo(),
-                                               RequiredArgs::All);
+        CGM.getTypes().arrangeBuiltinFunctionCall(E->getType(), Args);
     llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FuncInfo);
     llvm::Constant *Func = CGM.CreateRuntimeFunction(FTy, LibCallName);
     return EmitCall(FuncInfo, Func, ReturnValueSlot(), Args);
@@ -1320,30 +1411,27 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
       switch (ord) {
       case 0:  // memory_order_relaxed
       default: // invalid order
-        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
-                                         Ptr, NewVal,
-                                         llvm::Monotonic);
+        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg, Ptr, NewVal,
+                                         llvm::AtomicOrdering::Monotonic);
         break;
-      case 1:  // memory_order_consume
-      case 2:  // memory_order_acquire
-        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
-                                         Ptr, NewVal,
-                                         llvm::Acquire);
+      case 1: // memory_order_consume
+      case 2: // memory_order_acquire
+        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg, Ptr, NewVal,
+                                         llvm::AtomicOrdering::Acquire);
         break;
-      case 3:  // memory_order_release
-        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
-                                         Ptr, NewVal,
-                                         llvm::Release);
+      case 3: // memory_order_release
+        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg, Ptr, NewVal,
+                                         llvm::AtomicOrdering::Release);
         break;
-      case 4:  // memory_order_acq_rel
-        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
-                                         Ptr, NewVal,
-                                         llvm::AcquireRelease);
+      case 4: // memory_order_acq_rel
+
+        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg, Ptr, NewVal,
+                                         llvm::AtomicOrdering::AcquireRelease);
         break;
-      case 5:  // memory_order_seq_cst
-        Result = Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
-                                         Ptr, NewVal,
-                                         llvm::SequentiallyConsistent);
+      case 5: // memory_order_seq_cst
+        Result = Builder.CreateAtomicRMW(
+            llvm::AtomicRMWInst::Xchg, Ptr, NewVal,
+            llvm::AtomicOrdering::SequentiallyConsistent);
         break;
       }
       Result->setVolatile(Volatile);
@@ -1360,9 +1448,9 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
       createBasicBlock("seqcst", CurFn)
     };
     llvm::AtomicOrdering Orders[5] = {
-      llvm::Monotonic, llvm::Acquire, llvm::Release,
-      llvm::AcquireRelease, llvm::SequentiallyConsistent
-    };
+        llvm::AtomicOrdering::Monotonic, llvm::AtomicOrdering::Acquire,
+        llvm::AtomicOrdering::Release, llvm::AtomicOrdering::AcquireRelease,
+        llvm::AtomicOrdering::SequentiallyConsistent};
 
     Order = Builder.CreateIntCast(Order, Builder.getInt32Ty(), false);
     llvm::SwitchInst *SI = Builder.CreateSwitch(Order, BBs[0]);
@@ -1406,13 +1494,13 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
       switch (ord) {
       case 0:  // memory_order_relaxed
       default: // invalid order
-        Store->setOrdering(llvm::Monotonic);
+        Store->setOrdering(llvm::AtomicOrdering::Monotonic);
         break;
       case 3:  // memory_order_release
-        Store->setOrdering(llvm::Release);
+        Store->setOrdering(llvm::AtomicOrdering::Release);
         break;
       case 5:  // memory_order_seq_cst
-        Store->setOrdering(llvm::SequentiallyConsistent);
+        Store->setOrdering(llvm::AtomicOrdering::SequentiallyConsistent);
         break;
       }
       return RValue::get(nullptr);
@@ -1426,8 +1514,8 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
       createBasicBlock("seqcst", CurFn)
     };
     llvm::AtomicOrdering Orders[3] = {
-      llvm::Monotonic, llvm::Release, llvm::SequentiallyConsistent
-    };
+        llvm::AtomicOrdering::Monotonic, llvm::AtomicOrdering::Release,
+        llvm::AtomicOrdering::SequentiallyConsistent};
 
     Order = Builder.CreateIntCast(Order, Builder.getInt32Ty(), false);
     llvm::SwitchInst *SI = Builder.CreateSwitch(Order, BBs[0]);
@@ -1466,16 +1554,17 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
         break;
       case 1:  // memory_order_consume
       case 2:  // memory_order_acquire
-        Builder.CreateFence(llvm::Acquire, Scope);
+        Builder.CreateFence(llvm::AtomicOrdering::Acquire, Scope);
         break;
       case 3:  // memory_order_release
-        Builder.CreateFence(llvm::Release, Scope);
+        Builder.CreateFence(llvm::AtomicOrdering::Release, Scope);
         break;
       case 4:  // memory_order_acq_rel
-        Builder.CreateFence(llvm::AcquireRelease, Scope);
+        Builder.CreateFence(llvm::AtomicOrdering::AcquireRelease, Scope);
         break;
       case 5:  // memory_order_seq_cst
-        Builder.CreateFence(llvm::SequentiallyConsistent, Scope);
+        Builder.CreateFence(llvm::AtomicOrdering::SequentiallyConsistent,
+                            Scope);
         break;
       }
       return RValue::get(nullptr);
@@ -1492,23 +1581,23 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     llvm::SwitchInst *SI = Builder.CreateSwitch(Order, ContBB);
 
     Builder.SetInsertPoint(AcquireBB);
-    Builder.CreateFence(llvm::Acquire, Scope);
+    Builder.CreateFence(llvm::AtomicOrdering::Acquire, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32(1), AcquireBB);
     SI->addCase(Builder.getInt32(2), AcquireBB);
 
     Builder.SetInsertPoint(ReleaseBB);
-    Builder.CreateFence(llvm::Release, Scope);
+    Builder.CreateFence(llvm::AtomicOrdering::Release, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32(3), ReleaseBB);
 
     Builder.SetInsertPoint(AcqRelBB);
-    Builder.CreateFence(llvm::AcquireRelease, Scope);
+    Builder.CreateFence(llvm::AtomicOrdering::AcquireRelease, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32(4), AcqRelBB);
 
     Builder.SetInsertPoint(SeqCstBB);
-    Builder.CreateFence(llvm::SequentiallyConsistent, Scope);
+    Builder.CreateFence(llvm::AtomicOrdering::SequentiallyConsistent, Scope);
     Builder.CreateBr(ContBB);
     SI->addCase(Builder.getInt32(5), SeqCstBB);
 
@@ -1794,7 +1883,7 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
       break;
     }
 
-    
+
     llvm::Value *Carry;
     llvm::Value *Sum = EmitOverflowIntrinsic(*this, IntrinsicId, X, Y, Carry);
     Builder.CreateStore(Sum, SumOutPtr);
@@ -1839,9 +1928,10 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
     llvm::Value *Comparand =
       Builder.CreatePtrToInt(EmitScalarExpr(E->getArg(2)), IntType);
 
-    auto Result = Builder.CreateAtomicCmpXchg(Destination, Comparand, Exchange,
-                                              SequentiallyConsistent,
-                                              SequentiallyConsistent);
+    auto Result =
+        Builder.CreateAtomicCmpXchg(Destination, Comparand, Exchange,
+                                    AtomicOrdering::SequentiallyConsistent,
+                                    AtomicOrdering::SequentiallyConsistent);
     Result->setVolatile(true);
 
     return RValue::get(Builder.CreateIntToPtr(Builder.CreateExtractValue(Result,
@@ -1853,44 +1943,47 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
         EmitScalarExpr(E->getArg(0)),
         EmitScalarExpr(E->getArg(2)),
         EmitScalarExpr(E->getArg(1)),
-        SequentiallyConsistent,
-        SequentiallyConsistent);
+        AtomicOrdering::SequentiallyConsistent,
+        AtomicOrdering::SequentiallyConsistent);
       CXI->setVolatile(true);
       return RValue::get(Builder.CreateExtractValue(CXI, 0));
   }
   case Builtin::BI_InterlockedIncrement: {
+    llvm::Type *IntTy = ConvertType(E->getType());
     AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
       AtomicRMWInst::Add,
       EmitScalarExpr(E->getArg(0)),
-      ConstantInt::get(Int32Ty, 1),
-      llvm::SequentiallyConsistent);
+      ConstantInt::get(IntTy, 1),
+      llvm::AtomicOrdering::SequentiallyConsistent);
     RMWI->setVolatile(true);
-    return RValue::get(Builder.CreateAdd(RMWI, ConstantInt::get(Int32Ty, 1)));
+    return RValue::get(Builder.CreateAdd(RMWI, ConstantInt::get(IntTy, 1)));
   }
   case Builtin::BI_InterlockedDecrement: {
+    llvm::Type *IntTy = ConvertType(E->getType());
     AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
       AtomicRMWInst::Sub,
       EmitScalarExpr(E->getArg(0)),
-      ConstantInt::get(Int32Ty, 1),
-      llvm::SequentiallyConsistent);
+      ConstantInt::get(IntTy, 1),
+      llvm::AtomicOrdering::SequentiallyConsistent);
     RMWI->setVolatile(true);
-    return RValue::get(Builder.CreateSub(RMWI, ConstantInt::get(Int32Ty, 1)));
+    return RValue::get(Builder.CreateSub(RMWI, ConstantInt::get(IntTy, 1)));
   }
   case Builtin::BI_InterlockedExchangeAdd: {
     AtomicRMWInst *RMWI = Builder.CreateAtomicRMW(
       AtomicRMWInst::Add,
       EmitScalarExpr(E->getArg(0)),
       EmitScalarExpr(E->getArg(1)),
-      llvm::SequentiallyConsistent);
+      llvm::AtomicOrdering::SequentiallyConsistent);
     RMWI->setVolatile(true);
     return RValue::get(RMWI);
   }
   case Builtin::BI__readfsdword: {
+    llvm::Type *IntTy = ConvertType(E->getType());
     Value *IntToPtr =
       Builder.CreateIntToPtr(EmitScalarExpr(E->getArg(0)),
-                             llvm::PointerType::get(CGM.Int32Ty, 257));
+                             llvm::PointerType::get(IntTy, 257));
     LoadInst *Load =
-        Builder.CreateAlignedLoad(IntToPtr, /*Align=*/4, /*isVolatile=*/true);
+        Builder.CreateDefaultAlignedLoad(IntToPtr, /*isVolatile=*/true);
     return RValue::get(Load);
   }
 
@@ -1963,6 +2056,322 @@ RValue CodeGenFunction::EmitBuiltinExpr(const FunctionDecl *FD,
       return RValue::get(llvm::ConstantExpr::getBitCast(GV, CGM.Int8PtrTy));
     break;
   }
+
+  // OpenCL v2.0 s6.13.16.2, Built-in pipe read and write functions
+  case Builtin::BIread_pipe:
+  case Builtin::BIwrite_pipe: {
+    Value *Arg0 = EmitScalarExpr(E->getArg(0)),
+          *Arg1 = EmitScalarExpr(E->getArg(1));
+
+    // Type of the generic packet parameter.
+    unsigned GenericAS =
+        getContext().getTargetAddressSpace(LangAS::opencl_generic);
+    llvm::Type *I8PTy = llvm::PointerType::get(
+        llvm::Type::getInt8Ty(getLLVMContext()), GenericAS);
+
+    // Testing which overloaded version we should generate the call for.
+    if (2U == E->getNumArgs()) {
+      const char *Name = (BuiltinID == Builtin::BIread_pipe) ? "__read_pipe_2"
+                                                             : "__write_pipe_2";
+      // Creating a generic function type to be able to call with any builtin or
+      // user defined type.
+      llvm::Type *ArgTys[] = {Arg0->getType(), I8PTy};
+      llvm::FunctionType *FTy = llvm::FunctionType::get(
+          Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
+      Value *BCast = Builder.CreatePointerCast(Arg1, I8PTy);
+      return RValue::get(Builder.CreateCall(
+          CGM.CreateRuntimeFunction(FTy, Name), {Arg0, BCast}));
+    } else {
+      assert(4 == E->getNumArgs() &&
+             "Illegal number of parameters to pipe function");
+      const char *Name = (BuiltinID == Builtin::BIread_pipe) ? "__read_pipe_4"
+                                                             : "__write_pipe_4";
+
+      llvm::Type *ArgTys[] = {Arg0->getType(), Arg1->getType(), Int32Ty, I8PTy};
+      Value *Arg2 = EmitScalarExpr(E->getArg(2)),
+            *Arg3 = EmitScalarExpr(E->getArg(3));
+      llvm::FunctionType *FTy = llvm::FunctionType::get(
+          Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
+      Value *BCast = Builder.CreatePointerCast(Arg3, I8PTy);
+      // We know the third argument is an integer type, but we may need to cast
+      // it to i32.
+      if (Arg2->getType() != Int32Ty)
+        Arg2 = Builder.CreateZExtOrTrunc(Arg2, Int32Ty);
+      return RValue::get(Builder.CreateCall(
+          CGM.CreateRuntimeFunction(FTy, Name), {Arg0, Arg1, Arg2, BCast}));
+    }
+  }
+  // OpenCL v2.0 s6.13.16 ,s9.17.3.5 - Built-in pipe reserve read and write
+  // functions
+  case Builtin::BIreserve_read_pipe:
+  case Builtin::BIreserve_write_pipe:
+  case Builtin::BIwork_group_reserve_read_pipe:
+  case Builtin::BIwork_group_reserve_write_pipe:
+  case Builtin::BIsub_group_reserve_read_pipe:
+  case Builtin::BIsub_group_reserve_write_pipe: {
+    // Composing the mangled name for the function.
+    const char *Name;
+    if (BuiltinID == Builtin::BIreserve_read_pipe)
+      Name = "__reserve_read_pipe";
+    else if (BuiltinID == Builtin::BIreserve_write_pipe)
+      Name = "__reserve_write_pipe";
+    else if (BuiltinID == Builtin::BIwork_group_reserve_read_pipe)
+      Name = "__work_group_reserve_read_pipe";
+    else if (BuiltinID == Builtin::BIwork_group_reserve_write_pipe)
+      Name = "__work_group_reserve_write_pipe";
+    else if (BuiltinID == Builtin::BIsub_group_reserve_read_pipe)
+      Name = "__sub_group_reserve_read_pipe";
+    else
+      Name = "__sub_group_reserve_write_pipe";
+
+    Value *Arg0 = EmitScalarExpr(E->getArg(0)),
+          *Arg1 = EmitScalarExpr(E->getArg(1));
+    llvm::Type *ReservedIDTy = ConvertType(getContext().OCLReserveIDTy);
+
+    // Building the generic function prototype.
+    llvm::Type *ArgTys[] = {Arg0->getType(), Int32Ty};
+    llvm::FunctionType *FTy = llvm::FunctionType::get(
+        ReservedIDTy, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
+    // We know the second argument is an integer type, but we may need to cast
+    // it to i32.
+    if (Arg1->getType() != Int32Ty)
+      Arg1 = Builder.CreateZExtOrTrunc(Arg1, Int32Ty);
+    return RValue::get(
+        Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name), {Arg0, Arg1}));
+  }
+  // OpenCL v2.0 s6.13.16, s9.17.3.5 - Built-in pipe commit read and write
+  // functions
+  case Builtin::BIcommit_read_pipe:
+  case Builtin::BIcommit_write_pipe:
+  case Builtin::BIwork_group_commit_read_pipe:
+  case Builtin::BIwork_group_commit_write_pipe:
+  case Builtin::BIsub_group_commit_read_pipe:
+  case Builtin::BIsub_group_commit_write_pipe: {
+    const char *Name;
+    if (BuiltinID == Builtin::BIcommit_read_pipe)
+      Name = "__commit_read_pipe";
+    else if (BuiltinID == Builtin::BIcommit_write_pipe)
+      Name = "__commit_write_pipe";
+    else if (BuiltinID == Builtin::BIwork_group_commit_read_pipe)
+      Name = "__work_group_commit_read_pipe";
+    else if (BuiltinID == Builtin::BIwork_group_commit_write_pipe)
+      Name = "__work_group_commit_write_pipe";
+    else if (BuiltinID == Builtin::BIsub_group_commit_read_pipe)
+      Name = "__sub_group_commit_read_pipe";
+    else
+      Name = "__sub_group_commit_write_pipe";
+
+    Value *Arg0 = EmitScalarExpr(E->getArg(0)),
+          *Arg1 = EmitScalarExpr(E->getArg(1));
+
+    // Building the generic function prototype.
+    llvm::Type *ArgTys[] = {Arg0->getType(), Arg1->getType()};
+    llvm::FunctionType *FTy =
+        llvm::FunctionType::get(llvm::Type::getVoidTy(getLLVMContext()),
+                                llvm::ArrayRef<llvm::Type *>(ArgTys), false);
+
+    return RValue::get(
+        Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name), {Arg0, Arg1}));
+  }
+  // OpenCL v2.0 s6.13.16.4 Built-in pipe query functions
+  case Builtin::BIget_pipe_num_packets:
+  case Builtin::BIget_pipe_max_packets: {
+    const char *Name;
+    if (BuiltinID == Builtin::BIget_pipe_num_packets)
+      Name = "__get_pipe_num_packets";
+    else
+      Name = "__get_pipe_max_packets";
+
+    // Building the generic function prototype.
+    Value *Arg0 = EmitScalarExpr(E->getArg(0));
+    llvm::Type *ArgTys[] = {Arg0->getType()};
+    llvm::FunctionType *FTy = llvm::FunctionType::get(
+        Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
+
+    return RValue::get(
+        Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name), {Arg0}));
+  }
+
+  // OpenCL v2.0 s6.13.9 - Address space qualifier functions.
+  case Builtin::BIto_global:
+  case Builtin::BIto_local:
+  case Builtin::BIto_private: {
+    auto Arg0 = EmitScalarExpr(E->getArg(0));
+    auto NewArgT = llvm::PointerType::get(Int8Ty,
+      CGM.getContext().getTargetAddressSpace(LangAS::opencl_generic));
+    auto NewRetT = llvm::PointerType::get(Int8Ty,
+      CGM.getContext().getTargetAddressSpace(
+        E->getType()->getPointeeType().getAddressSpace()));
+    auto FTy = llvm::FunctionType::get(NewRetT, {NewArgT}, false);
+    llvm::Value *NewArg;
+    if (Arg0->getType()->getPointerAddressSpace() !=
+        NewArgT->getPointerAddressSpace())
+      NewArg = Builder.CreateAddrSpaceCast(Arg0, NewArgT);
+    else
+      NewArg = Builder.CreateBitOrPointerCast(Arg0, NewArgT);
+    auto NewCall = Builder.CreateCall(CGM.CreateRuntimeFunction(FTy,
+      E->getDirectCallee()->getName()), {NewArg});
+    return RValue::get(Builder.CreateBitOrPointerCast(NewCall,
+      ConvertType(E->getType())));
+  }
+
+  // OpenCL v2.0, s6.13.17 - Enqueue kernel function.
+  // It contains four different overload formats specified in Table 6.13.17.1.
+  case Builtin::BIenqueue_kernel: {
+    StringRef Name; // Generated function call name
+    unsigned NumArgs = E->getNumArgs();
+
+    llvm::Type *QueueTy = ConvertType(getContext().OCLQueueTy);
+    llvm::Type *RangeTy = ConvertType(getContext().OCLNDRangeTy);
+
+    llvm::Value *Queue = EmitScalarExpr(E->getArg(0));
+    llvm::Value *Flags = EmitScalarExpr(E->getArg(1));
+    llvm::Value *Range = EmitScalarExpr(E->getArg(2));
+
+    if (NumArgs == 4) {
+      // The most basic form of the call with parameters:
+      // queue_t, kernel_enqueue_flags_t, ndrange_t, block(void)
+      Name = "__enqueue_kernel_basic";
+      llvm::Type *ArgTys[] = {QueueTy, Int32Ty, RangeTy, Int8PtrTy};
+      llvm::FunctionType *FTy = llvm::FunctionType::get(
+          Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys, 4), false);
+
+      llvm::Value *Block =
+          Builder.CreateBitCast(EmitScalarExpr(E->getArg(3)), Int8PtrTy);
+
+      return RValue::get(Builder.CreateCall(
+          CGM.CreateRuntimeFunction(FTy, Name), {Queue, Flags, Range, Block}));
+    }
+    assert(NumArgs >= 5 && "Invalid enqueue_kernel signature");
+
+    // Could have events and/or vaargs.
+    if (E->getArg(3)->getType()->isBlockPointerType()) {
+      // No events passed, but has variadic arguments.
+      Name = "__enqueue_kernel_vaargs";
+      llvm::Value *Block =
+          Builder.CreateBitCast(EmitScalarExpr(E->getArg(3)), Int8PtrTy);
+      // Create a vector of the arguments, as well as a constant value to
+      // express to the runtime the number of variadic arguments.
+      std::vector<llvm::Value *> Args = {Queue, Flags, Range, Block,
+                                         ConstantInt::get(IntTy, NumArgs - 4)};
+      std::vector<llvm::Type *> ArgTys = {QueueTy, IntTy, RangeTy, Int8PtrTy,
+                                          IntTy};
+
+      // Add the variadics.
+      for (unsigned I = 4; I < NumArgs; ++I) {
+        llvm::Value *ArgSize = EmitScalarExpr(E->getArg(I));
+        unsigned TypeSizeInBytes =
+            getContext()
+                .getTypeSizeInChars(E->getArg(I)->getType())
+                .getQuantity();
+        Args.push_back(TypeSizeInBytes < 4
+                           ? Builder.CreateZExt(ArgSize, Int32Ty)
+                           : ArgSize);
+      }
+
+      llvm::FunctionType *FTy = llvm::FunctionType::get(
+          Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), true);
+      return RValue::get(
+          Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name),
+                             llvm::ArrayRef<llvm::Value *>(Args)));
+    }
+    // Any calls now have event arguments passed.
+    if (NumArgs >= 7) {
+      llvm::Type *EventTy = ConvertType(getContext().OCLClkEventTy);
+      unsigned AS4 =
+          E->getArg(4)->getType()->isArrayType()
+              ? E->getArg(4)->getType().getAddressSpace()
+              : E->getArg(4)->getType()->getPointeeType().getAddressSpace();
+      llvm::Type *EventPtrAS4Ty =
+          EventTy->getPointerTo(CGM.getContext().getTargetAddressSpace(AS4));
+      unsigned AS5 =
+          E->getArg(5)->getType()->getPointeeType().getAddressSpace();
+      llvm::Type *EventPtrAS5Ty =
+          EventTy->getPointerTo(CGM.getContext().getTargetAddressSpace(AS5));
+
+      llvm::Value *NumEvents = EmitScalarExpr(E->getArg(3));
+      llvm::Value *EventList =
+          E->getArg(4)->getType()->isArrayType()
+              ? EmitArrayToPointerDecay(E->getArg(4)).getPointer()
+              : EmitScalarExpr(E->getArg(4));
+      llvm::Value *ClkEvent = EmitScalarExpr(E->getArg(5));
+      llvm::Value *Block =
+          Builder.CreateBitCast(EmitScalarExpr(E->getArg(6)), Int8PtrTy);
+
+      std::vector<llvm::Type *> ArgTys = {
+          QueueTy,       Int32Ty,       RangeTy,  Int32Ty,
+          EventPtrAS4Ty, EventPtrAS5Ty, Int8PtrTy};
+      std::vector<llvm::Value *> Args = {Queue,     Flags,    Range, NumEvents,
+                                         EventList, ClkEvent, Block};
+
+      if (NumArgs == 7) {
+        // Has events but no variadics.
+        Name = "__enqueue_kernel_basic_events";
+        llvm::FunctionType *FTy = llvm::FunctionType::get(
+            Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), false);
+        return RValue::get(
+            Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name),
+                               llvm::ArrayRef<llvm::Value *>(Args)));
+      }
+      // Has event info and variadics
+      // Pass the number of variadics to the runtime function too.
+      Args.push_back(ConstantInt::get(Int32Ty, NumArgs - 7));
+      ArgTys.push_back(Int32Ty);
+      Name = "__enqueue_kernel_events_vaargs";
+
+      // Add the variadics.
+      for (unsigned I = 7; I < NumArgs; ++I) {
+        llvm::Value *ArgSize = EmitScalarExpr(E->getArg(I));
+        unsigned TypeSizeInBytes =
+            getContext()
+                .getTypeSizeInChars(E->getArg(I)->getType())
+                .getQuantity();
+        Args.push_back(TypeSizeInBytes < 4
+                           ? Builder.CreateZExt(ArgSize, Int32Ty)
+                           : ArgSize);
+      }
+      llvm::FunctionType *FTy = llvm::FunctionType::get(
+          Int32Ty, llvm::ArrayRef<llvm::Type *>(ArgTys), true);
+      return RValue::get(
+          Builder.CreateCall(CGM.CreateRuntimeFunction(FTy, Name),
+                             llvm::ArrayRef<llvm::Value *>(Args)));
+    }
+  }
+  // OpenCL v2.0 s6.13.17.6 - Kernel query functions need bitcast of block
+  // parameter.
+  case Builtin::BIget_kernel_work_group_size: {
+    Value *Arg = EmitScalarExpr(E->getArg(0));
+    Arg = Builder.CreateBitCast(Arg, Int8PtrTy);
+    return RValue::get(
+        Builder.CreateCall(CGM.CreateRuntimeFunction(
+                               llvm::FunctionType::get(IntTy, Int8PtrTy, false),
+                               "__get_kernel_work_group_size_impl"),
+                           Arg));
+  }
+  case Builtin::BIget_kernel_preferred_work_group_size_multiple: {
+    Value *Arg = EmitScalarExpr(E->getArg(0));
+    Arg = Builder.CreateBitCast(Arg, Int8PtrTy);
+    return RValue::get(Builder.CreateCall(
+        CGM.CreateRuntimeFunction(
+            llvm::FunctionType::get(IntTy, Int8PtrTy, false),
+            "__get_kernel_preferred_work_group_multiple_impl"),
+        Arg));
+  }
+  case Builtin::BIprintf:
+    if (getLangOpts().CUDA && getLangOpts().CUDAIsDevice)
+      return EmitCUDADevicePrintfCallExpr(E, ReturnValue);
+    break;
+  case Builtin::BI__builtin_canonicalize:
+  case Builtin::BI__builtin_canonicalizef:
+  case Builtin::BI__builtin_canonicalizel:
+    return RValue::get(emitUnaryBuiltin(*this, E, Intrinsic::canonicalize));
+
+  case Builtin::BI__builtin_thread_pointer: {
+    if (!getContext().getTargetInfo().isTLSSupported())
+      CGM.ErrorUnsupported(E, "__builtin_thread_pointer");
+    // Fall through - it's already mapped to the intrinsic by GCCBuiltin.
+    break;
+  }
   }
 
   // If this is an alias for a lib function (e.g. __builtin_sin), emit
@@ -2155,7 +2564,7 @@ static llvm::VectorType *GetFloatNeonType(CodeGenFunction *CGF,
 }
 
 Value *CodeGenFunction::EmitNeonSplat(Value *V, Constant *C) {
-  unsigned nElts = cast<llvm::VectorType>(V->getType())->getNumElements();
+  unsigned nElts = V->getType()->getVectorNumElements();
   Value* SV = llvm::ConstantVector::getSplat(nElts, C);
   return Builder.CreateShuffleVector(V, V, SV, "lane");
 }
@@ -3073,14 +3482,13 @@ Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(
   case NEON::BI__builtin_neon_vext_v:
   case NEON::BI__builtin_neon_vextq_v: {
     int CV = cast<ConstantInt>(Ops[2])->getSExtValue();
-    SmallVector<Constant*, 16> Indices;
+    SmallVector<uint32_t, 16> Indices;
     for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i)
-      Indices.push_back(ConstantInt::get(Int32Ty, i+CV));
+      Indices.push_back(i+CV);
 
     Ops[0] = Builder.CreateBitCast(Ops[0], Ty);
     Ops[1] = Builder.CreateBitCast(Ops[1], Ty);
-    Value *SV = llvm::ConstantVector::get(Indices);
-    return Builder.CreateShuffleVector(Ops[0], Ops[1], SV, "vext");
+    return Builder.CreateShuffleVector(Ops[0], Ops[1], Indices, "vext");
   }
   case NEON::BI__builtin_neon_vfma_v:
   case NEON::BI__builtin_neon_vfmaq_v: {
@@ -3278,14 +3686,13 @@ Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(
     Value *SV = nullptr;
 
     for (unsigned vi = 0; vi != 2; ++vi) {
-      SmallVector<Constant*, 16> Indices;
+      SmallVector<uint32_t, 16> Indices;
       for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
-        Indices.push_back(Builder.getInt32(i+vi));
-        Indices.push_back(Builder.getInt32(i+e+vi));
+        Indices.push_back(i+vi);
+        Indices.push_back(i+e+vi);
       }
       Value *Addr = Builder.CreateConstInBoundsGEP1_32(Ty, Ops[0], vi);
-      SV = llvm::ConstantVector::get(Indices);
-      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], SV, "vtrn");
+      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], Indices, "vtrn");
       SV = Builder.CreateDefaultAlignedStore(SV, Addr);
     }
     return SV;
@@ -3307,13 +3714,12 @@ Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(
     Value *SV = nullptr;
 
     for (unsigned vi = 0; vi != 2; ++vi) {
-      SmallVector<Constant*, 16> Indices;
+      SmallVector<uint32_t, 16> Indices;
       for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i)
-        Indices.push_back(ConstantInt::get(Int32Ty, 2*i+vi));
+        Indices.push_back(2*i+vi);
 
       Value *Addr = Builder.CreateConstInBoundsGEP1_32(Ty, Ops[0], vi);
-      SV = llvm::ConstantVector::get(Indices);
-      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], SV, "vuzp");
+      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], Indices, "vuzp");
       SV = Builder.CreateDefaultAlignedStore(SV, Addr);
     }
     return SV;
@@ -3326,14 +3732,13 @@ Value *CodeGenFunction::EmitCommonNeonBuiltinExpr(
     Value *SV = nullptr;
 
     for (unsigned vi = 0; vi != 2; ++vi) {
-      SmallVector<Constant*, 16> Indices;
+      SmallVector<uint32_t, 16> Indices;
       for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
-        Indices.push_back(ConstantInt::get(Int32Ty, (i + vi*e) >> 1));
-        Indices.push_back(ConstantInt::get(Int32Ty, ((i + vi*e) >> 1)+e));
+        Indices.push_back((i + vi*e) >> 1);
+        Indices.push_back(((i + vi*e) >> 1)+e);
       }
       Value *Addr = Builder.CreateConstInBoundsGEP1_32(Ty, Ops[0], vi);
-      SV = llvm::ConstantVector::get(Indices);
-      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], SV, "vzip");
+      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], Indices, "vzip");
       SV = Builder.CreateDefaultAlignedStore(SV, Addr);
     }
     return SV;
@@ -3381,19 +3786,19 @@ static Value *packTBLDVectorList(CodeGenFunction &CGF, ArrayRef<Value *> Ops,
   if (ExtOp)
     TblOps.push_back(ExtOp);
 
-  // Build a vector containing sequential number like (0, 1, 2, ..., 15)  
-  SmallVector<Constant*, 16> Indices;
+  // Build a vector containing sequential number like (0, 1, 2, ..., 15)
+  SmallVector<uint32_t, 16> Indices;
   llvm::VectorType *TblTy = cast<llvm::VectorType>(Ops[0]->getType());
   for (unsigned i = 0, e = TblTy->getNumElements(); i != e; ++i) {
-    Indices.push_back(ConstantInt::get(CGF.Int32Ty, 2*i));
-    Indices.push_back(ConstantInt::get(CGF.Int32Ty, 2*i+1));
+    Indices.push_back(2*i);
+    Indices.push_back(2*i+1);
   }
-  Value *SV = llvm::ConstantVector::get(Indices);
 
   int PairPos = 0, End = Ops.size() - 1;
   while (PairPos < End) {
     TblOps.push_back(CGF.Builder.CreateShuffleVector(Ops[PairPos],
-                                                     Ops[PairPos+1], SV, Name));
+                                                     Ops[PairPos+1], Indices,
+                                                     Name));
     PairPos += 2;
   }
 
@@ -3402,13 +3807,13 @@ static Value *packTBLDVectorList(CodeGenFunction &CGF, ArrayRef<Value *> Ops,
   if (PairPos == End) {
     Value *ZeroTbl = ConstantAggregateZero::get(TblTy);
     TblOps.push_back(CGF.Builder.CreateShuffleVector(Ops[PairPos],
-                                                     ZeroTbl, SV, Name));
+                                                     ZeroTbl, Indices, Name));
   }
 
   Function *TblF;
   TblOps.push_back(IndexOp);
   TblF = CGF.CGM.getIntrinsic(IntID, ResTy);
-  
+
   return CGF.EmitNeonCall(TblF, TblOps, Name);
 }
 
@@ -3452,7 +3857,9 @@ Value *CodeGenFunction::GetValueForARMHint(unsigned BuiltinID) {
 static Value *EmitSpecialRegisterBuiltin(CodeGenFunction &CGF,
                                          const CallExpr *E,
                                          llvm::Type *RegisterType,
-                                         llvm::Type *ValueType, bool IsRead) {
+                                         llvm::Type *ValueType,
+                                         bool IsRead,
+                                         StringRef SysReg = "") {
   // write and register intrinsics only support 32 and 64 bit operations.
   assert((RegisterType->isIntegerTy(32) || RegisterType->isIntegerTy(64))
           && "Unsupported size for register.");
@@ -3461,8 +3868,10 @@ static Value *EmitSpecialRegisterBuiltin(CodeGenFunction &CGF,
   CodeGen::CodeGenModule &CGM = CGF.CGM;
   LLVMContext &Context = CGM.getLLVMContext();
 
-  const Expr *SysRegStrExpr = E->getArg(0)->IgnoreParenCasts();
-  StringRef SysReg = cast<StringLiteral>(SysRegStrExpr)->getString();
+  if (SysReg.empty()) {
+    const Expr *SysRegStrExpr = E->getArg(0)->IgnoreParenCasts();
+    SysReg = cast<StringLiteral>(SysRegStrExpr)->getString();
+  }
 
   llvm::Metadata *Ops[] = { llvm::MDString::get(Context, SysReg) };
   llvm::MDNode *RegName = llvm::MDNode::get(Context, Ops);
@@ -3602,6 +4011,74 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
     return EmitNounwindRuntimeCall(CGM.CreateRuntimeFunction(FTy, Name), Ops);
   }
 
+  if (BuiltinID == ARM::BI__builtin_arm_mcrr ||
+      BuiltinID == ARM::BI__builtin_arm_mcrr2) {
+    Function *F;
+
+    switch (BuiltinID) {
+    default: llvm_unreachable("unexpected builtin");
+    case ARM::BI__builtin_arm_mcrr:
+      F = CGM.getIntrinsic(Intrinsic::arm_mcrr);
+      break;
+    case ARM::BI__builtin_arm_mcrr2:
+      F = CGM.getIntrinsic(Intrinsic::arm_mcrr2);
+      break;
+    }
+
+    // MCRR{2} instruction has 5 operands but
+    // the intrinsic has 4 because Rt and Rt2
+    // are represented as a single unsigned 64
+    // bit integer in the intrinsic definition
+    // but internally it's represented as 2 32
+    // bit integers.
+
+    Value *Coproc = EmitScalarExpr(E->getArg(0));
+    Value *Opc1 = EmitScalarExpr(E->getArg(1));
+    Value *RtAndRt2 = EmitScalarExpr(E->getArg(2));
+    Value *CRm = EmitScalarExpr(E->getArg(3));
+
+    Value *C1 = llvm::ConstantInt::get(Int64Ty, 32);
+    Value *Rt = Builder.CreateTruncOrBitCast(RtAndRt2, Int32Ty);
+    Value *Rt2 = Builder.CreateLShr(RtAndRt2, C1);
+    Rt2 = Builder.CreateTruncOrBitCast(Rt2, Int32Ty);
+
+    return Builder.CreateCall(F, {Coproc, Opc1, Rt, Rt2, CRm});
+  }
+
+  if (BuiltinID == ARM::BI__builtin_arm_mrrc ||
+      BuiltinID == ARM::BI__builtin_arm_mrrc2) {
+    Function *F;
+
+    switch (BuiltinID) {
+    default: llvm_unreachable("unexpected builtin");
+    case ARM::BI__builtin_arm_mrrc:
+      F = CGM.getIntrinsic(Intrinsic::arm_mrrc);
+      break;
+    case ARM::BI__builtin_arm_mrrc2:
+      F = CGM.getIntrinsic(Intrinsic::arm_mrrc2);
+      break;
+    }
+
+    Value *Coproc = EmitScalarExpr(E->getArg(0));
+    Value *Opc1 = EmitScalarExpr(E->getArg(1));
+    Value *CRm  = EmitScalarExpr(E->getArg(2));
+    Value *RtAndRt2 = Builder.CreateCall(F, {Coproc, Opc1, CRm});
+
+    // Returns an unsigned 64 bit integer, represented
+    // as two 32 bit integers.
+
+    Value *Rt = Builder.CreateExtractValue(RtAndRt2, 1);
+    Value *Rt1 = Builder.CreateExtractValue(RtAndRt2, 0);
+    Rt = Builder.CreateZExt(Rt, Int64Ty);
+    Rt1 = Builder.CreateZExt(Rt1, Int64Ty);
+
+    Value *ShiftCast = llvm::ConstantInt::get(Int64Ty, 32);
+    RtAndRt2 = Builder.CreateShl(Rt, ShiftCast, "shl", true);
+    RtAndRt2 = Builder.CreateOr(RtAndRt2, Rt1);
+
+    return Builder.CreateBitCast(RtAndRt2, ConvertType(E->getType()));
+  }
+
   if (BuiltinID == ARM::BI__builtin_arm_ldrexd ||
       ((BuiltinID == ARM::BI__builtin_arm_ldrex ||
         BuiltinID == ARM::BI__builtin_arm_ldaex) &&
@@ -3914,7 +4391,7 @@ Value *CodeGenFunction::EmitARMBuiltinExpr(unsigned BuiltinID,
   // the first argument, but the LLVM intrinsic expects it as the third one.
   case ARM::BI_MoveToCoprocessor:
   case ARM::BI_MoveToCoprocessor2: {
-    Function *F = CGM.getIntrinsic(BuiltinID == ARM::BI_MoveToCoprocessor ? 
+    Function *F = CGM.getIntrinsic(BuiltinID == ARM::BI_MoveToCoprocessor ?
                                    Intrinsic::arm_mcr : Intrinsic::arm_mcr2);
     return Builder.CreateCall(F, {Ops[1], Ops[2], Ops[0],
                                   Ops[3], Ops[4], Ops[5]});
@@ -4478,11 +4955,6 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     return Builder.CreateCall(F);
   }
 
-  if (BuiltinID == AArch64::BI__builtin_thread_pointer) {
-    Function *F = CGM.getIntrinsic(Intrinsic::aarch64_thread_pointer);
-    return Builder.CreateCall(F);
-  }
-
   // CRC32
   Intrinsic::ID CRCIntrinsicID = Intrinsic::not_intrinsic;
   switch (BuiltinID) {
@@ -5150,22 +5622,6 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     Ops[2] = Builder.CreateExtractElement(Ops[2], Ops[3], "extract");
     return Builder.CreateCall(F, {Ops[1], Ops[2], Ops[0]});
   }
-  case NEON::BI__builtin_neon_vfms_v:
-  case NEON::BI__builtin_neon_vfmsq_v: {  // Only used for FP types
-    // FIXME: probably remove when we no longer support aarch64_simd.h
-    // (arm_neon.h delegates to vfma).
-
-    // The ARM builtins (and instructions) have the addend as the first
-    // operand, but the 'fma' intrinsics have it last. Swap it around here.
-    Value *Subtrahend = Ops[0];
-    Value *Multiplicand = Ops[2];
-    Ops[0] = Multiplicand;
-    Ops[2] = Subtrahend;
-    Ops[1] = Builder.CreateBitCast(Ops[1], VTy);
-    Ops[1] = Builder.CreateFNeg(Ops[1]);
-    Int = Intrinsic::fma;
-    return EmitNeonCall(CGM.getIntrinsic(Int, Ty), Ops, "fmls");
-  }
   case NEON::BI__builtin_neon_vmull_v:
     // FIXME: improve sharing scheme to cope with 3 alternative LLVM intrinsics.
     Int = usgn ? Intrinsic::aarch64_neon_umull : Intrinsic::aarch64_neon_smull;
@@ -5988,14 +6444,13 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     Value *SV = nullptr;
 
     for (unsigned vi = 0; vi != 2; ++vi) {
-      SmallVector<Constant*, 16> Indices;
+      SmallVector<uint32_t, 16> Indices;
       for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
-        Indices.push_back(ConstantInt::get(Int32Ty, i+vi));
-        Indices.push_back(ConstantInt::get(Int32Ty, i+e+vi));
+        Indices.push_back(i+vi);
+        Indices.push_back(i+e+vi);
       }
       Value *Addr = Builder.CreateConstInBoundsGEP1_32(Ty, Ops[0], vi);
-      SV = llvm::ConstantVector::get(Indices);
-      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], SV, "vtrn");
+      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], Indices, "vtrn");
       SV = Builder.CreateDefaultAlignedStore(SV, Addr);
     }
     return SV;
@@ -6008,13 +6463,12 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     Value *SV = nullptr;
 
     for (unsigned vi = 0; vi != 2; ++vi) {
-      SmallVector<Constant*, 16> Indices;
+      SmallVector<uint32_t, 16> Indices;
       for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i)
-        Indices.push_back(ConstantInt::get(Int32Ty, 2*i+vi));
+        Indices.push_back(2*i+vi);
 
       Value *Addr = Builder.CreateConstInBoundsGEP1_32(Ty, Ops[0], vi);
-      SV = llvm::ConstantVector::get(Indices);
-      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], SV, "vuzp");
+      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], Indices, "vuzp");
       SV = Builder.CreateDefaultAlignedStore(SV, Addr);
     }
     return SV;
@@ -6027,14 +6481,13 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
     Value *SV = nullptr;
 
     for (unsigned vi = 0; vi != 2; ++vi) {
-      SmallVector<Constant*, 16> Indices;
+      SmallVector<uint32_t, 16> Indices;
       for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
-        Indices.push_back(ConstantInt::get(Int32Ty, (i + vi*e) >> 1));
-        Indices.push_back(ConstantInt::get(Int32Ty, ((i + vi*e) >> 1)+e));
+        Indices.push_back((i + vi*e) >> 1);
+        Indices.push_back(((i + vi*e) >> 1)+e);
       }
       Value *Addr = Builder.CreateConstInBoundsGEP1_32(Ty, Ops[0], vi);
-      SV = llvm::ConstantVector::get(Indices);
-      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], SV, "vzip");
+      SV = Builder.CreateShuffleVector(Ops[1], Ops[2], Indices, "vzip");
       SV = Builder.CreateDefaultAlignedStore(SV, Addr);
     }
     return SV;
@@ -6110,6 +6563,118 @@ BuildVector(ArrayRef<llvm::Value*> Ops) {
   return Result;
 }
 
+// Convert the mask from an integer type to a vector of i1.
+static Value *getMaskVecValue(CodeGenFunction &CGF, Value *Mask,
+                              unsigned NumElts) {
+
+  llvm::VectorType *MaskTy = llvm::VectorType::get(CGF.Builder.getInt1Ty(),
+                         cast<IntegerType>(Mask->getType())->getBitWidth());
+  Value *MaskVec = CGF.Builder.CreateBitCast(Mask, MaskTy);
+
+  // If we have less than 8 elements, then the starting mask was an i8 and
+  // we need to extract down to the right number of elements.
+  if (NumElts < 8) {
+    uint32_t Indices[4];
+    for (unsigned i = 0; i != NumElts; ++i)
+      Indices[i] = i;
+    MaskVec = CGF.Builder.CreateShuffleVector(MaskVec, MaskVec,
+                                             makeArrayRef(Indices, NumElts),
+                                             "extract");
+  }
+  return MaskVec;
+}
+
+static Value *EmitX86MaskedStore(CodeGenFunction &CGF,
+                                 SmallVectorImpl<Value *> &Ops,
+                                 unsigned Align) {
+  // Cast the pointer to right type.
+  Ops[0] = CGF.Builder.CreateBitCast(Ops[0],
+                               llvm::PointerType::getUnqual(Ops[1]->getType()));
+
+  // If the mask is all ones just emit a regular store.
+  if (const auto *C = dyn_cast<Constant>(Ops[2]))
+    if (C->isAllOnesValue())
+      return CGF.Builder.CreateAlignedStore(Ops[1], Ops[0], Align);
+
+  Value *MaskVec = getMaskVecValue(CGF, Ops[2],
+                                   Ops[1]->getType()->getVectorNumElements());
+
+  return CGF.Builder.CreateMaskedStore(Ops[1], Ops[0], Align, MaskVec);
+}
+
+static Value *EmitX86MaskedLoad(CodeGenFunction &CGF,
+                                SmallVectorImpl<Value *> &Ops, unsigned Align) {
+  // Cast the pointer to right type.
+  Ops[0] = CGF.Builder.CreateBitCast(Ops[0],
+                               llvm::PointerType::getUnqual(Ops[1]->getType()));
+
+  // If the mask is all ones just emit a regular store.
+  if (const auto *C = dyn_cast<Constant>(Ops[2]))
+    if (C->isAllOnesValue())
+      return CGF.Builder.CreateAlignedLoad(Ops[0], Align);
+
+  Value *MaskVec = getMaskVecValue(CGF, Ops[2],
+                                   Ops[1]->getType()->getVectorNumElements());
+
+  return CGF.Builder.CreateMaskedLoad(Ops[0], Align, MaskVec, Ops[1]);
+}
+
+static Value *EmitX86Select(CodeGenFunction &CGF,
+                            Value *Mask, Value *Op0, Value *Op1) {
+
+  // If the mask is all ones just return first argument.
+  if (const auto *C = dyn_cast<Constant>(Mask))
+    if (C->isAllOnesValue())
+      return Op0;
+
+  Mask = getMaskVecValue(CGF, Mask, Op0->getType()->getVectorNumElements());
+
+  return CGF.Builder.CreateSelect(Mask, Op0, Op1);
+}
+
+static Value *EmitX86MaskedCompare(CodeGenFunction &CGF, unsigned CC,
+                                   bool Signed, SmallVectorImpl<Value *> &Ops) {
+  unsigned NumElts = Ops[0]->getType()->getVectorNumElements();
+  Value *Cmp;
+
+  if (CC == 3) {
+    Cmp = Constant::getNullValue(
+                       llvm::VectorType::get(CGF.Builder.getInt1Ty(), NumElts));
+  } else if (CC == 7) {
+    Cmp = Constant::getAllOnesValue(
+                       llvm::VectorType::get(CGF.Builder.getInt1Ty(), NumElts));
+  } else {
+    ICmpInst::Predicate Pred;
+    switch (CC) {
+    default: llvm_unreachable("Unknown condition code");
+    case 0: Pred = ICmpInst::ICMP_EQ;  break;
+    case 1: Pred = Signed ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; break;
+    case 2: Pred = Signed ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE; break;
+    case 4: Pred = ICmpInst::ICMP_NE;  break;
+    case 5: Pred = Signed ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE; break;
+    case 6: Pred = Signed ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; break;
+    }
+    Cmp = CGF.Builder.CreateICmp(Pred, Ops[0], Ops[1]);
+  }
+
+  const auto *C = dyn_cast<Constant>(Ops.back());
+  if (!C || !C->isAllOnesValue())
+    Cmp = CGF.Builder.CreateAnd(Cmp, getMaskVecValue(CGF, Ops.back(), NumElts));
+
+  if (NumElts < 8) {
+    uint32_t Indices[8];
+    for (unsigned i = 0; i != NumElts; ++i)
+      Indices[i] = i;
+    for (unsigned i = NumElts; i != 8; ++i)
+      Indices[i] = i % NumElts + NumElts;
+    Cmp = CGF.Builder.CreateShuffleVector(
+        Cmp, llvm::Constant::getNullValue(Cmp->getType()), Indices);
+  }
+  return CGF.Builder.CreateBitCast(Cmp,
+                                   IntegerType::get(CGF.getLLVMContext(),
+                                                    std::max(NumElts, 8U)));
+}
+
 Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
                                            const CallExpr *E) {
   if (BuiltinID == X86::BI__builtin_ms_va_start ||
@@ -6160,6 +6725,31 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     Ops.push_back(llvm::ConstantInt::get(getLLVMContext(), Result));
   }
 
+  // These exist so that the builtin that takes an immediate can be bounds
+  // checked by clang to avoid passing bad immediates to the backend. Since
+  // AVX has a larger immediate than SSE we would need separate builtins to
+  // do the different bounds checking. Rather than create a clang specific
+  // SSE only builtin, this implements eight separate builtins to match gcc
+  // implementation.
+  auto getCmpIntrinsicCall = [this, &Ops](Intrinsic::ID ID, unsigned Imm) {
+    Ops.push_back(llvm::ConstantInt::get(Int8Ty, Imm));
+    llvm::Function *F = CGM.getIntrinsic(ID);
+    return Builder.CreateCall(F, Ops);
+  };
+
+  // For the vector forms of FP comparisons, translate the builtins directly to
+  // IR.
+  // TODO: The builtins could be removed if the SSE header files used vector
+  // extension comparisons directly (vector ordered/unordered may need
+  // additional support via __builtin_isnan()).
+  auto getVectorFCmpIR = [this, &Ops](CmpInst::Predicate Pred) {
+    Value *Cmp = Builder.CreateFCmp(Pred, Ops[0], Ops[1]);
+    llvm::VectorType *FPVecTy = cast<llvm::VectorType>(Ops[0]->getType());
+    llvm::VectorType *IntVecTy = llvm::VectorType::getInteger(FPVecTy);
+    Value *Sext = Builder.CreateSExt(Cmp, IntVecTy);
+    return Builder.CreateBitCast(Sext, FPVecTy);
+  };
+
   switch (BuiltinID) {
   default: return nullptr;
   case X86::BI__builtin_cpu_supports: {
@@ -6188,6 +6778,16 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
       AVX512F,
       BMI,
       BMI2,
+      AES,
+      PCLMUL,
+      AVX512VL,
+      AVX512BW,
+      AVX512DQ,
+      AVX512CD,
+      AVX512ER,
+      AVX512PF,
+      AVX512VBMI,
+      AVX512IFMA,
       MAX
     };
 
@@ -6198,6 +6798,7 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
                               .Case("sse", X86Features::SSE)
                               .Case("sse2", X86Features::SSE2)
                               .Case("sse3", X86Features::SSE3)
+                              .Case("ssse3", X86Features::SSSE3)
                               .Case("sse4.1", X86Features::SSE4_1)
                               .Case("sse4.2", X86Features::SSE4_2)
                               .Case("avx", X86Features::AVX)
@@ -6209,6 +6810,16 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
                               .Case("avx512f", X86Features::AVX512F)
                               .Case("bmi", X86Features::BMI)
                               .Case("bmi2", X86Features::BMI2)
+                              .Case("aes", X86Features::AES)
+                              .Case("pclmul", X86Features::PCLMUL)
+                              .Case("avx512vl", X86Features::AVX512VL)
+                              .Case("avx512bw", X86Features::AVX512BW)
+                              .Case("avx512dq", X86Features::AVX512DQ)
+                              .Case("avx512cd", X86Features::AVX512CD)
+                              .Case("avx512er", X86Features::AVX512ER)
+                              .Case("avx512pf", X86Features::AVX512PF)
+                              .Case("avx512vbmi", X86Features::AVX512VBMI)
+                              .Case("avx512ifma", X86Features::AVX512IFMA)
                               .Default(X86Features::MAX);
     assert(Feature != X86Features::MAX && "Invalid feature!");
 
@@ -6237,7 +6848,7 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
 
     // Check the value of the bit corresponding to the feature requested.
     Value *Bitset = Builder.CreateAnd(
-        Features, llvm::ConstantInt::get(Int32Ty, 1 << Feature));
+        Features, llvm::ConstantInt::get(Int32Ty, 1ULL << Feature));
     return Builder.CreateICmpNE(Bitset, llvm::ConstantInt::get(Int32Ty, 0));
   }
   case X86::BI_mm_prefetch: {
@@ -6312,6 +6923,78 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     Ops.push_back(Mlo);
     return Builder.CreateCall(CGM.getIntrinsic(ID), Ops);
   }
+  case X86::BI__builtin_ia32_storedqudi128_mask:
+  case X86::BI__builtin_ia32_storedqusi128_mask:
+  case X86::BI__builtin_ia32_storedquhi128_mask:
+  case X86::BI__builtin_ia32_storedquqi128_mask:
+  case X86::BI__builtin_ia32_storeupd128_mask:
+  case X86::BI__builtin_ia32_storeups128_mask:
+  case X86::BI__builtin_ia32_storedqudi256_mask:
+  case X86::BI__builtin_ia32_storedqusi256_mask:
+  case X86::BI__builtin_ia32_storedquhi256_mask:
+  case X86::BI__builtin_ia32_storedquqi256_mask:
+  case X86::BI__builtin_ia32_storeupd256_mask:
+  case X86::BI__builtin_ia32_storeups256_mask:
+  case X86::BI__builtin_ia32_storedqudi512_mask:
+  case X86::BI__builtin_ia32_storedqusi512_mask:
+  case X86::BI__builtin_ia32_storedquhi512_mask:
+  case X86::BI__builtin_ia32_storedquqi512_mask:
+  case X86::BI__builtin_ia32_storeupd512_mask:
+  case X86::BI__builtin_ia32_storeups512_mask:
+    return EmitX86MaskedStore(*this, Ops, 1);
+
+  case X86::BI__builtin_ia32_movdqa32store128_mask:
+  case X86::BI__builtin_ia32_movdqa64store128_mask:
+  case X86::BI__builtin_ia32_storeaps128_mask:
+  case X86::BI__builtin_ia32_storeapd128_mask:
+  case X86::BI__builtin_ia32_movdqa32store256_mask:
+  case X86::BI__builtin_ia32_movdqa64store256_mask:
+  case X86::BI__builtin_ia32_storeaps256_mask:
+  case X86::BI__builtin_ia32_storeapd256_mask:
+  case X86::BI__builtin_ia32_movdqa32store512_mask:
+  case X86::BI__builtin_ia32_movdqa64store512_mask:
+  case X86::BI__builtin_ia32_storeaps512_mask:
+  case X86::BI__builtin_ia32_storeapd512_mask: {
+    unsigned Align =
+      getContext().getTypeAlignInChars(E->getArg(1)->getType()).getQuantity();
+    return EmitX86MaskedStore(*this, Ops, Align);
+  }
+  case X86::BI__builtin_ia32_loadups128_mask:
+  case X86::BI__builtin_ia32_loadups256_mask:
+  case X86::BI__builtin_ia32_loadups512_mask:
+  case X86::BI__builtin_ia32_loadupd128_mask:
+  case X86::BI__builtin_ia32_loadupd256_mask:
+  case X86::BI__builtin_ia32_loadupd512_mask:
+  case X86::BI__builtin_ia32_loaddquqi128_mask:
+  case X86::BI__builtin_ia32_loaddquqi256_mask:
+  case X86::BI__builtin_ia32_loaddquqi512_mask:
+  case X86::BI__builtin_ia32_loaddquhi128_mask:
+  case X86::BI__builtin_ia32_loaddquhi256_mask:
+  case X86::BI__builtin_ia32_loaddquhi512_mask:
+  case X86::BI__builtin_ia32_loaddqusi128_mask:
+  case X86::BI__builtin_ia32_loaddqusi256_mask:
+  case X86::BI__builtin_ia32_loaddqusi512_mask:
+  case X86::BI__builtin_ia32_loaddqudi128_mask:
+  case X86::BI__builtin_ia32_loaddqudi256_mask:
+  case X86::BI__builtin_ia32_loaddqudi512_mask:
+    return EmitX86MaskedLoad(*this, Ops, 1);
+
+  case X86::BI__builtin_ia32_loadaps128_mask:
+  case X86::BI__builtin_ia32_loadaps256_mask:
+  case X86::BI__builtin_ia32_loadaps512_mask:
+  case X86::BI__builtin_ia32_loadapd128_mask:
+  case X86::BI__builtin_ia32_loadapd256_mask:
+  case X86::BI__builtin_ia32_loadapd512_mask:
+  case X86::BI__builtin_ia32_movdqa32load128_mask:
+  case X86::BI__builtin_ia32_movdqa32load256_mask:
+  case X86::BI__builtin_ia32_movdqa32load512_mask:
+  case X86::BI__builtin_ia32_movdqa64load128_mask:
+  case X86::BI__builtin_ia32_movdqa64load256_mask:
+  case X86::BI__builtin_ia32_movdqa64load512_mask: {
+    unsigned Align =
+      getContext().getTypeAlignInChars(E->getArg(1)->getType()).getQuantity();
+    return EmitX86MaskedLoad(*this, Ops, Align);
+  }
   case X86::BI__builtin_ia32_storehps:
   case X86::BI__builtin_ia32_storelps: {
     llvm::Type *PtrTy = llvm::PointerType::getUnqual(Int64Ty);
@@ -6330,103 +7013,50 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     return Builder.CreateDefaultAlignedStore(Ops[1], Ops[0]);
   }
   case X86::BI__builtin_ia32_palignr128:
-  case X86::BI__builtin_ia32_palignr256: {
+  case X86::BI__builtin_ia32_palignr256:
+  case X86::BI__builtin_ia32_palignr128_mask:
+  case X86::BI__builtin_ia32_palignr256_mask:
+  case X86::BI__builtin_ia32_palignr512_mask: {
     unsigned ShiftVal = cast<llvm::ConstantInt>(Ops[2])->getZExtValue();
 
-    unsigned NumElts =
-      cast<llvm::VectorType>(Ops[0]->getType())->getNumElements();
+    unsigned NumElts = Ops[0]->getType()->getVectorNumElements();
     assert(NumElts % 16 == 0);
-    unsigned NumLanes = NumElts / 16;
-    unsigned NumLaneElts = NumElts / NumLanes;
 
     // If palignr is shifting the pair of vectors more than the size of two
     // lanes, emit zero.
-    if (ShiftVal >= (2 * NumLaneElts))
+    if (ShiftVal >= 32)
       return llvm::Constant::getNullValue(ConvertType(E->getType()));
 
     // If palignr is shifting the pair of input vectors more than one lane,
     // but less than two lanes, convert to shifting in zeroes.
-    if (ShiftVal > NumLaneElts) {
-      ShiftVal -= NumLaneElts;
+    if (ShiftVal > 16) {
+      ShiftVal -= 16;
       Ops[1] = Ops[0];
       Ops[0] = llvm::Constant::getNullValue(Ops[0]->getType());
     }
 
-    uint32_t Indices[32];
+    uint32_t Indices[64];
     // 256-bit palignr operates on 128-bit lanes so we need to handle that
-    for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
-      for (unsigned i = 0; i != NumLaneElts; ++i) {
+    for (unsigned l = 0; l != NumElts; l += 16) {
+      for (unsigned i = 0; i != 16; ++i) {
         unsigned Idx = ShiftVal + i;
-        if (Idx >= NumLaneElts)
-          Idx += NumElts - NumLaneElts; // End of lane, switch operand.
+        if (Idx >= 16)
+          Idx += NumElts - 16; // End of lane, switch operand.
         Indices[l + i] = Idx + l;
       }
     }
 
-    Value *SV = llvm::ConstantDataVector::get(getLLVMContext(),
-                                              makeArrayRef(Indices, NumElts));
-    return Builder.CreateShuffleVector(Ops[1], Ops[0], SV, "palignr");
-  }
-  case X86::BI__builtin_ia32_pslldqi256: {
-    // Shift value is in bits so divide by 8.
-    unsigned shiftVal = cast<llvm::ConstantInt>(Ops[1])->getZExtValue() >> 3;
+    Value *Align = Builder.CreateShuffleVector(Ops[1], Ops[0],
+                                               makeArrayRef(Indices, NumElts),
+                                               "palignr");
 
-    // If pslldq is shifting the vector more than 15 bytes, emit zero.
-    if (shiftVal >= 16)
-      return llvm::Constant::getNullValue(ConvertType(E->getType()));
-
-    uint32_t Indices[32];
-    // 256-bit pslldq operates on 128-bit lanes so we need to handle that
-    for (unsigned l = 0; l != 32; l += 16) {
-      for (unsigned i = 0; i != 16; ++i) {
-        unsigned Idx = 32 + i - shiftVal;
-        if (Idx < 32) Idx -= 16; // end of lane, switch operand.
-        Indices[l + i] = Idx + l;
-      }
-    }
-
-    llvm::Type *VecTy = llvm::VectorType::get(Int8Ty, 32);
-    Ops[0] = Builder.CreateBitCast(Ops[0], VecTy, "cast");
-    Value *Zero = llvm::Constant::getNullValue(VecTy);
+    // If this isn't a masked builtin, just return the align operation.
+    if (Ops.size() == 3)
+      return Align;
 
-    Value *SV = llvm::ConstantDataVector::get(getLLVMContext(), Indices);
-    SV = Builder.CreateShuffleVector(Zero, Ops[0], SV, "pslldq");
-    llvm::Type *ResultType = ConvertType(E->getType());
-    return Builder.CreateBitCast(SV, ResultType, "cast");
+    return EmitX86Select(*this, Ops[4], Align, Ops[3]);
   }
-  case X86::BI__builtin_ia32_psrldqi256: {
-    // Shift value is in bits so divide by 8.
-    unsigned shiftVal = cast<llvm::ConstantInt>(Ops[1])->getZExtValue() >> 3;
-
-    // If psrldq is shifting the vector more than 15 bytes, emit zero.
-    if (shiftVal >= 16)
-      return llvm::Constant::getNullValue(ConvertType(E->getType()));
 
-    uint32_t Indices[32];
-    // 256-bit psrldq operates on 128-bit lanes so we need to handle that
-    for (unsigned l = 0; l != 32; l += 16) {
-      for (unsigned i = 0; i != 16; ++i) {
-        unsigned Idx = i + shiftVal;
-        if (Idx >= 16) Idx += 16; // end of lane, switch operand.
-        Indices[l + i] = Idx + l;
-      }
-    }
-
-    llvm::Type *VecTy = llvm::VectorType::get(Int8Ty, 32);
-    Ops[0] = Builder.CreateBitCast(Ops[0], VecTy, "cast");
-    Value *Zero = llvm::Constant::getNullValue(VecTy);
-
-    Value *SV = llvm::ConstantDataVector::get(getLLVMContext(), Indices);
-    SV = Builder.CreateShuffleVector(Ops[0], Zero, SV, "psrldq");
-    llvm::Type *ResultType = ConvertType(E->getType());
-    return Builder.CreateBitCast(SV, ResultType, "cast");
-  }
-  case X86::BI__builtin_ia32_movntps:
-  case X86::BI__builtin_ia32_movntps256:
-  case X86::BI__builtin_ia32_movntpd:
-  case X86::BI__builtin_ia32_movntpd256:
-  case X86::BI__builtin_ia32_movntdq:
-  case X86::BI__builtin_ia32_movntdq256:
   case X86::BI__builtin_ia32_movnti:
   case X86::BI__builtin_ia32_movnti64: {
     llvm::MDNode *Node = llvm::MDNode::get(
@@ -6439,17 +7069,156 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
     StoreInst *SI = Builder.CreateDefaultAlignedStore(Ops[1], BC);
     SI->setMetadata(CGM.getModule().getMDKindID("nontemporal"), Node);
 
-    // If the operand is an integer, we can't assume alignment. Otherwise,
-    // assume natural alignment.
-    QualType ArgTy = E->getArg(1)->getType();
-    unsigned Align;
-    if (ArgTy->isIntegerType())
-      Align = 1;
-    else
-      Align = getContext().getTypeSizeInChars(ArgTy).getQuantity();
-    SI->setAlignment(Align);
+    // No alignment for scalar intrinsic store.
+    SI->setAlignment(1);
+    return SI;
+  }
+  case X86::BI__builtin_ia32_movntsd:
+  case X86::BI__builtin_ia32_movntss: {
+    llvm::MDNode *Node = llvm::MDNode::get(
+        getLLVMContext(), llvm::ConstantAsMetadata::get(Builder.getInt32(1)));
+
+    // Extract the 0'th element of the source vector.
+    Value *Scl = Builder.CreateExtractElement(Ops[1], (uint64_t)0, "extract");
+
+    // Convert the type of the pointer to a pointer to the stored type.
+    Value *BC = Builder.CreateBitCast(Ops[0],
+                                llvm::PointerType::getUnqual(Scl->getType()),
+                                      "cast");
+
+    // Unaligned nontemporal store of the scalar value.
+    StoreInst *SI = Builder.CreateDefaultAlignedStore(Scl, BC);
+    SI->setMetadata(CGM.getModule().getMDKindID("nontemporal"), Node);
+    SI->setAlignment(1);
     return SI;
   }
+
+  case X86::BI__builtin_ia32_selectb_128:
+  case X86::BI__builtin_ia32_selectb_256:
+  case X86::BI__builtin_ia32_selectb_512:
+  case X86::BI__builtin_ia32_selectw_128:
+  case X86::BI__builtin_ia32_selectw_256:
+  case X86::BI__builtin_ia32_selectw_512:
+  case X86::BI__builtin_ia32_selectd_128:
+  case X86::BI__builtin_ia32_selectd_256:
+  case X86::BI__builtin_ia32_selectd_512:
+  case X86::BI__builtin_ia32_selectq_128:
+  case X86::BI__builtin_ia32_selectq_256:
+  case X86::BI__builtin_ia32_selectq_512:
+  case X86::BI__builtin_ia32_selectps_128:
+  case X86::BI__builtin_ia32_selectps_256:
+  case X86::BI__builtin_ia32_selectps_512:
+  case X86::BI__builtin_ia32_selectpd_128:
+  case X86::BI__builtin_ia32_selectpd_256:
+  case X86::BI__builtin_ia32_selectpd_512:
+    return EmitX86Select(*this, Ops[0], Ops[1], Ops[2]);
+  case X86::BI__builtin_ia32_pcmpeqb128_mask:
+  case X86::BI__builtin_ia32_pcmpeqb256_mask:
+  case X86::BI__builtin_ia32_pcmpeqb512_mask:
+  case X86::BI__builtin_ia32_pcmpeqw128_mask:
+  case X86::BI__builtin_ia32_pcmpeqw256_mask:
+  case X86::BI__builtin_ia32_pcmpeqw512_mask:
+  case X86::BI__builtin_ia32_pcmpeqd128_mask:
+  case X86::BI__builtin_ia32_pcmpeqd256_mask:
+  case X86::BI__builtin_ia32_pcmpeqd512_mask:
+  case X86::BI__builtin_ia32_pcmpeqq128_mask:
+  case X86::BI__builtin_ia32_pcmpeqq256_mask:
+  case X86::BI__builtin_ia32_pcmpeqq512_mask:
+    return EmitX86MaskedCompare(*this, 0, false, Ops);
+  case X86::BI__builtin_ia32_pcmpgtb128_mask:
+  case X86::BI__builtin_ia32_pcmpgtb256_mask:
+  case X86::BI__builtin_ia32_pcmpgtb512_mask:
+  case X86::BI__builtin_ia32_pcmpgtw128_mask:
+  case X86::BI__builtin_ia32_pcmpgtw256_mask:
+  case X86::BI__builtin_ia32_pcmpgtw512_mask:
+  case X86::BI__builtin_ia32_pcmpgtd128_mask:
+  case X86::BI__builtin_ia32_pcmpgtd256_mask:
+  case X86::BI__builtin_ia32_pcmpgtd512_mask:
+  case X86::BI__builtin_ia32_pcmpgtq128_mask:
+  case X86::BI__builtin_ia32_pcmpgtq256_mask:
+  case X86::BI__builtin_ia32_pcmpgtq512_mask:
+    return EmitX86MaskedCompare(*this, 6, true, Ops);
+  case X86::BI__builtin_ia32_cmpb128_mask:
+  case X86::BI__builtin_ia32_cmpb256_mask:
+  case X86::BI__builtin_ia32_cmpb512_mask:
+  case X86::BI__builtin_ia32_cmpw128_mask:
+  case X86::BI__builtin_ia32_cmpw256_mask:
+  case X86::BI__builtin_ia32_cmpw512_mask:
+  case X86::BI__builtin_ia32_cmpd128_mask:
+  case X86::BI__builtin_ia32_cmpd256_mask:
+  case X86::BI__builtin_ia32_cmpd512_mask:
+  case X86::BI__builtin_ia32_cmpq128_mask:
+  case X86::BI__builtin_ia32_cmpq256_mask:
+  case X86::BI__builtin_ia32_cmpq512_mask: {
+    unsigned CC = cast<llvm::ConstantInt>(Ops[2])->getZExtValue() & 0x7;
+    return EmitX86MaskedCompare(*this, CC, true, Ops);
+  }
+  case X86::BI__builtin_ia32_ucmpb128_mask:
+  case X86::BI__builtin_ia32_ucmpb256_mask:
+  case X86::BI__builtin_ia32_ucmpb512_mask:
+  case X86::BI__builtin_ia32_ucmpw128_mask:
+  case X86::BI__builtin_ia32_ucmpw256_mask:
+  case X86::BI__builtin_ia32_ucmpw512_mask:
+  case X86::BI__builtin_ia32_ucmpd128_mask:
+  case X86::BI__builtin_ia32_ucmpd256_mask:
+  case X86::BI__builtin_ia32_ucmpd512_mask:
+  case X86::BI__builtin_ia32_ucmpq128_mask:
+  case X86::BI__builtin_ia32_ucmpq256_mask:
+  case X86::BI__builtin_ia32_ucmpq512_mask: {
+    unsigned CC = cast<llvm::ConstantInt>(Ops[2])->getZExtValue() & 0x7;
+    return EmitX86MaskedCompare(*this, CC, false, Ops);
+  }
+
+  case X86::BI__builtin_ia32_vplzcntd_128_mask:
+  case X86::BI__builtin_ia32_vplzcntd_256_mask:
+  case X86::BI__builtin_ia32_vplzcntd_512_mask:
+  case X86::BI__builtin_ia32_vplzcntq_128_mask:
+  case X86::BI__builtin_ia32_vplzcntq_256_mask:
+  case X86::BI__builtin_ia32_vplzcntq_512_mask: {
+    Function *F = CGM.getIntrinsic(Intrinsic::ctlz, Ops[0]->getType());
+    return EmitX86Select(*this, Ops[2],
+                         Builder.CreateCall(F, {Ops[0],Builder.getInt1(false)}),
+                         Ops[1]);
+  }
+
+  // TODO: Handle 64/512-bit vector widths of min/max.
+  case X86::BI__builtin_ia32_pmaxsb128:
+  case X86::BI__builtin_ia32_pmaxsw128:
+  case X86::BI__builtin_ia32_pmaxsd128:
+  case X86::BI__builtin_ia32_pmaxsb256:
+  case X86::BI__builtin_ia32_pmaxsw256:
+  case X86::BI__builtin_ia32_pmaxsd256: {
+    Value *Cmp = Builder.CreateICmp(ICmpInst::ICMP_SGT, Ops[0], Ops[1]);
+    return Builder.CreateSelect(Cmp, Ops[0], Ops[1]);
+  }
+  case X86::BI__builtin_ia32_pmaxub128:
+  case X86::BI__builtin_ia32_pmaxuw128:
+  case X86::BI__builtin_ia32_pmaxud128:
+  case X86::BI__builtin_ia32_pmaxub256:
+  case X86::BI__builtin_ia32_pmaxuw256:
+  case X86::BI__builtin_ia32_pmaxud256: {
+    Value *Cmp = Builder.CreateICmp(ICmpInst::ICMP_UGT, Ops[0], Ops[1]);
+    return Builder.CreateSelect(Cmp, Ops[0], Ops[1]);
+  }
+  case X86::BI__builtin_ia32_pminsb128:
+  case X86::BI__builtin_ia32_pminsw128:
+  case X86::BI__builtin_ia32_pminsd128:
+  case X86::BI__builtin_ia32_pminsb256:
+  case X86::BI__builtin_ia32_pminsw256:
+  case X86::BI__builtin_ia32_pminsd256: {
+    Value *Cmp = Builder.CreateICmp(ICmpInst::ICMP_SLT, Ops[0], Ops[1]);
+    return Builder.CreateSelect(Cmp, Ops[0], Ops[1]);
+  }
+  case X86::BI__builtin_ia32_pminub128:
+  case X86::BI__builtin_ia32_pminuw128:
+  case X86::BI__builtin_ia32_pminud128:
+  case X86::BI__builtin_ia32_pminub256:
+  case X86::BI__builtin_ia32_pminuw256:
+  case X86::BI__builtin_ia32_pminud256: {
+    Value *Cmp = Builder.CreateICmp(ICmpInst::ICMP_ULT, Ops[0], Ops[1]);
+    return Builder.CreateSelect(Cmp, Ops[0], Ops[1]);
+  }
+
   // 3DNow!
   case X86::BI__builtin_ia32_pswapdsf:
   case X86::BI__builtin_ia32_pswapdsi: {
@@ -6492,154 +7261,107 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
                                       Ops[0]);
     return Builder.CreateExtractValue(Call, 1);
   }
-  // SSE comparison intrisics
+
+  // SSE packed comparison intrinsics
   case X86::BI__builtin_ia32_cmpeqps:
+  case X86::BI__builtin_ia32_cmpeqpd:
+    return getVectorFCmpIR(CmpInst::FCMP_OEQ);
   case X86::BI__builtin_ia32_cmpltps:
+  case X86::BI__builtin_ia32_cmpltpd:
+    return getVectorFCmpIR(CmpInst::FCMP_OLT);
   case X86::BI__builtin_ia32_cmpleps:
+  case X86::BI__builtin_ia32_cmplepd:
+    return getVectorFCmpIR(CmpInst::FCMP_OLE);
   case X86::BI__builtin_ia32_cmpunordps:
+  case X86::BI__builtin_ia32_cmpunordpd:
+    return getVectorFCmpIR(CmpInst::FCMP_UNO);
   case X86::BI__builtin_ia32_cmpneqps:
+  case X86::BI__builtin_ia32_cmpneqpd:
+    return getVectorFCmpIR(CmpInst::FCMP_UNE);
   case X86::BI__builtin_ia32_cmpnltps:
+  case X86::BI__builtin_ia32_cmpnltpd:
+    return getVectorFCmpIR(CmpInst::FCMP_UGE);
   case X86::BI__builtin_ia32_cmpnleps:
+  case X86::BI__builtin_ia32_cmpnlepd:
+    return getVectorFCmpIR(CmpInst::FCMP_UGT);
   case X86::BI__builtin_ia32_cmpordps:
+  case X86::BI__builtin_ia32_cmpordpd:
+    return getVectorFCmpIR(CmpInst::FCMP_ORD);
+  case X86::BI__builtin_ia32_cmpps:
+  case X86::BI__builtin_ia32_cmpps256:
+  case X86::BI__builtin_ia32_cmppd:
+  case X86::BI__builtin_ia32_cmppd256: {
+    unsigned CC = cast<llvm::ConstantInt>(Ops[2])->getZExtValue();
+    // If this one of the SSE immediates, we can use native IR.
+    if (CC < 8) {
+      FCmpInst::Predicate Pred;
+      switch (CC) {
+      case 0: Pred = FCmpInst::FCMP_OEQ; break;
+      case 1: Pred = FCmpInst::FCMP_OLT; break;
+      case 2: Pred = FCmpInst::FCMP_OLE; break;
+      case 3: Pred = FCmpInst::FCMP_UNO; break;
+      case 4: Pred = FCmpInst::FCMP_UNE; break;
+      case 5: Pred = FCmpInst::FCMP_UGE; break;
+      case 6: Pred = FCmpInst::FCMP_UGT; break;
+      case 7: Pred = FCmpInst::FCMP_ORD; break;
+      }
+      return getVectorFCmpIR(Pred);
+    }
+
+    // We can't handle 8-31 immediates with native IR, use the intrinsic.
+    Intrinsic::ID ID;
+    switch (BuiltinID) {
+    default: llvm_unreachable("Unsupported intrinsic!");
+    case X86::BI__builtin_ia32_cmpps:
+      ID = Intrinsic::x86_sse_cmp_ps;
+      break;
+    case X86::BI__builtin_ia32_cmpps256:
+      ID = Intrinsic::x86_avx_cmp_ps_256;
+      break;
+    case X86::BI__builtin_ia32_cmppd:
+      ID = Intrinsic::x86_sse2_cmp_pd;
+      break;
+    case X86::BI__builtin_ia32_cmppd256:
+      ID = Intrinsic::x86_avx_cmp_pd_256;
+      break;
+    }
+
+    return Builder.CreateCall(CGM.getIntrinsic(ID), Ops);
+  }
+
+  // SSE scalar comparison intrinsics
   case X86::BI__builtin_ia32_cmpeqss:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 0);
   case X86::BI__builtin_ia32_cmpltss:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 1);
   case X86::BI__builtin_ia32_cmpless:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 2);
   case X86::BI__builtin_ia32_cmpunordss:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 3);
   case X86::BI__builtin_ia32_cmpneqss:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 4);
   case X86::BI__builtin_ia32_cmpnltss:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 5);
   case X86::BI__builtin_ia32_cmpnless:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 6);
   case X86::BI__builtin_ia32_cmpordss:
-  case X86::BI__builtin_ia32_cmpeqpd:
-  case X86::BI__builtin_ia32_cmpltpd:
-  case X86::BI__builtin_ia32_cmplepd:
-  case X86::BI__builtin_ia32_cmpunordpd:
-  case X86::BI__builtin_ia32_cmpneqpd:
-  case X86::BI__builtin_ia32_cmpnltpd:
-  case X86::BI__builtin_ia32_cmpnlepd:
-  case X86::BI__builtin_ia32_cmpordpd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse_cmp_ss, 7);
   case X86::BI__builtin_ia32_cmpeqsd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 0);
   case X86::BI__builtin_ia32_cmpltsd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 1);
   case X86::BI__builtin_ia32_cmplesd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 2);
   case X86::BI__builtin_ia32_cmpunordsd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 3);
   case X86::BI__builtin_ia32_cmpneqsd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 4);
   case X86::BI__builtin_ia32_cmpnltsd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 5);
   case X86::BI__builtin_ia32_cmpnlesd:
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 6);
   case X86::BI__builtin_ia32_cmpordsd:
-    // These exist so that the builtin that takes an immediate can be bounds
-    // checked by clang to avoid passing bad immediates to the backend. Since
-    // AVX has a larger immediate than SSE we would need separate builtins to
-    // do the different bounds checking. Rather than create a clang specific
-    // SSE only builtin, this implements eight separate builtins to match gcc
-    // implementation.
-
-    // Choose the immediate.
-    unsigned Imm;
-    switch (BuiltinID) {
-    default: llvm_unreachable("Unsupported intrinsic!");
-    case X86::BI__builtin_ia32_cmpeqps:
-    case X86::BI__builtin_ia32_cmpeqss:
-    case X86::BI__builtin_ia32_cmpeqpd:
-    case X86::BI__builtin_ia32_cmpeqsd:
-      Imm = 0;
-      break;
-    case X86::BI__builtin_ia32_cmpltps:
-    case X86::BI__builtin_ia32_cmpltss:
-    case X86::BI__builtin_ia32_cmpltpd:
-    case X86::BI__builtin_ia32_cmpltsd:
-      Imm = 1;
-      break;
-    case X86::BI__builtin_ia32_cmpleps:
-    case X86::BI__builtin_ia32_cmpless:
-    case X86::BI__builtin_ia32_cmplepd:
-    case X86::BI__builtin_ia32_cmplesd:
-      Imm = 2;
-      break;
-    case X86::BI__builtin_ia32_cmpunordps:
-    case X86::BI__builtin_ia32_cmpunordss:
-    case X86::BI__builtin_ia32_cmpunordpd:
-    case X86::BI__builtin_ia32_cmpunordsd:
-      Imm = 3;
-      break;
-    case X86::BI__builtin_ia32_cmpneqps:
-    case X86::BI__builtin_ia32_cmpneqss:
-    case X86::BI__builtin_ia32_cmpneqpd:
-    case X86::BI__builtin_ia32_cmpneqsd:
-      Imm = 4;
-      break;
-    case X86::BI__builtin_ia32_cmpnltps:
-    case X86::BI__builtin_ia32_cmpnltss:
-    case X86::BI__builtin_ia32_cmpnltpd:
-    case X86::BI__builtin_ia32_cmpnltsd:
-      Imm = 5;
-      break;
-    case X86::BI__builtin_ia32_cmpnleps:
-    case X86::BI__builtin_ia32_cmpnless:
-    case X86::BI__builtin_ia32_cmpnlepd:
-    case X86::BI__builtin_ia32_cmpnlesd:
-      Imm = 6;
-      break;
-    case X86::BI__builtin_ia32_cmpordps:
-    case X86::BI__builtin_ia32_cmpordss:
-    case X86::BI__builtin_ia32_cmpordpd:
-    case X86::BI__builtin_ia32_cmpordsd:
-      Imm = 7;
-      break;
-    }
-
-    // Choose the intrinsic ID.
-    const char *name;
-    Intrinsic::ID ID;
-    switch (BuiltinID) {
-    default: llvm_unreachable("Unsupported intrinsic!");
-    case X86::BI__builtin_ia32_cmpeqps:
-    case X86::BI__builtin_ia32_cmpltps:
-    case X86::BI__builtin_ia32_cmpleps:
-    case X86::BI__builtin_ia32_cmpunordps:
-    case X86::BI__builtin_ia32_cmpneqps:
-    case X86::BI__builtin_ia32_cmpnltps:
-    case X86::BI__builtin_ia32_cmpnleps:
-    case X86::BI__builtin_ia32_cmpordps:
-      name = "cmpps";
-      ID = Intrinsic::x86_sse_cmp_ps;
-      break;
-    case X86::BI__builtin_ia32_cmpeqss:
-    case X86::BI__builtin_ia32_cmpltss:
-    case X86::BI__builtin_ia32_cmpless:
-    case X86::BI__builtin_ia32_cmpunordss:
-    case X86::BI__builtin_ia32_cmpneqss:
-    case X86::BI__builtin_ia32_cmpnltss:
-    case X86::BI__builtin_ia32_cmpnless:
-    case X86::BI__builtin_ia32_cmpordss:
-      name = "cmpss";
-      ID = Intrinsic::x86_sse_cmp_ss;
-      break;
-    case X86::BI__builtin_ia32_cmpeqpd:
-    case X86::BI__builtin_ia32_cmpltpd:
-    case X86::BI__builtin_ia32_cmplepd:
-    case X86::BI__builtin_ia32_cmpunordpd:
-    case X86::BI__builtin_ia32_cmpneqpd:
-    case X86::BI__builtin_ia32_cmpnltpd:
-    case X86::BI__builtin_ia32_cmpnlepd:
-    case X86::BI__builtin_ia32_cmpordpd:
-      name = "cmppd";
-      ID = Intrinsic::x86_sse2_cmp_pd;
-      break;
-    case X86::BI__builtin_ia32_cmpeqsd:
-    case X86::BI__builtin_ia32_cmpltsd:
-    case X86::BI__builtin_ia32_cmplesd:
-    case X86::BI__builtin_ia32_cmpunordsd:
-    case X86::BI__builtin_ia32_cmpneqsd:
-    case X86::BI__builtin_ia32_cmpnltsd:
-    case X86::BI__builtin_ia32_cmpnlesd:
-    case X86::BI__builtin_ia32_cmpordsd:
-      name = "cmpsd";
-      ID = Intrinsic::x86_sse2_cmp_sd;
-      break;
-    }
-
-    Ops.push_back(llvm::ConstantInt::get(Int8Ty, Imm));
-    llvm::Function *F = CGM.getIntrinsic(ID);
-    return Builder.CreateCall(F, Ops, name);
+    return getCmpIntrinsicCall(Intrinsic::x86_sse2_cmp_sd, 7);
   }
 }
 
@@ -6812,6 +7534,16 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
     llvm::Function *F = CGM.getIntrinsic(ID, ResultType);
     return Builder.CreateCall(F, X);
   }
+
+  // Absolute value
+  case PPC::BI__builtin_vsx_xvabsdp:
+  case PPC::BI__builtin_vsx_xvabssp: {
+    llvm::Type *ResultType = ConvertType(E->getType());
+    Value *X = EmitScalarExpr(E->getArg(0));
+    llvm::Function *F = CGM.getIntrinsic(Intrinsic::fabs, ResultType);
+    return Builder.CreateCall(F, X);
+  }
+
   // FMA variations
   case PPC::BI__builtin_vsx_xvmaddadp:
   case PPC::BI__builtin_vsx_xvmaddasp:
@@ -6851,44 +7583,11 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
   }
 }
 
-// Emit an intrinsic that has 1 float or double.
-static Value *emitUnaryFPBuiltin(CodeGenFunction &CGF,
-                                 const CallExpr *E,
-                                 unsigned IntrinsicID) {
-  llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
-
-  Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
-  return CGF.Builder.CreateCall(F, Src0);
-}
-
-// Emit an intrinsic that has 3 float or double operands.
-static Value *emitTernaryFPBuiltin(CodeGenFunction &CGF,
-                                   const CallExpr *E,
-                                   unsigned IntrinsicID) {
-  llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
-  llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
-  llvm::Value *Src2 = CGF.EmitScalarExpr(E->getArg(2));
-
-  Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
-  return CGF.Builder.CreateCall(F, {Src0, Src1, Src2});
-}
-
-// Emit an intrinsic that has 1 float or double operand, and 1 integer.
-static Value *emitFPIntBuiltin(CodeGenFunction &CGF,
-                               const CallExpr *E,
-                               unsigned IntrinsicID) {
-  llvm::Value *Src0 = CGF.EmitScalarExpr(E->getArg(0));
-  llvm::Value *Src1 = CGF.EmitScalarExpr(E->getArg(1));
-
-  Value *F = CGF.CGM.getIntrinsic(IntrinsicID, Src0->getType());
-  return CGF.Builder.CreateCall(F, {Src0, Src1});
-}
-
 Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
                                               const CallExpr *E) {
   switch (BuiltinID) {
-  case AMDGPU::BI__builtin_amdgpu_div_scale:
-  case AMDGPU::BI__builtin_amdgpu_div_scalef: {
+  case AMDGPU::BI__builtin_amdgcn_div_scale:
+  case AMDGPU::BI__builtin_amdgcn_div_scalef: {
     // Translate from the intrinsics's struct return to the builtin's out
     // argument.
 
@@ -6898,7 +7597,7 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
     llvm::Value *Y = EmitScalarExpr(E->getArg(1));
     llvm::Value *Z = EmitScalarExpr(E->getArg(2));
 
-    llvm::Value *Callee = CGM.getIntrinsic(Intrinsic::AMDGPU_div_scale,
+    llvm::Value *Callee = CGM.getIntrinsic(Intrinsic::amdgcn_div_scale,
                                            X->getType());
 
     llvm::Value *Tmp = Builder.CreateCall(Callee, {X, Y, Z});
@@ -6913,40 +7612,85 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
     Builder.CreateStore(FlagExt, FlagOutPtr);
     return Result;
   }
-  case AMDGPU::BI__builtin_amdgpu_div_fmas:
-  case AMDGPU::BI__builtin_amdgpu_div_fmasf: {
+  case AMDGPU::BI__builtin_amdgcn_div_fmas:
+  case AMDGPU::BI__builtin_amdgcn_div_fmasf: {
     llvm::Value *Src0 = EmitScalarExpr(E->getArg(0));
     llvm::Value *Src1 = EmitScalarExpr(E->getArg(1));
     llvm::Value *Src2 = EmitScalarExpr(E->getArg(2));
     llvm::Value *Src3 = EmitScalarExpr(E->getArg(3));
 
-    llvm::Value *F = CGM.getIntrinsic(Intrinsic::AMDGPU_div_fmas,
+    llvm::Value *F = CGM.getIntrinsic(Intrinsic::amdgcn_div_fmas,
                                       Src0->getType());
     llvm::Value *Src3ToBool = Builder.CreateIsNotNull(Src3);
     return Builder.CreateCall(F, {Src0, Src1, Src2, Src3ToBool});
   }
-  case AMDGPU::BI__builtin_amdgpu_div_fixup:
-  case AMDGPU::BI__builtin_amdgpu_div_fixupf:
-    return emitTernaryFPBuiltin(*this, E, Intrinsic::AMDGPU_div_fixup);
-  case AMDGPU::BI__builtin_amdgpu_trig_preop:
-  case AMDGPU::BI__builtin_amdgpu_trig_preopf:
-    return emitFPIntBuiltin(*this, E, Intrinsic::AMDGPU_trig_preop);
-  case AMDGPU::BI__builtin_amdgpu_rcp:
-  case AMDGPU::BI__builtin_amdgpu_rcpf:
-    return emitUnaryFPBuiltin(*this, E, Intrinsic::AMDGPU_rcp);
-  case AMDGPU::BI__builtin_amdgpu_rsq:
-  case AMDGPU::BI__builtin_amdgpu_rsqf:
-    return emitUnaryFPBuiltin(*this, E, Intrinsic::AMDGPU_rsq);
-  case AMDGPU::BI__builtin_amdgpu_rsq_clamped:
-  case AMDGPU::BI__builtin_amdgpu_rsq_clampedf:
-    return emitUnaryFPBuiltin(*this, E, Intrinsic::AMDGPU_rsq_clamped);
-  case AMDGPU::BI__builtin_amdgpu_ldexp:
-  case AMDGPU::BI__builtin_amdgpu_ldexpf:
-    return emitFPIntBuiltin(*this, E, Intrinsic::AMDGPU_ldexp);
-  case AMDGPU::BI__builtin_amdgpu_class:
-  case AMDGPU::BI__builtin_amdgpu_classf:
-    return emitFPIntBuiltin(*this, E, Intrinsic::AMDGPU_class);
-   default:
+  case AMDGPU::BI__builtin_amdgcn_div_fixup:
+  case AMDGPU::BI__builtin_amdgcn_div_fixupf:
+    return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_div_fixup);
+  case AMDGPU::BI__builtin_amdgcn_trig_preop:
+  case AMDGPU::BI__builtin_amdgcn_trig_preopf:
+    return emitFPIntBuiltin(*this, E, Intrinsic::amdgcn_trig_preop);
+  case AMDGPU::BI__builtin_amdgcn_rcp:
+  case AMDGPU::BI__builtin_amdgcn_rcpf:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_rcp);
+  case AMDGPU::BI__builtin_amdgcn_rsq:
+  case AMDGPU::BI__builtin_amdgcn_rsqf:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_rsq);
+  case AMDGPU::BI__builtin_amdgcn_rsq_clamp:
+  case AMDGPU::BI__builtin_amdgcn_rsq_clampf:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_rsq_clamp);
+  case AMDGPU::BI__builtin_amdgcn_sinf:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_sin);
+  case AMDGPU::BI__builtin_amdgcn_cosf:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_cos);
+  case AMDGPU::BI__builtin_amdgcn_log_clampf:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_log_clamp);
+  case AMDGPU::BI__builtin_amdgcn_ldexp:
+  case AMDGPU::BI__builtin_amdgcn_ldexpf:
+    return emitFPIntBuiltin(*this, E, Intrinsic::amdgcn_ldexp);
+  case AMDGPU::BI__builtin_amdgcn_frexp_mant:
+  case AMDGPU::BI__builtin_amdgcn_frexp_mantf: {
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_frexp_mant);
+  }
+  case AMDGPU::BI__builtin_amdgcn_frexp_exp:
+  case AMDGPU::BI__builtin_amdgcn_frexp_expf: {
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_frexp_exp);
+  }
+  case AMDGPU::BI__builtin_amdgcn_fract:
+  case AMDGPU::BI__builtin_amdgcn_fractf:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_fract);
+  case AMDGPU::BI__builtin_amdgcn_lerp:
+    return emitTernaryBuiltin(*this, E, Intrinsic::amdgcn_lerp);
+  case AMDGPU::BI__builtin_amdgcn_class:
+  case AMDGPU::BI__builtin_amdgcn_classf:
+    return emitFPIntBuiltin(*this, E, Intrinsic::amdgcn_class);
+
+  case AMDGPU::BI__builtin_amdgcn_read_exec: {
+    CallInst *CI = cast<CallInst>(
+      EmitSpecialRegisterBuiltin(*this, E, Int64Ty, Int64Ty, true, "exec"));
+    CI->setConvergent();
+    return CI;
+  }
+
+  // amdgcn workitem
+  case AMDGPU::BI__builtin_amdgcn_workitem_id_x:
+    return emitRangedBuiltin(*this, Intrinsic::amdgcn_workitem_id_x, 0, 1024);
+  case AMDGPU::BI__builtin_amdgcn_workitem_id_y:
+    return emitRangedBuiltin(*this, Intrinsic::amdgcn_workitem_id_y, 0, 1024);
+  case AMDGPU::BI__builtin_amdgcn_workitem_id_z:
+    return emitRangedBuiltin(*this, Intrinsic::amdgcn_workitem_id_z, 0, 1024);
+
+  // r600 intrinsics
+  case AMDGPU::BI__builtin_r600_recipsqrt_ieee:
+  case AMDGPU::BI__builtin_r600_recipsqrt_ieeef:
+    return emitUnaryBuiltin(*this, E, Intrinsic::r600_recipsqrt_ieee);
+  case AMDGPU::BI__builtin_r600_read_tidig_x:
+    return emitRangedBuiltin(*this, Intrinsic::r600_read_tidig_x, 0, 1024);
+  case AMDGPU::BI__builtin_r600_read_tidig_y:
+    return emitRangedBuiltin(*this, Intrinsic::r600_read_tidig_y, 0, 1024);
+  case AMDGPU::BI__builtin_r600_read_tidig_z:
+    return emitRangedBuiltin(*this, Intrinsic::r600_read_tidig_z, 0, 1024);
+  default:
     return nullptr;
   }
 }
@@ -7196,6 +7940,17 @@ Value *CodeGenFunction::EmitSystemZBuiltinExpr(unsigned BuiltinID,
 
 Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
                                              const CallExpr *E) {
+  auto MakeLdg = [&](unsigned IntrinsicID) {
+    Value *Ptr = EmitScalarExpr(E->getArg(0));
+    AlignmentSource AlignSource;
+    clang::CharUnits Align =
+        getNaturalPointeeTypeAlignment(E->getArg(0)->getType(), &AlignSource);
+    return Builder.CreateCall(
+        CGM.getIntrinsic(IntrinsicID, {Ptr->getType()->getPointerElementType(),
+                                       Ptr->getType()}),
+        {Ptr, ConstantInt::get(Builder.getInt32Ty(), Align.getQuantity())});
+  };
+
   switch (BuiltinID) {
   case NVPTX::BI__nvvm_atom_add_gen_i:
   case NVPTX::BI__nvvm_atom_add_gen_l:
@@ -7264,6 +8019,56 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
     return Builder.CreateCall(FnALAF32, {Ptr, Val});
   }
 
+  case NVPTX::BI__nvvm_atom_inc_gen_ui: {
+    Value *Ptr = EmitScalarExpr(E->getArg(0));
+    Value *Val = EmitScalarExpr(E->getArg(1));
+    Value *FnALI32 =
+        CGM.getIntrinsic(Intrinsic::nvvm_atomic_load_inc_32, Ptr->getType());
+    return Builder.CreateCall(FnALI32, {Ptr, Val});
+  }
+
+  case NVPTX::BI__nvvm_atom_dec_gen_ui: {
+    Value *Ptr = EmitScalarExpr(E->getArg(0));
+    Value *Val = EmitScalarExpr(E->getArg(1));
+    Value *FnALD32 =
+        CGM.getIntrinsic(Intrinsic::nvvm_atomic_load_dec_32, Ptr->getType());
+    return Builder.CreateCall(FnALD32, {Ptr, Val});
+  }
+
+  case NVPTX::BI__nvvm_ldg_c:
+  case NVPTX::BI__nvvm_ldg_c2:
+  case NVPTX::BI__nvvm_ldg_c4:
+  case NVPTX::BI__nvvm_ldg_s:
+  case NVPTX::BI__nvvm_ldg_s2:
+  case NVPTX::BI__nvvm_ldg_s4:
+  case NVPTX::BI__nvvm_ldg_i:
+  case NVPTX::BI__nvvm_ldg_i2:
+  case NVPTX::BI__nvvm_ldg_i4:
+  case NVPTX::BI__nvvm_ldg_l:
+  case NVPTX::BI__nvvm_ldg_ll:
+  case NVPTX::BI__nvvm_ldg_ll2:
+  case NVPTX::BI__nvvm_ldg_uc:
+  case NVPTX::BI__nvvm_ldg_uc2:
+  case NVPTX::BI__nvvm_ldg_uc4:
+  case NVPTX::BI__nvvm_ldg_us:
+  case NVPTX::BI__nvvm_ldg_us2:
+  case NVPTX::BI__nvvm_ldg_us4:
+  case NVPTX::BI__nvvm_ldg_ui:
+  case NVPTX::BI__nvvm_ldg_ui2:
+  case NVPTX::BI__nvvm_ldg_ui4:
+  case NVPTX::BI__nvvm_ldg_ul:
+  case NVPTX::BI__nvvm_ldg_ull:
+  case NVPTX::BI__nvvm_ldg_ull2:
+    // PTX Interoperability section 2.2: "For a vector with an even number of
+    // elements, its alignment is set to number of elements times the alignment
+    // of its member: n*alignof(t)."
+    return MakeLdg(Intrinsic::nvvm_ldg_global_i);
+  case NVPTX::BI__nvvm_ldg_f:
+  case NVPTX::BI__nvvm_ldg_f2:
+  case NVPTX::BI__nvvm_ldg_f4:
+  case NVPTX::BI__nvvm_ldg_d:
+  case NVPTX::BI__nvvm_ldg_d2:
+    return MakeLdg(Intrinsic::nvvm_ldg_global_f);
   default:
     return nullptr;
   }
@@ -7272,9 +8077,9 @@ Value *CodeGenFunction::EmitNVPTXBuiltinExpr(unsigned BuiltinID,
 Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID,
                                                    const CallExpr *E) {
   switch (BuiltinID) {
-  case WebAssembly::BI__builtin_wasm_memory_size: {
+  case WebAssembly::BI__builtin_wasm_current_memory: {
     llvm::Type *ResultType = ConvertType(E->getType());
-    Value *Callee = CGM.getIntrinsic(Intrinsic::wasm_memory_size, ResultType);
+    Value *Callee = CGM.getIntrinsic(Intrinsic::wasm_current_memory, ResultType);
     return Builder.CreateCall(Callee);
   }
   case WebAssembly::BI__builtin_wasm_grow_memory: {
diff --git a/lib/CodeGen/CGCUDABuiltin.cpp b/lib/CodeGen/CGCUDABuiltin.cpp
new file mode 100644
index 000000000000..ea3b888635c3
--- /dev/null
+++ b/lib/CodeGen/CGCUDABuiltin.cpp
@@ -0,0 +1,117 @@
+//===----- CGCUDABuiltin.cpp - Codegen for CUDA builtins ------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Generates code for built-in CUDA calls which are not runtime-specific.
+// (Runtime-specific codegen lives in CGCUDARuntime.)
+//
+//===----------------------------------------------------------------------===//
+
+#include "CodeGenFunction.h"
+#include "clang/Basic/Builtins.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/IR/Instruction.h"
+#include "llvm/Support/MathExtras.h"
+
+using namespace clang;
+using namespace CodeGen;
+
+static llvm::Function *GetVprintfDeclaration(llvm::Module &M) {
+  llvm::Type *ArgTypes[] = {llvm::Type::getInt8PtrTy(M.getContext()),
+                            llvm::Type::getInt8PtrTy(M.getContext())};
+  llvm::FunctionType *VprintfFuncType = llvm::FunctionType::get(
+      llvm::Type::getInt32Ty(M.getContext()), ArgTypes, false);
+
+  if (auto* F = M.getFunction("vprintf")) {
+    // Our CUDA system header declares vprintf with the right signature, so
+    // nobody else should have been able to declare vprintf with a bogus
+    // signature.
+    assert(F->getFunctionType() == VprintfFuncType);
+    return F;
+  }
+
+  // vprintf doesn't already exist; create a declaration and insert it into the
+  // module.
+  return llvm::Function::Create(
+      VprintfFuncType, llvm::GlobalVariable::ExternalLinkage, "vprintf", &M);
+}
+
+// Transforms a call to printf into a call to the NVPTX vprintf syscall (which
+// isn't particularly special; it's invoked just like a regular function).
+// vprintf takes two args: A format string, and a pointer to a buffer containing
+// the varargs.
+//
+// For example, the call
+//
+//   printf("format string", arg1, arg2, arg3);
+//
+// is converted into something resembling
+//
+//   struct Tmp {
+//     Arg1 a1;
+//     Arg2 a2;
+//     Arg3 a3;
+//   };
+//   char* buf = alloca(sizeof(Tmp));
+//   *(Tmp*)buf = {a1, a2, a3};
+//   vprintf("format string", buf);
+//
+// buf is aligned to the max of {alignof(Arg1), ...}.  Furthermore, each of the
+// args is itself aligned to its preferred alignment.
+//
+// Note that by the time this function runs, E's args have already undergone the
+// standard C vararg promotion (short -> int, float -> double, etc.).
+RValue
+CodeGenFunction::EmitCUDADevicePrintfCallExpr(const CallExpr *E,
+                                              ReturnValueSlot ReturnValue) {
+  assert(getLangOpts().CUDA);
+  assert(getLangOpts().CUDAIsDevice);
+  assert(E->getBuiltinCallee() == Builtin::BIprintf);
+  assert(E->getNumArgs() >= 1); // printf always has at least one arg.
+
+  const llvm::DataLayout &DL = CGM.getDataLayout();
+  llvm::LLVMContext &Ctx = CGM.getLLVMContext();
+
+  CallArgList Args;
+  EmitCallArgs(Args,
+               E->getDirectCallee()->getType()->getAs<FunctionProtoType>(),
+               E->arguments(), E->getDirectCallee(),
+               /* ParamsToSkip = */ 0);
+
+  // We don't know how to emit non-scalar varargs.
+  if (std::any_of(Args.begin() + 1, Args.end(),
+                  [](const CallArg &A) { return !A.RV.isScalar(); })) {
+    CGM.ErrorUnsupported(E, "non-scalar arg to printf");
+    return RValue::get(llvm::ConstantInt::get(IntTy, 0));
+  }
+
+  // Construct and fill the args buffer that we'll pass to vprintf.
+  llvm::Value *BufferPtr;
+  if (Args.size() <= 1) {
+    // If there are no args, pass a null pointer to vprintf.
+    BufferPtr = llvm::ConstantPointerNull::get(llvm::Type::getInt8PtrTy(Ctx));
+  } else {
+    llvm::SmallVector<llvm::Type *, 8> ArgTypes;
+    for (unsigned I = 1, NumArgs = Args.size(); I < NumArgs; ++I)
+      ArgTypes.push_back(Args[I].RV.getScalarVal()->getType());
+    llvm::Type *AllocaTy = llvm::StructType::create(ArgTypes, "printf_args");
+    llvm::Value *Alloca = CreateTempAlloca(AllocaTy);
+
+    for (unsigned I = 1, NumArgs = Args.size(); I < NumArgs; ++I) {
+      llvm::Value *P = Builder.CreateStructGEP(AllocaTy, Alloca, I - 1);
+      llvm::Value *Arg = Args[I].RV.getScalarVal();
+      Builder.CreateAlignedStore(Arg, P, DL.getPrefTypeAlignment(Arg->getType()));
+    }
+    BufferPtr = Builder.CreatePointerCast(Alloca, llvm::Type::getInt8PtrTy(Ctx));
+  }
+
+  // Invoke vprintf and return.
+  llvm::Function* VprintfFunc = GetVprintfDeclaration(CGM.getModule());
+  return RValue::get(
+      Builder.CreateCall(VprintfFunc, {Args[0].RV.getScalarVal(), BufferPtr}));
+}
diff --git a/lib/CodeGen/CGCUDANV.cpp b/lib/CodeGen/CGCUDANV.cpp
index 045e19b189dc..6a04d4eea784 100644
--- a/lib/CodeGen/CGCUDANV.cpp
+++ b/lib/CodeGen/CGCUDANV.cpp
@@ -38,6 +38,7 @@ private:
   llvm::Module &TheModule;
   /// Keeps track of kernel launch stubs emitted in this module
   llvm::SmallVector<llvm::Function *, 16> EmittedKernels;
+  llvm::SmallVector<std::pair<llvm::GlobalVariable *, unsigned>, 16> DeviceVars;
   /// Keeps track of variables containing handles of GPU binaries. Populated by
   /// ModuleCtorFunction() and used to create corresponding cleanup calls in
   /// ModuleDtorFunction()
@@ -47,7 +48,7 @@ private:
   llvm::Constant *getLaunchFn() const;
 
   /// Creates a function to register all kernel stubs generated in this module.
-  llvm::Function *makeRegisterKernelsFn();
+  llvm::Function *makeRegisterGlobalsFn();
 
   /// Helper function that generates a constant string and returns a pointer to
   /// the start of the string.  The result of this function can be used anywhere
@@ -68,6 +69,10 @@ public:
   CGNVCUDARuntime(CodeGenModule &CGM);
 
   void emitDeviceStub(CodeGenFunction &CGF, FunctionArgList &Args) override;
+  void registerDeviceVar(llvm::GlobalVariable &Var, unsigned Flags) override {
+    DeviceVars.push_back(std::make_pair(&Var, Flags));
+  }
+
   /// Creates module constructor function
   llvm::Function *makeModuleCtorFunction() override;
   /// Creates module destructor function
@@ -93,10 +98,7 @@ CGNVCUDARuntime::CGNVCUDARuntime(CodeGenModule &CGM)
 
 llvm::Constant *CGNVCUDARuntime::getSetupArgumentFn() const {
   // cudaError_t cudaSetupArgument(void *, size_t, size_t)
-  std::vector<llvm::Type*> Params;
-  Params.push_back(VoidPtrTy);
-  Params.push_back(SizeTy);
-  Params.push_back(SizeTy);
+  llvm::Type *Params[] = {VoidPtrTy, SizeTy, SizeTy};
   return CGM.CreateRuntimeFunction(llvm::FunctionType::get(IntTy,
                                                            Params, false),
                                    "cudaSetupArgument");
@@ -158,19 +160,28 @@ void CGNVCUDARuntime::emitDeviceStubBody(CodeGenFunction &CGF,
   CGF.EmitBlock(EndBlock);
 }
 
-/// Creates internal function to register all kernel stubs generated in this
-/// module with the CUDA runtime.
+/// Creates a function that sets up state on the host side for CUDA objects that
+/// have a presence on both the host and device sides. Specifically, registers
+/// the host side of kernel functions and device global variables with the CUDA
+/// runtime.
 /// \code
-/// void __cuda_register_kernels(void** GpuBinaryHandle) {
+/// void __cuda_register_globals(void** GpuBinaryHandle) {
 ///    __cudaRegisterFunction(GpuBinaryHandle,Kernel0,...);
 ///    ...
 ///    __cudaRegisterFunction(GpuBinaryHandle,KernelM,...);
+///    __cudaRegisterVar(GpuBinaryHandle, GlobalVar0, ...);
+///    ...
+///    __cudaRegisterVar(GpuBinaryHandle, GlobalVarN, ...);
 /// }
 /// \endcode
-llvm::Function *CGNVCUDARuntime::makeRegisterKernelsFn() {
+llvm::Function *CGNVCUDARuntime::makeRegisterGlobalsFn() {
+  // No need to register anything
+  if (EmittedKernels.empty() && DeviceVars.empty())
+    return nullptr;
+
   llvm::Function *RegisterKernelsFunc = llvm::Function::Create(
       llvm::FunctionType::get(VoidTy, VoidPtrPtrTy, false),
-      llvm::GlobalValue::InternalLinkage, "__cuda_register_kernels", &TheModule);
+      llvm::GlobalValue::InternalLinkage, "__cuda_register_globals", &TheModule);
   llvm::BasicBlock *EntryBB =
       llvm::BasicBlock::Create(Context, "entry", RegisterKernelsFunc);
   CGBuilderTy Builder(CGM, Context);
@@ -178,7 +189,7 @@ llvm::Function *CGNVCUDARuntime::makeRegisterKernelsFn() {
 
   // void __cudaRegisterFunction(void **, const char *, char *, const char *,
   //                             int, uint3*, uint3*, dim3*, dim3*, int*)
-  std::vector<llvm::Type *> RegisterFuncParams = {
+  llvm::Type *RegisterFuncParams[] = {
       VoidPtrPtrTy, CharPtrTy, CharPtrTy, CharPtrTy, IntTy,
       VoidPtrTy,    VoidPtrTy, VoidPtrTy, VoidPtrTy, IntTy->getPointerTo()};
   llvm::Constant *RegisterFunc = CGM.CreateRuntimeFunction(
@@ -186,18 +197,44 @@ llvm::Function *CGNVCUDARuntime::makeRegisterKernelsFn() {
       "__cudaRegisterFunction");
 
   // Extract GpuBinaryHandle passed as the first argument passed to
-  // __cuda_register_kernels() and generate __cudaRegisterFunction() call for
+  // __cuda_register_globals() and generate __cudaRegisterFunction() call for
   // each emitted kernel.
   llvm::Argument &GpuBinaryHandlePtr = *RegisterKernelsFunc->arg_begin();
   for (llvm::Function *Kernel : EmittedKernels) {
     llvm::Constant *KernelName = makeConstantString(Kernel->getName());
     llvm::Constant *NullPtr = llvm::ConstantPointerNull::get(VoidPtrTy);
-    llvm::Value *args[] = {
+    llvm::Value *Args[] = {
         &GpuBinaryHandlePtr, Builder.CreateBitCast(Kernel, VoidPtrTy),
         KernelName, KernelName, llvm::ConstantInt::get(IntTy, -1), NullPtr,
         NullPtr, NullPtr, NullPtr,
         llvm::ConstantPointerNull::get(IntTy->getPointerTo())};
-    Builder.CreateCall(RegisterFunc, args);
+    Builder.CreateCall(RegisterFunc, Args);
+  }
+
+  // void __cudaRegisterVar(void **, char *, char *, const char *,
+  //                        int, int, int, int)
+  llvm::Type *RegisterVarParams[] = {VoidPtrPtrTy, CharPtrTy, CharPtrTy,
+                                     CharPtrTy,    IntTy,     IntTy,
+                                     IntTy,        IntTy};
+  llvm::Constant *RegisterVar = CGM.CreateRuntimeFunction(
+      llvm::FunctionType::get(IntTy, RegisterVarParams, false),
+      "__cudaRegisterVar");
+  for (auto &Pair : DeviceVars) {
+    llvm::GlobalVariable *Var = Pair.first;
+    unsigned Flags = Pair.second;
+    llvm::Constant *VarName = makeConstantString(Var->getName());
+    uint64_t VarSize =
+        CGM.getDataLayout().getTypeAllocSize(Var->getValueType());
+    llvm::Value *Args[] = {
+        &GpuBinaryHandlePtr,
+        Builder.CreateBitCast(Var, VoidPtrTy),
+        VarName,
+        VarName,
+        llvm::ConstantInt::get(IntTy, (Flags & ExternDeviceVar) ? 1 : 0),
+        llvm::ConstantInt::get(IntTy, VarSize),
+        llvm::ConstantInt::get(IntTy, (Flags & ConstantDeviceVar) ? 1 : 0),
+        llvm::ConstantInt::get(IntTy, 0)};
+    Builder.CreateCall(RegisterVar, Args);
   }
 
   Builder.CreateRetVoid();
@@ -208,15 +245,19 @@ llvm::Function *CGNVCUDARuntime::makeRegisterKernelsFn() {
 /// \code
 /// void __cuda_module_ctor(void*) {
 ///     Handle0 = __cudaRegisterFatBinary(GpuBinaryBlob0);
-///     __cuda_register_kernels(Handle0);
+///     __cuda_register_globals(Handle0);
 ///     ...
 ///     HandleN = __cudaRegisterFatBinary(GpuBinaryBlobN);
-///     __cuda_register_kernels(HandleN);
+///     __cuda_register_globals(HandleN);
 /// }
 /// \endcode
 llvm::Function *CGNVCUDARuntime::makeModuleCtorFunction() {
-  // void __cuda_register_kernels(void* handle);
-  llvm::Function *RegisterKernelsFunc = makeRegisterKernelsFn();
+  // No need to generate ctors/dtors if there are no GPU binaries.
+  if (CGM.getCodeGenOpts().CudaGpuBinaryFileNames.empty())
+    return nullptr;
+
+  // void __cuda_register_globals(void* handle);
+  llvm::Function *RegisterGlobalsFunc = makeRegisterGlobalsFn();
   // void ** __cudaRegisterFatBinary(void *);
   llvm::Constant *RegisterFatbinFunc = CGM.CreateRuntimeFunction(
       llvm::FunctionType::get(VoidPtrPtrTy, VoidPtrTy, false),
@@ -259,6 +300,8 @@ llvm::Function *CGNVCUDARuntime::makeModuleCtorFunction() {
         TheModule, FatbinWrapperTy, true, llvm::GlobalValue::InternalLinkage,
         llvm::ConstantStruct::get(FatbinWrapperTy, Values),
         "__cuda_fatbin_wrapper");
+    // NVIDIA's cuobjdump looks for fatbins in this section.
+    FatbinWrapper->setSection(".nvFatBinSegment");
 
     // GpuBinaryHandle = __cudaRegisterFatBinary(&FatbinWrapper);
     llvm::CallInst *RegisterFatbinCall = CtorBuilder.CreateCall(
@@ -270,8 +313,9 @@ llvm::Function *CGNVCUDARuntime::makeModuleCtorFunction() {
     CtorBuilder.CreateAlignedStore(RegisterFatbinCall, GpuBinaryHandle,
                                    CGM.getPointerAlign());
 
-    // Call __cuda_register_kernels(GpuBinaryHandle);
-    CtorBuilder.CreateCall(RegisterKernelsFunc, RegisterFatbinCall);
+    // Call __cuda_register_globals(GpuBinaryHandle);
+    if (RegisterGlobalsFunc)
+      CtorBuilder.CreateCall(RegisterGlobalsFunc, RegisterFatbinCall);
 
     // Save GpuBinaryHandle so we can unregister it in destructor.
     GpuBinaryHandles.push_back(GpuBinaryHandle);
@@ -291,6 +335,10 @@ llvm::Function *CGNVCUDARuntime::makeModuleCtorFunction() {
 /// }
 /// \endcode
 llvm::Function *CGNVCUDARuntime::makeModuleDtorFunction() {
+  // No need for destructor if we don't have handles to unregister.
+  if (GpuBinaryHandles.empty())
+    return nullptr;
+
   // void __cudaUnregisterFatBinary(void ** handle);
   llvm::Constant *UnregisterFatbinFunc = CGM.CreateRuntimeFunction(
       llvm::FunctionType::get(VoidTy, VoidPtrPtrTy, false),
diff --git a/lib/CodeGen/CGCUDARuntime.h b/lib/CodeGen/CGCUDARuntime.h
index dcacf9703277..0168f4f9e942 100644
--- a/lib/CodeGen/CGCUDARuntime.h
+++ b/lib/CodeGen/CGCUDARuntime.h
@@ -18,6 +18,7 @@
 
 namespace llvm {
 class Function;
+class GlobalVariable;
 }
 
 namespace clang {
@@ -37,6 +38,12 @@ protected:
   CodeGenModule &CGM;
 
 public:
+  // Global variable properties that must be passed to CUDA runtime.
+  enum DeviceVarFlags {
+    ExternDeviceVar = 0x01,   // extern
+    ConstantDeviceVar = 0x02, // __constant__
+  };
+
   CGCUDARuntime(CodeGenModule &CGM) : CGM(CGM) {}
   virtual ~CGCUDARuntime();
 
@@ -46,6 +53,7 @@ public:
 
   /// Emits a kernel launch stub.
   virtual void emitDeviceStub(CodeGenFunction &CGF, FunctionArgList &Args) = 0;
+  virtual void registerDeviceVar(llvm::GlobalVariable &Var, unsigned Flags) = 0;
 
   /// Constructs and returns a module initialization function or nullptr if it's
   /// not needed. Must be called after all kernels have been emitted.
diff --git a/lib/CodeGen/CGCXX.cpp b/lib/CodeGen/CGCXX.cpp
index 6847df9b749b..40f1bc426ff7 100644
--- a/lib/CodeGen/CGCXX.cpp
+++ b/lib/CodeGen/CGCXX.cpp
@@ -164,7 +164,7 @@ bool CodeGenModule::TryEmitDefinitionAsAlias(GlobalDecl AliasDecl,
     // members with attribute "AlwaysInline" and expect no reference to
     // be generated. It is desirable to reenable this optimisation after
     // corresponding LLVM changes.
-    Replacements[MangledName] = Aliasee;
+    addReplacement(MangledName, Aliasee);
     return false;
   }
 
diff --git a/lib/CodeGen/CGCXXABI.h b/lib/CodeGen/CGCXXABI.h
index 3f240b1802b8..9e10ec068e09 100644
--- a/lib/CodeGen/CGCXXABI.h
+++ b/lib/CodeGen/CGCXXABI.h
@@ -106,6 +106,16 @@ public:
 
   virtual bool hasMostDerivedReturn(GlobalDecl GD) const { return false; }
 
+  /// Returns true if the target allows calling a function through a pointer
+  /// with a different signature than the actual function (or equivalently,
+  /// bitcasting a function or function pointer to a different function type).
+  /// In principle in the most general case this could depend on the target, the
+  /// calling convention, and the actual types of the arguments and return
+  /// value. Here it just means whether the signature mismatch could *ever* be
+  /// allowed; in other words, does the target do strict checking of signatures
+  /// for all calls.
+  virtual bool canCallMismatchedFunctionType() const { return true; }
+
   /// If the C++ ABI requires the given type be returned in a particular way,
   /// this method sets RetAI and returns true.
   virtual bool classifyReturnType(CGFunctionInfo &FI) const = 0;
@@ -326,6 +336,12 @@ public:
   virtual void addImplicitStructorParams(CodeGenFunction &CGF, QualType &ResTy,
                                          FunctionArgList &Params) = 0;
 
+  /// Get the ABI-specific "this" parameter adjustment to apply in the prologue
+  /// of a virtual function.
+  virtual CharUnits getVirtualFunctionPrologueThisAdjustment(GlobalDecl GD) {
+    return CharUnits::Zero();
+  }
+
   /// Perform ABI-specific "this" parameter adjustment in a virtual function
   /// prologue.
   virtual llvm::Value *adjustThisParameterInVirtualFunctionPrologue(
diff --git a/lib/CodeGen/CGCall.cpp b/lib/CodeGen/CGCall.cpp
index 935985049c01..242b5962070a 100644
--- a/lib/CodeGen/CGCall.cpp
+++ b/lib/CodeGen/CGCall.cpp
@@ -14,6 +14,7 @@
 
 #include "CGCall.h"
 #include "ABIInfo.h"
+#include "CGBlocks.h"
 #include "CGCXXABI.h"
 #include "CGCleanup.h"
 #include "CodeGenFunction.h"
@@ -25,9 +26,11 @@
 #include "clang/Basic/TargetBuiltins.h"
 #include "clang/Basic/TargetInfo.h"
 #include "clang/CodeGen/CGFunctionInfo.h"
+#include "clang/CodeGen/SwiftCallingConv.h"
 #include "clang/Frontend/CodeGenOptions.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/IR/Attributes.h"
+#include "llvm/IR/CallingConv.h"
 #include "llvm/IR/CallSite.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/InlineAsm.h"
@@ -39,7 +42,7 @@ using namespace CodeGen;
 
 /***/
 
-static unsigned ClangCallConvToLLVMCallConv(CallingConv CC) {
+unsigned CodeGenTypes::ClangCallConvToLLVMCallConv(CallingConv CC) {
   switch (CC) {
   default: return llvm::CallingConv::C;
   case CC_X86StdCall: return llvm::CallingConv::X86_StdCall;
@@ -55,7 +58,10 @@ static unsigned ClangCallConvToLLVMCallConv(CallingConv CC) {
   // TODO: Add support for __vectorcall to LLVM.
   case CC_X86VectorCall: return llvm::CallingConv::X86_VectorCall;
   case CC_SpirFunction: return llvm::CallingConv::SPIR_FUNC;
-  case CC_SpirKernel: return llvm::CallingConv::SPIR_KERNEL;
+  case CC_OpenCLKernel: return CGM.getTargetCodeGenInfo().getOpenCLKernelCallingConv();
+  case CC_PreserveMost: return llvm::CallingConv::PreserveMost;
+  case CC_PreserveAll: return llvm::CallingConv::PreserveAll;
+  case CC_Swift: return llvm::CallingConv::Swift;
   }
 }
 
@@ -90,15 +96,25 @@ CodeGenTypes::arrangeFreeFunctionType(CanQual<FunctionNoProtoType> FTNP) {
   return arrangeLLVMFunctionInfo(FTNP->getReturnType().getUnqualifiedType(),
                                  /*instanceMethod=*/false,
                                  /*chainCall=*/false, None,
-                                 FTNP->getExtInfo(), RequiredArgs(0));
+                                 FTNP->getExtInfo(), {}, RequiredArgs(0));
 }
 
 /// Adds the formal paramaters in FPT to the given prefix. If any parameter in
 /// FPT has pass_object_size attrs, then we'll add parameters for those, too.
 static void appendParameterTypes(const CodeGenTypes &CGT,
                                  SmallVectorImpl<CanQualType> &prefix,
-                                 const CanQual<FunctionProtoType> &FPT,
+              SmallVectorImpl<FunctionProtoType::ExtParameterInfo> &paramInfos,
+                                 CanQual<FunctionProtoType> FPT,
                                  const FunctionDecl *FD) {
+  // Fill out paramInfos.
+  if (FPT->hasExtParameterInfos() || !paramInfos.empty()) {
+    assert(paramInfos.size() <= prefix.size());
+    auto protoParamInfos = FPT->getExtParameterInfos();
+    paramInfos.reserve(prefix.size() + protoParamInfos.size());
+    paramInfos.resize(prefix.size());
+    paramInfos.append(protoParamInfos.begin(), protoParamInfos.end());
+  }
+
   // Fast path: unknown target.
   if (FD == nullptr) {
     prefix.append(FPT->param_type_begin(), FPT->param_type_end());
@@ -125,13 +141,17 @@ arrangeLLVMFunctionInfo(CodeGenTypes &CGT, bool instanceMethod,
                         SmallVectorImpl<CanQualType> &prefix,
                         CanQual<FunctionProtoType> FTP,
                         const FunctionDecl *FD) {
-  RequiredArgs required = RequiredArgs::forPrototypePlus(FTP, prefix.size());
+  SmallVector<FunctionProtoType::ExtParameterInfo, 16> paramInfos;
+  RequiredArgs Required =
+      RequiredArgs::forPrototypePlus(FTP, prefix.size(), FD);
   // FIXME: Kill copy.
-  appendParameterTypes(CGT, prefix, FTP, FD);
+  appendParameterTypes(CGT, prefix, paramInfos, FTP, FD);
   CanQualType resultType = FTP->getReturnType().getUnqualifiedType();
+
   return CGT.arrangeLLVMFunctionInfo(resultType, instanceMethod,
                                      /*chainCall=*/false, prefix,
-                                     FTP->getExtInfo(), required);
+                                     FTP->getExtInfo(), paramInfos,
+                                     Required);
 }
 
 /// Arrange the argument and result information for a value of the
@@ -173,6 +193,12 @@ static CallingConv getCallingConventionForDecl(const Decl *D, bool IsWindows) {
   if (D->hasAttr<SysVABIAttr>())
     return IsWindows ? CC_X86_64SysV : CC_C;
 
+  if (D->hasAttr<PreserveMostAttr>())
+    return CC_PreserveMost;
+
+  if (D->hasAttr<PreserveAllAttr>())
+    return CC_PreserveAll;
+
   return CC_C;
 }
 
@@ -219,16 +245,33 @@ CodeGenTypes::arrangeCXXMethodDeclaration(const CXXMethodDecl *MD) {
   return arrangeFreeFunctionType(prototype, MD);
 }
 
+bool CodeGenTypes::inheritingCtorHasParams(
+    const InheritedConstructor &Inherited, CXXCtorType Type) {
+  // Parameters are unnecessary if we're constructing a base class subobject
+  // and the inherited constructor lives in a virtual base.
+  return Type == Ctor_Complete ||
+         !Inherited.getShadowDecl()->constructsVirtualBase() ||
+         !Target.getCXXABI().hasConstructorVariants();
+  }
+
 const CGFunctionInfo &
 CodeGenTypes::arrangeCXXStructorDeclaration(const CXXMethodDecl *MD,
                                             StructorType Type) {
 
   SmallVector<CanQualType, 16> argTypes;
+  SmallVector<FunctionProtoType::ExtParameterInfo, 16> paramInfos;
   argTypes.push_back(GetThisType(Context, MD->getParent()));
 
+  bool PassParams = true;
+
   GlobalDecl GD;
   if (auto *CD = dyn_cast<CXXConstructorDecl>(MD)) {
     GD = GlobalDecl(CD, toCXXCtorType(Type));
+
+    // A base class inheriting constructor doesn't get forwarded arguments
+    // needed to construct a virtual base (or base class thereof).
+    if (auto Inherited = CD->getInheritedConstructor())
+      PassParams = inheritingCtorHasParams(Inherited, toCXXCtorType(Type));
   } else {
     auto *DD = dyn_cast<CXXDestructorDecl>(MD);
     GD = GlobalDecl(DD, toCXXDtorType(Type));
@@ -237,12 +280,14 @@ CodeGenTypes::arrangeCXXStructorDeclaration(const CXXMethodDecl *MD,
   CanQual<FunctionProtoType> FTP = GetFormalType(MD);
 
   // Add the formal parameters.
-  appendParameterTypes(*this, argTypes, FTP, MD);
+  if (PassParams)
+    appendParameterTypes(*this, argTypes, paramInfos, FTP, MD);
 
   TheCXXABI.buildStructorSignature(MD, Type, argTypes);
 
   RequiredArgs required =
-      (MD->isVariadic() ? RequiredArgs(argTypes.size()) : RequiredArgs::All);
+      (PassParams && MD->isVariadic() ? RequiredArgs(argTypes.size())
+                                      : RequiredArgs::All);
 
   FunctionType::ExtInfo extInfo = FTP->getExtInfo();
   CanQualType resultType = TheCXXABI.HasThisReturn(GD)
@@ -252,7 +297,53 @@ CodeGenTypes::arrangeCXXStructorDeclaration(const CXXMethodDecl *MD,
                                      : Context.VoidTy;
   return arrangeLLVMFunctionInfo(resultType, /*instanceMethod=*/true,
                                  /*chainCall=*/false, argTypes, extInfo,
-                                 required);
+                                 paramInfos, required);
+}
+
+static SmallVector<CanQualType, 16>
+getArgTypesForCall(ASTContext &ctx, const CallArgList &args) {
+  SmallVector<CanQualType, 16> argTypes;
+  for (auto &arg : args)
+    argTypes.push_back(ctx.getCanonicalParamType(arg.Ty));
+  return argTypes;
+}
+
+static SmallVector<CanQualType, 16>
+getArgTypesForDeclaration(ASTContext &ctx, const FunctionArgList &args) {
+  SmallVector<CanQualType, 16> argTypes;
+  for (auto &arg : args)
+    argTypes.push_back(ctx.getCanonicalParamType(arg->getType()));
+  return argTypes;
+}
+
+static void addExtParameterInfosForCall(
+         llvm::SmallVectorImpl<FunctionProtoType::ExtParameterInfo> &paramInfos,
+                                        const FunctionProtoType *proto,
+                                        unsigned prefixArgs,
+                                        unsigned totalArgs) {
+  assert(proto->hasExtParameterInfos());
+  assert(paramInfos.size() <= prefixArgs);
+  assert(proto->getNumParams() + prefixArgs <= totalArgs);
+
+  // Add default infos for any prefix args that don't already have infos.
+  paramInfos.resize(prefixArgs);
+
+  // Add infos for the prototype.
+  auto protoInfos = proto->getExtParameterInfos();
+  paramInfos.append(protoInfos.begin(), protoInfos.end());
+
+  // Add default infos for the variadic arguments.
+  paramInfos.resize(totalArgs);
+}
+
+static llvm::SmallVector<FunctionProtoType::ExtParameterInfo, 16>
+getExtParameterInfosForCall(const FunctionProtoType *proto,
+                            unsigned prefixArgs, unsigned totalArgs) {
+  llvm::SmallVector<FunctionProtoType::ExtParameterInfo, 16> result;
+  if (proto->hasExtParameterInfos()) {
+    addExtParameterInfosForCall(result, proto, prefixArgs, totalArgs);
+  }
+  return result;
 }
 
 /// Arrange a call to a C++ method, passing the given arguments.
@@ -267,7 +358,7 @@ CodeGenTypes::arrangeCXXConstructorCall(const CallArgList &args,
     ArgTypes.push_back(Context.getCanonicalParamType(Arg.Ty));
 
   CanQual<FunctionProtoType> FPT = GetFormalType(D);
-  RequiredArgs Required = RequiredArgs::forPrototypePlus(FPT, 1 + ExtraArgs);
+  RequiredArgs Required = RequiredArgs::forPrototypePlus(FPT, 1 + ExtraArgs, D);
   GlobalDecl GD(D, CtorKind);
   CanQualType ResultType = TheCXXABI.HasThisReturn(GD)
                                ? ArgTypes.front()
@@ -276,9 +367,11 @@ CodeGenTypes::arrangeCXXConstructorCall(const CallArgList &args,
                                      : Context.VoidTy;
 
   FunctionType::ExtInfo Info = FPT->getExtInfo();
+  auto ParamInfos = getExtParameterInfosForCall(FPT.getTypePtr(), 1 + ExtraArgs,
+                                                ArgTypes.size());
   return arrangeLLVMFunctionInfo(ResultType, /*instanceMethod=*/true,
                                  /*chainCall=*/false, ArgTypes, Info,
-                                 Required);
+                                 ParamInfos, Required);
 }
 
 /// Arrange the argument and result information for the declaration or
@@ -299,7 +392,7 @@ CodeGenTypes::arrangeFunctionDeclaration(const FunctionDecl *FD) {
     CanQual<FunctionNoProtoType> noProto = FTy.getAs<FunctionNoProtoType>();
     return arrangeLLVMFunctionInfo(
         noProto->getReturnType(), /*instanceMethod=*/false,
-        /*chainCall=*/false, None, noProto->getExtInfo(), RequiredArgs::All);
+        /*chainCall=*/false, None, noProto->getExtInfo(), {},RequiredArgs::All);
   }
 
   assert(isa<FunctionProtoType>(FTy));
@@ -328,7 +421,7 @@ CodeGenTypes::arrangeObjCMessageSendSignature(const ObjCMethodDecl *MD,
   argTys.push_back(Context.getCanonicalParamType(receiverType));
   argTys.push_back(Context.getCanonicalParamType(Context.getObjCSelType()));
   // FIXME: Kill copy?
-  for (const auto *I : MD->params()) {
+  for (const auto *I : MD->parameters()) {
     argTys.push_back(Context.getCanonicalParamType(I->getType()));
   }
 
@@ -345,7 +438,18 @@ CodeGenTypes::arrangeObjCMessageSendSignature(const ObjCMethodDecl *MD,
 
   return arrangeLLVMFunctionInfo(
       GetReturnType(MD->getReturnType()), /*instanceMethod=*/false,
-      /*chainCall=*/false, argTys, einfo, required);
+      /*chainCall=*/false, argTys, einfo, {}, required);
+}
+
+const CGFunctionInfo &
+CodeGenTypes::arrangeUnprototypedObjCMessageSend(QualType returnType,
+                                                 const CallArgList &args) {
+  auto argTypes = getArgTypesForCall(Context, args);
+  FunctionType::ExtInfo einfo;
+
+  return arrangeLLVMFunctionInfo(
+      GetReturnType(returnType), /*instanceMethod=*/false,
+      /*chainCall=*/false, argTypes, einfo, {}, RequiredArgs::All);
 }
 
 const CGFunctionInfo &
@@ -374,7 +478,7 @@ CodeGenTypes::arrangeMSMemberPointerThunk(const CXXMethodDecl *MD) {
   CanQualType ArgTys[] = { GetThisType(Context, MD->getParent()) };
   return arrangeLLVMFunctionInfo(Context.VoidTy, /*instanceMethod=*/false,
                                  /*chainCall=*/false, ArgTys,
-                                 FTP->getExtInfo(), RequiredArgs(1));
+                                 FTP->getExtInfo(), {}, RequiredArgs(1));
 }
 
 const CGFunctionInfo &
@@ -394,7 +498,8 @@ CodeGenTypes::arrangeMSCtorClosure(const CXXConstructorDecl *CD,
       /*IsVariadic=*/false, /*IsCXXMethod=*/true);
   return arrangeLLVMFunctionInfo(Context.VoidTy, /*instanceMethod=*/true,
                                  /*chainCall=*/false, ArgTys,
-                                 FunctionType::ExtInfo(CC), RequiredArgs::All);
+                                 FunctionType::ExtInfo(CC), {},
+                                 RequiredArgs::All);
 }
 
 /// Arrange a call as unto a free function, except possibly with an
@@ -408,6 +513,8 @@ arrangeFreeFunctionLikeCall(CodeGenTypes &CGT,
                             bool chainCall) {
   assert(args.size() >= numExtraRequiredArgs);
 
+  llvm::SmallVector<FunctionProtoType::ExtParameterInfo, 16> paramInfos;
+
   // In most cases, there are no optional arguments.
   RequiredArgs required = RequiredArgs::All;
 
@@ -417,6 +524,10 @@ arrangeFreeFunctionLikeCall(CodeGenTypes &CGT,
     if (proto->isVariadic())
       required = RequiredArgs(proto->getNumParams() + numExtraRequiredArgs);
 
+    if (proto->hasExtParameterInfos())
+      addExtParameterInfosForCall(paramInfos, proto, numExtraRequiredArgs,
+                                  args.size());
+
   // If we don't have a prototype at all, but we're supposed to
   // explicitly use the variadic convention for unprototyped calls,
   // treat all of the arguments as required but preserve the nominal
@@ -433,7 +544,8 @@ arrangeFreeFunctionLikeCall(CodeGenTypes &CGT,
     argTypes.push_back(CGT.getContext().getCanonicalParamType(arg.Ty));
   return CGT.arrangeLLVMFunctionInfo(GetReturnType(fnType->getReturnType()),
                                      /*instanceMethod=*/false, chainCall,
-                                     argTypes, fnType->getExtInfo(), required);
+                                     argTypes, fnType->getExtInfo(), paramInfos,
+                                     required);
 }
 
 /// Figure out the rules for calling a function with the given formal
@@ -448,7 +560,7 @@ CodeGenTypes::arrangeFreeFunctionCall(const CallArgList &args,
                                      chainCall ? 1 : 0, chainCall);
 }
 
-/// A block function call is essentially a free-function call with an
+/// A block function is essentially a free function with an
 /// extra implicit argument.
 const CGFunctionInfo &
 CodeGenTypes::arrangeBlockFunctionCall(const CallArgList &args,
@@ -458,54 +570,99 @@ CodeGenTypes::arrangeBlockFunctionCall(const CallArgList &args,
 }
 
 const CGFunctionInfo &
-CodeGenTypes::arrangeFreeFunctionCall(QualType resultType,
-                                      const CallArgList &args,
-                                      FunctionType::ExtInfo info,
-                                      RequiredArgs required) {
+CodeGenTypes::arrangeBlockFunctionDeclaration(const FunctionProtoType *proto,
+                                              const FunctionArgList &params) {
+  auto paramInfos = getExtParameterInfosForCall(proto, 1, params.size());
+  auto argTypes = getArgTypesForDeclaration(Context, params);
+
+  return arrangeLLVMFunctionInfo(
+      GetReturnType(proto->getReturnType()),
+      /*instanceMethod*/ false, /*chainCall*/ false, argTypes,
+      proto->getExtInfo(), paramInfos,
+      RequiredArgs::forPrototypePlus(proto, 1, nullptr));
+}
+
+const CGFunctionInfo &
+CodeGenTypes::arrangeBuiltinFunctionCall(QualType resultType,
+                                         const CallArgList &args) {
   // FIXME: Kill copy.
   SmallVector<CanQualType, 16> argTypes;
   for (const auto &Arg : args)
     argTypes.push_back(Context.getCanonicalParamType(Arg.Ty));
   return arrangeLLVMFunctionInfo(
       GetReturnType(resultType), /*instanceMethod=*/false,
-      /*chainCall=*/false, argTypes, info, required);
+      /*chainCall=*/false, argTypes, FunctionType::ExtInfo(),
+      /*paramInfos=*/ {}, RequiredArgs::All);
 }
 
-/// Arrange a call to a C++ method, passing the given arguments.
 const CGFunctionInfo &
-CodeGenTypes::arrangeCXXMethodCall(const CallArgList &args,
-                                   const FunctionProtoType *FPT,
-                                   RequiredArgs required) {
-  // FIXME: Kill copy.
-  SmallVector<CanQualType, 16> argTypes;
-  for (const auto &Arg : args)
-    argTypes.push_back(Context.getCanonicalParamType(Arg.Ty));
+CodeGenTypes::arrangeBuiltinFunctionDeclaration(QualType resultType,
+                                                const FunctionArgList &args) {
+  auto argTypes = getArgTypesForDeclaration(Context, args);
+
+  return arrangeLLVMFunctionInfo(
+      GetReturnType(resultType), /*instanceMethod=*/false, /*chainCall=*/false,
+      argTypes, FunctionType::ExtInfo(), {}, RequiredArgs::All);
+}
 
-  FunctionType::ExtInfo info = FPT->getExtInfo();
+const CGFunctionInfo &
+CodeGenTypes::arrangeBuiltinFunctionDeclaration(CanQualType resultType,
+                                              ArrayRef<CanQualType> argTypes) {
   return arrangeLLVMFunctionInfo(
-      GetReturnType(FPT->getReturnType()), /*instanceMethod=*/true,
-      /*chainCall=*/false, argTypes, info, required);
+      resultType, /*instanceMethod=*/false, /*chainCall=*/false,
+      argTypes, FunctionType::ExtInfo(), {}, RequiredArgs::All);
 }
 
-const CGFunctionInfo &CodeGenTypes::arrangeFreeFunctionDeclaration(
-    QualType resultType, const FunctionArgList &args,
-    const FunctionType::ExtInfo &info, bool isVariadic) {
+/// Arrange a call to a C++ method, passing the given arguments.
+const CGFunctionInfo &
+CodeGenTypes::arrangeCXXMethodCall(const CallArgList &args,
+                                   const FunctionProtoType *proto,
+                                   RequiredArgs required) {
+  unsigned numRequiredArgs =
+    (proto->isVariadic() ? required.getNumRequiredArgs() : args.size());
+  unsigned numPrefixArgs = numRequiredArgs - proto->getNumParams();
+  auto paramInfos =
+    getExtParameterInfosForCall(proto, numPrefixArgs, args.size());
+
   // FIXME: Kill copy.
-  SmallVector<CanQualType, 16> argTypes;
-  for (auto Arg : args)
-    argTypes.push_back(Context.getCanonicalParamType(Arg->getType()));
+  auto argTypes = getArgTypesForCall(Context, args);
 
-  RequiredArgs required =
-    (isVariadic ? RequiredArgs(args.size()) : RequiredArgs::All);
+  FunctionType::ExtInfo info = proto->getExtInfo();
   return arrangeLLVMFunctionInfo(
-      GetReturnType(resultType), /*instanceMethod=*/false,
-      /*chainCall=*/false, argTypes, info, required);
+      GetReturnType(proto->getReturnType()), /*instanceMethod=*/true,
+      /*chainCall=*/false, argTypes, info, paramInfos, required);
 }
 
 const CGFunctionInfo &CodeGenTypes::arrangeNullaryFunction() {
   return arrangeLLVMFunctionInfo(
       getContext().VoidTy, /*instanceMethod=*/false, /*chainCall=*/false,
-      None, FunctionType::ExtInfo(), RequiredArgs::All);
+      None, FunctionType::ExtInfo(), {}, RequiredArgs::All);
+}
+
+const CGFunctionInfo &
+CodeGenTypes::arrangeCall(const CGFunctionInfo &signature,
+                          const CallArgList &args) {
+  assert(signature.arg_size() <= args.size());
+  if (signature.arg_size() == args.size())
+    return signature;
+
+  SmallVector<FunctionProtoType::ExtParameterInfo, 16> paramInfos;
+  auto sigParamInfos = signature.getExtParameterInfos();
+  if (!sigParamInfos.empty()) {
+    paramInfos.append(sigParamInfos.begin(), sigParamInfos.end());
+    paramInfos.resize(args.size());
+  }
+
+  auto argTypes = getArgTypesForCall(Context, args);
+
+  assert(signature.getRequiredArgs().allowsOptionalArgs());
+  return arrangeLLVMFunctionInfo(signature.getReturnType(),
+                                 signature.isInstanceMethod(),
+                                 signature.isChainCall(),
+                                 argTypes,
+                                 signature.getExtInfo(),
+                                 paramInfos,
+                                 signature.getRequiredArgs());
 }
 
 /// Arrange the argument and result information for an abstract value
@@ -517,25 +674,26 @@ CodeGenTypes::arrangeLLVMFunctionInfo(CanQualType resultType,
                                       bool chainCall,
                                       ArrayRef<CanQualType> argTypes,
                                       FunctionType::ExtInfo info,
+                     ArrayRef<FunctionProtoType::ExtParameterInfo> paramInfos,
                                       RequiredArgs required) {
   assert(std::all_of(argTypes.begin(), argTypes.end(),
                      std::mem_fun_ref(&CanQualType::isCanonicalAsParam)));
 
-  unsigned CC = ClangCallConvToLLVMCallConv(info.getCC());
-
   // Lookup or create unique function info.
   llvm::FoldingSetNodeID ID;
-  CGFunctionInfo::Profile(ID, instanceMethod, chainCall, info, required,
-                          resultType, argTypes);
+  CGFunctionInfo::Profile(ID, instanceMethod, chainCall, info, paramInfos,
+                          required, resultType, argTypes);
 
   void *insertPos = nullptr;
   CGFunctionInfo *FI = FunctionInfos.FindNodeOrInsertPos(ID, insertPos);
   if (FI)
     return *FI;
 
+  unsigned CC = ClangCallConvToLLVMCallConv(info.getCC());
+
   // Construct the function info.  We co-allocate the ArgInfos.
   FI = CGFunctionInfo::create(CC, instanceMethod, chainCall, info,
-                              resultType, argTypes, required);
+                              paramInfos, resultType, argTypes, required);
   FunctionInfos.InsertNode(FI, insertPos);
 
   bool inserted = FunctionsBeingProcessed.insert(FI).second;
@@ -543,7 +701,11 @@ CodeGenTypes::arrangeLLVMFunctionInfo(CanQualType resultType,
   assert(inserted && "Recursively being processed?");
   
   // Compute ABI information.
-  getABIInfo().computeInfo(*FI);
+  if (info.getCC() != CC_Swift) {
+    getABIInfo().computeInfo(*FI);
+  } else {
+    swiftcall::computeABIInfo(CGM, *FI);
+  }
 
   // Loop over all of the computed argument and return value info.  If any of
   // them are direct or extend without a specified coerce type, specify the
@@ -566,11 +728,16 @@ CGFunctionInfo *CGFunctionInfo::create(unsigned llvmCC,
                                        bool instanceMethod,
                                        bool chainCall,
                                        const FunctionType::ExtInfo &info,
+                                       ArrayRef<ExtParameterInfo> paramInfos,
                                        CanQualType resultType,
                                        ArrayRef<CanQualType> argTypes,
                                        RequiredArgs required) {
-  void *buffer = operator new(sizeof(CGFunctionInfo) +
-                              sizeof(ArgInfo) * (argTypes.size() + 1));
+  assert(paramInfos.empty() || paramInfos.size() == argTypes.size());
+
+  void *buffer =
+    operator new(totalSizeToAlloc<ArgInfo,             ExtParameterInfo>(
+                                  argTypes.size() + 1, paramInfos.size()));
+
   CGFunctionInfo *FI = new(buffer) CGFunctionInfo();
   FI->CallingConvention = llvmCC;
   FI->EffectiveCallingConvention = llvmCC;
@@ -585,9 +752,12 @@ CGFunctionInfo *CGFunctionInfo::create(unsigned llvmCC,
   FI->ArgStruct = nullptr;
   FI->ArgStructAlign = 0;
   FI->NumArgs = argTypes.size();
+  FI->HasExtParameterInfos = !paramInfos.empty();
   FI->getArgsBuffer()[0].type = resultType;
   for (unsigned i = 0, e = argTypes.size(); i != e; ++i)
     FI->getArgsBuffer()[i + 1].type = argTypes[i];
+  for (unsigned i = 0, e = paramInfos.size(); i != e; ++i)
+    FI->getExtParameterInfosBuffer()[i] = paramInfos[i];
   return FI;
 }
 
@@ -634,7 +804,8 @@ struct RecordExpansion : TypeExpansion {
 
   RecordExpansion(SmallVector<const CXXBaseSpecifier *, 1> &&Bases,
                   SmallVector<const FieldDecl *, 1> &&Fields)
-      : TypeExpansion(TEK_Record), Bases(Bases), Fields(Fields) {}
+      : TypeExpansion(TEK_Record), Bases(std::move(Bases)),
+        Fields(std::move(Fields)) {}
   static bool classof(const TypeExpansion *TE) {
     return TE->Kind == TEK_Record;
   }
@@ -773,7 +944,7 @@ static void forConstantArrayExpansion(CodeGenFunction &CGF,
 }
 
 void CodeGenFunction::ExpandTypeFromArgs(
-    QualType Ty, LValue LV, SmallVectorImpl<llvm::Argument *>::iterator &AI) {
+    QualType Ty, LValue LV, SmallVectorImpl<llvm::Value *>::iterator &AI) {
   assert(LV.isSimple() &&
          "Unexpected non-simple lvalue during struct expansion.");
 
@@ -798,7 +969,7 @@ void CodeGenFunction::ExpandTypeFromArgs(
     }
     for (auto FD : RExp->Fields) {
       // FIXME: What are the right qualifiers here?
-      LValue SubLV = EmitLValueForField(LV, FD);
+      LValue SubLV = EmitLValueForFieldInitialization(LV, FD);
       ExpandTypeFromArgs(FD->getType(), SubLV, AI);
     }
   } else if (isa<ComplexExpansion>(Exp.get())) {
@@ -1220,11 +1391,13 @@ void ClangToLLVMArgMapping::construct(const ASTContext &Context,
       // ignore and inalloca doesn't have matching LLVM parameters.
       IRArgs.NumberOfArgs = 0;
       break;
-    case ABIArgInfo::Expand: {
+    case ABIArgInfo::CoerceAndExpand:
+      IRArgs.NumberOfArgs = AI.getCoerceAndExpandTypeSequence().size();
+      break;
+    case ABIArgInfo::Expand:
       IRArgs.NumberOfArgs = getExpansionSize(ArgType, Context);
       break;
     }
-    }
 
     if (IRArgs.NumberOfArgs > 0) {
       IRArgs.FirstArgIndex = IRArgNo;
@@ -1323,6 +1496,10 @@ CodeGenTypes::GetFunctionType(const CGFunctionInfo &FI) {
   case ABIArgInfo::Ignore:
     resultType = llvm::Type::getVoidTy(getLLVMContext());
     break;
+
+  case ABIArgInfo::CoerceAndExpand:
+    resultType = retAI.getUnpaddedCoerceAndExpandType();
+    break;
   }
 
   ClangToLLVMArgMapping IRFunctionArgs(getContext(), FI, true);
@@ -1390,6 +1567,15 @@ CodeGenTypes::GetFunctionType(const CGFunctionInfo &FI) {
       break;
     }
 
+    case ABIArgInfo::CoerceAndExpand: {
+      auto ArgTypesIter = ArgTypes.begin() + FirstIRArg;
+      for (auto EltTy : ArgInfo.getCoerceAndExpandTypeSequence()) {
+        *ArgTypesIter++ = EltTy;
+      }
+      assert(ArgTypesIter == ArgTypes.begin() + FirstIRArg + NumIRArgs);
+      break;
+    }
+
     case ABIArgInfo::Expand:
       auto ArgTypesIter = ArgTypes.begin() + FirstIRArg;
       getExpandedTypes(it->type, ArgTypesIter);
@@ -1450,6 +1636,7 @@ void CodeGenModule::ConstructAttributeList(
 
   const Decl *TargetDecl = CalleeInfo.getCalleeDecl();
 
+  bool HasAnyX86InterruptAttr = false;
   // FIXME: handle sseregparm someday...
   if (TargetDecl) {
     if (TargetDecl->hasAttr<ReturnsTwiceAttr>())
@@ -1487,6 +1674,7 @@ void CodeGenModule::ConstructAttributeList(
     if (TargetDecl->hasAttr<ReturnsNonNullAttr>())
       RetAttrs.addAttribute(llvm::Attribute::NonNull);
 
+    HasAnyX86InterruptAttr = TargetDecl->hasAttr<AnyX86InterruptAttr>();
     HasOptnone = TargetDecl->hasAttr<OptimizeNoneAttr>();
   }
 
@@ -1526,10 +1714,11 @@ void CodeGenModule::ConstructAttributeList(
     }
 
     bool DisableTailCalls =
-        CodeGenOpts.DisableTailCalls ||
+        CodeGenOpts.DisableTailCalls || HasAnyX86InterruptAttr ||
         (TargetDecl && TargetDecl->hasAttr<DisableTailCallsAttr>());
-    FuncAttrs.addAttribute("disable-tail-calls",
-                           llvm::toStringRef(DisableTailCalls));
+    FuncAttrs.addAttribute(
+        "disable-tail-calls",
+        llvm::toStringRef(DisableTailCalls));
 
     FuncAttrs.addAttribute("less-precise-fpmad",
                            llvm::toStringRef(CodeGenOpts.LessPreciseFPMAD));
@@ -1543,9 +1732,13 @@ void CodeGenModule::ConstructAttributeList(
                            llvm::toStringRef(CodeGenOpts.SoftFloat));
     FuncAttrs.addAttribute("stack-protector-buffer-size",
                            llvm::utostr(CodeGenOpts.SSPBufferSize));
+    FuncAttrs.addAttribute("no-signed-zeros-fp-math",
+                           llvm::toStringRef(CodeGenOpts.NoSignedZeros));
 
     if (CodeGenOpts.StackRealignment)
       FuncAttrs.addAttribute("stackrealign");
+    if (CodeGenOpts.Backchain)
+      FuncAttrs.addAttribute("backchain");
 
     // Add target-cpu and target-features attributes to functions. If
     // we have a decl for the function and it has a target attribute then
@@ -1594,6 +1787,18 @@ void CodeGenModule::ConstructAttributeList(
     }
   }
 
+  if (getLangOpts().CUDA && getLangOpts().CUDAIsDevice) {
+    // Conservatively, mark all functions and calls in CUDA as convergent
+    // (meaning, they may call an intrinsically convergent op, such as
+    // __syncthreads(), and so can't have certain optimizations applied around
+    // them).  LLVM will remove this attribute where it safely can.
+    FuncAttrs.addAttribute(llvm::Attribute::Convergent);
+
+    // Respect -fcuda-flush-denormals-to-zero.
+    if (getLangOpts().CUDADeviceFlushDenormalsToZero)
+      FuncAttrs.addAttribute("nvptx-f32ftz", "true");
+  }
+
   ClangToLLVMArgMapping IRFunctionArgs(getContext(), FI);
 
   QualType RetTy = FI.getReturnType();
@@ -1620,6 +1825,9 @@ void CodeGenModule::ConstructAttributeList(
     break;
   }
 
+  case ABIArgInfo::CoerceAndExpand:
+    break;
+
   case ABIArgInfo::Expand:
     llvm_unreachable("Invalid ABI kind for return argument");
   }
@@ -1639,10 +1847,13 @@ void CodeGenModule::ConstructAttributeList(
         getLLVMContext(), llvm::AttributeSet::ReturnIndex, RetAttrs));
   }
 
+  bool hasUsedSRet = false;
+
   // Attach attributes to sret.
   if (IRFunctionArgs.hasSRetArg()) {
     llvm::AttrBuilder SRETAttrs;
     SRETAttrs.addAttribute(llvm::Attribute::StructRet);
+    hasUsedSRet = true;
     if (RetAI.getInReg())
       SRETAttrs.addAttribute(llvm::Attribute::InReg);
     PAL.push_back(llvm::AttributeSet::get(
@@ -1727,7 +1938,8 @@ void CodeGenModule::ConstructAttributeList(
     }
     case ABIArgInfo::Ignore:
     case ABIArgInfo::Expand:
-      continue;
+    case ABIArgInfo::CoerceAndExpand:
+      break;
 
     case ABIArgInfo::InAlloca:
       // inalloca disables readnone and readonly.
@@ -1745,6 +1957,41 @@ void CodeGenModule::ConstructAttributeList(
         Attrs.addAttribute(llvm::Attribute::NonNull);
     }
 
+    switch (FI.getExtParameterInfo(ArgNo).getABI()) {
+    case ParameterABI::Ordinary:
+      break;
+
+    case ParameterABI::SwiftIndirectResult: {
+      // Add 'sret' if we haven't already used it for something, but
+      // only if the result is void.
+      if (!hasUsedSRet && RetTy->isVoidType()) {
+        Attrs.addAttribute(llvm::Attribute::StructRet);
+        hasUsedSRet = true;
+      }
+
+      // Add 'noalias' in either case.
+      Attrs.addAttribute(llvm::Attribute::NoAlias);
+
+      // Add 'dereferenceable' and 'alignment'.
+      auto PTy = ParamType->getPointeeType();
+      if (!PTy->isIncompleteType() && PTy->isConstantSizeType()) {
+        auto info = getContext().getTypeInfoInChars(PTy);
+        Attrs.addDereferenceableAttr(info.first.getQuantity());
+        Attrs.addAttribute(llvm::Attribute::getWithAlignment(getLLVMContext(),
+                                                 info.second.getQuantity()));
+      }
+      break;
+    }
+
+    case ParameterABI::SwiftErrorResult:
+      Attrs.addAttribute(llvm::Attribute::SwiftError);
+      break;
+
+    case ParameterABI::SwiftContext:
+      Attrs.addAttribute(llvm::Attribute::SwiftSelf);
+      break;
+    }
+
     if (Attrs.hasAttributes()) {
       unsigned FirstIRArg, NumIRArgs;
       std::tie(FirstIRArg, NumIRArgs) = IRFunctionArgs.getIRArgs(ArgNo);
@@ -1810,6 +2057,18 @@ static const NonNullAttr *getNonNullAttr(const Decl *FD, const ParmVarDecl *PVD,
   return nullptr;
 }
 
+namespace {
+  struct CopyBackSwiftError final : EHScopeStack::Cleanup {
+    Address Temp;
+    Address Arg;
+    CopyBackSwiftError(Address temp, Address arg) : Temp(temp), Arg(arg) {}
+    void Emit(CodeGenFunction &CGF, Flags flags) override {
+      llvm::Value *errorValue = CGF.Builder.CreateLoad(Temp);
+      CGF.Builder.CreateStore(errorValue, Arg);
+    }
+  };
+}
+
 void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
                                          llvm::Function *Fn,
                                          const FunctionArgList &Args) {
@@ -1835,7 +2094,7 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
 
   ClangToLLVMArgMapping IRFunctionArgs(CGM.getContext(), FI);
   // Flattened function arguments.
-  SmallVector<llvm::Argument *, 16> FnArgs;
+  SmallVector<llvm::Value *, 16> FnArgs;
   FnArgs.reserve(IRFunctionArgs.totalIRArgs());
   for (auto &Arg : Fn->args()) {
     FnArgs.push_back(&Arg);
@@ -1856,7 +2115,7 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
 
   // Name the struct return parameter.
   if (IRFunctionArgs.hasSRetArg()) {
-    auto AI = FnArgs[IRFunctionArgs.getSRetArgNo()];
+    auto AI = cast<llvm::Argument>(FnArgs[IRFunctionArgs.getSRetArgNo()]);
     AI->setName("agg.result");
     AI->addAttr(llvm::AttributeSet::get(getLLVMContext(), AI->getArgNo() + 1,
                                         llvm::Attribute::NoAlias));
@@ -1944,8 +2203,8 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
           ArgI.getCoerceToType() == ConvertType(Ty) &&
           ArgI.getDirectOffset() == 0) {
         assert(NumIRArgs == 1);
-        auto AI = FnArgs[FirstIRArg];
-        llvm::Value *V = AI;
+        llvm::Value *V = FnArgs[FirstIRArg];
+        auto AI = cast<llvm::Argument>(V);
 
         if (const ParmVarDecl *PVD = dyn_cast<ParmVarDecl>(Arg)) {
           if (getNonNullAttr(CurCodeDecl, PVD, PVD->getType(),
@@ -2014,6 +2273,25 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
                                               AI->getArgNo() + 1,
                                               llvm::Attribute::NoAlias));
 
+        // LLVM expects swifterror parameters to be used in very restricted
+        // ways.  Copy the value into a less-restricted temporary.
+        if (FI.getExtParameterInfo(ArgNo).getABI()
+              == ParameterABI::SwiftErrorResult) {
+          QualType pointeeTy = Ty->getPointeeType();
+          assert(pointeeTy->isPointerType());
+          Address temp =
+            CreateMemTemp(pointeeTy, getPointerAlign(), "swifterror.temp");
+          Address arg = Address(V, getContext().getTypeAlignInChars(pointeeTy));
+          llvm::Value *incomingErrorValue = Builder.CreateLoad(arg);
+          Builder.CreateStore(incomingErrorValue, temp);
+          V = temp.getPointer();
+
+          // Push a cleanup to copy the value back at the end of the function.
+          // The convention does not guarantee that the value will be written
+          // back if the function exits with an unwind exception.
+          EHStack.pushCleanup<CopyBackSwiftError>(NormalCleanup, temp, arg);
+        }
+
         // Ensure the argument is the correct type.
         if (V->getType() != ArgI.getCoerceToType())
           V = Builder.CreateBitCast(V, ArgI.getCoerceToType());
@@ -2100,6 +2378,29 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
       break;
     }
 
+    case ABIArgInfo::CoerceAndExpand: {
+      // Reconstruct into a temporary.
+      Address alloca = CreateMemTemp(Ty, getContext().getDeclAlign(Arg));
+      ArgVals.push_back(ParamValue::forIndirect(alloca));
+
+      auto coercionType = ArgI.getCoerceAndExpandType();
+      alloca = Builder.CreateElementBitCast(alloca, coercionType);
+      auto layout = CGM.getDataLayout().getStructLayout(coercionType);
+
+      unsigned argIndex = FirstIRArg;
+      for (unsigned i = 0, e = coercionType->getNumElements(); i != e; ++i) {
+        llvm::Type *eltType = coercionType->getElementType(i);
+        if (ABIArgInfo::isPaddingForCoerceAndExpand(eltType))
+          continue;
+
+        auto eltAddr = Builder.CreateStructGEP(alloca, i, layout);
+        auto elt = FnArgs[argIndex++];
+        Builder.CreateStore(elt, eltAddr);
+      }
+      assert(argIndex == FirstIRArg + NumIRArgs);
+      break;
+    }
+
     case ABIArgInfo::Expand: {
       // If this structure was expanded into multiple arguments then
       // we need to create a temporary and reconstruct it from the
@@ -2462,9 +2763,26 @@ void CodeGenFunction::EmitFunctionEpilog(const CGFunctionInfo &FI,
     // In ARC, end functions that return a retainable type with a call
     // to objc_autoreleaseReturnValue.
     if (AutoreleaseResult) {
+#ifndef NDEBUG
+      // Type::isObjCRetainabletype has to be called on a QualType that hasn't
+      // been stripped of the typedefs, so we cannot use RetTy here. Get the
+      // original return type of FunctionDecl, CurCodeDecl, and BlockDecl from
+      // CurCodeDecl or BlockInfo.
+      QualType RT;
+
+      if (auto *FD = dyn_cast<FunctionDecl>(CurCodeDecl))
+        RT = FD->getReturnType();
+      else if (auto *MD = dyn_cast<ObjCMethodDecl>(CurCodeDecl))
+        RT = MD->getReturnType();
+      else if (isa<BlockDecl>(CurCodeDecl))
+        RT = BlockInfo->BlockExpression->getFunctionType()->getReturnType();
+      else
+        llvm_unreachable("Unexpected function/method type");
+
       assert(getLangOpts().ObjCAutoRefCount &&
              !FI.isReturnsRetained() &&
-             RetTy->isObjCRetainableType());
+             RT->isObjCRetainableType());
+#endif
       RV = emitAutoreleaseOfResult(*this, RV);
     }
 
@@ -2473,6 +2791,40 @@ void CodeGenFunction::EmitFunctionEpilog(const CGFunctionInfo &FI,
   case ABIArgInfo::Ignore:
     break;
 
+  case ABIArgInfo::CoerceAndExpand: {
+    auto coercionType = RetAI.getCoerceAndExpandType();
+    auto layout = CGM.getDataLayout().getStructLayout(coercionType);
+
+    // Load all of the coerced elements out into results.
+    llvm::SmallVector<llvm::Value*, 4> results;
+    Address addr = Builder.CreateElementBitCast(ReturnValue, coercionType);
+    for (unsigned i = 0, e = coercionType->getNumElements(); i != e; ++i) {
+      auto coercedEltType = coercionType->getElementType(i);
+      if (ABIArgInfo::isPaddingForCoerceAndExpand(coercedEltType))
+        continue;
+
+      auto eltAddr = Builder.CreateStructGEP(addr, i, layout);
+      auto elt = Builder.CreateLoad(eltAddr);
+      results.push_back(elt);
+    }
+
+    // If we have one result, it's the single direct result type.
+    if (results.size() == 1) {
+      RV = results[0];
+
+    // Otherwise, we need to make a first-class aggregate.
+    } else {
+      // Construct a return type that lacks padding elements.
+      llvm::Type *returnType = RetAI.getUnpaddedCoerceAndExpandType();
+
+      RV = llvm::UndefValue::get(returnType);
+      for (unsigned i = 0, e = results.size(); i != e; ++i) {
+        RV = Builder.CreateInsertValue(RV, results[i], i);
+      }
+    }
+    break;
+  }
+
   case ABIArgInfo::Expand:
     llvm_unreachable("Invalid ABI kind for return argument");
   }
@@ -2536,23 +2888,15 @@ void CodeGenFunction::EmitDelegateCallArg(CallArgList &args,
 
   QualType type = param->getType();
 
-  // For the most part, we just need to load the alloca, except:
-  // 1) aggregate r-values are actually pointers to temporaries, and
-  // 2) references to non-scalars are pointers directly to the aggregate.
-  // I don't know why references to scalars are different here.
-  if (const ReferenceType *ref = type->getAs<ReferenceType>()) {
-    if (!hasScalarEvaluationKind(ref->getPointeeType()))
-      return args.add(RValue::getAggregate(local), type);
-
-    // Locals which are references to scalars are represented
-    // with allocas holding the pointer.
-    return args.add(RValue::get(Builder.CreateLoad(local)), type);
-  }
-
   assert(!isInAllocaArgument(CGM.getCXXABI(), type) &&
          "cannot emit delegate call arguments for inalloca arguments!");
 
-  args.add(convertTempToRValue(local, type, loc), type);
+  // For the most part, we just need to load the alloca, except that
+  // aggregate r-values are actually pointers to temporaries.
+  if (type->isReferenceType())
+    args.add(RValue::get(Builder.CreateLoad(local)), type);
+  else
+    args.add(convertTempToRValue(local, type, loc), type);
 }
 
 static bool isProvablyNull(llvm::Value *addr) {
@@ -2863,10 +3207,10 @@ void CodeGenFunction::EmitCallArgs(
     size_t CallArgsStart = Args.size();
     for (int I = ArgTypes.size() - 1; I >= 0; --I) {
       CallExpr::const_arg_iterator Arg = ArgRange.begin() + I;
+      MaybeEmitImplicitObjectSize(I, *Arg);
       EmitCallArg(Args, *Arg, ArgTypes[I]);
       EmitNonNullArgCheck(Args.back().RV, ArgTypes[I], (*Arg)->getExprLoc(),
                           CalleeDecl, ParamsToSkip + I);
-      MaybeEmitImplicitObjectSize(I, *Arg);
     }
 
     // Un-reverse the arguments we just evaluated so they match up with the LLVM
@@ -3046,24 +3390,13 @@ CodeGenFunction::EmitRuntimeCall(llvm::Value *callee,
   return EmitRuntimeCall(callee, None, name);
 }
 
-/// Emits a simple call (never an invoke) to the given runtime
-/// function.
-llvm::CallInst *
-CodeGenFunction::EmitRuntimeCall(llvm::Value *callee,
-                                 ArrayRef<llvm::Value*> args,
-                                 const llvm::Twine &name) {
-  llvm::CallInst *call = Builder.CreateCall(callee, args, name);
-  call->setCallingConv(getRuntimeCC());
-  return call;
-}
-
 // Calls which may throw must have operand bundles indicating which funclet
 // they are nested within.
 static void
-getBundlesForFunclet(llvm::Value *Callee,
-                     llvm::Instruction *CurrentFuncletPad,
+getBundlesForFunclet(llvm::Value *Callee, llvm::Instruction *CurrentFuncletPad,
                      SmallVectorImpl<llvm::OperandBundleDef> &BundleList) {
-  // There is no need for a funclet operand bundle if we aren't inside a funclet.
+  // There is no need for a funclet operand bundle if we aren't inside a
+  // funclet.
   if (!CurrentFuncletPad)
     return;
 
@@ -3075,6 +3408,19 @@ getBundlesForFunclet(llvm::Value *Callee,
   BundleList.emplace_back("funclet", CurrentFuncletPad);
 }
 
+/// Emits a simple call (never an invoke) to the given runtime function.
+llvm::CallInst *
+CodeGenFunction::EmitRuntimeCall(llvm::Value *callee,
+                                 ArrayRef<llvm::Value*> args,
+                                 const llvm::Twine &name) {
+  SmallVector<llvm::OperandBundleDef, 1> BundleList;
+  getBundlesForFunclet(callee, CurrentFuncletPad, BundleList);
+
+  llvm::CallInst *call = Builder.CreateCall(callee, args, BundleList, name);
+  call->setCallingConv(getRuntimeCC());
+  return call;
+}
+
 /// Emits a call or invoke to the given noreturn runtime function.
 void CodeGenFunction::EmitNoreturnRuntimeCallOrInvoke(llvm::Value *callee,
                                                ArrayRef<llvm::Value*> args) {
@@ -3098,8 +3444,7 @@ void CodeGenFunction::EmitNoreturnRuntimeCallOrInvoke(llvm::Value *callee,
   }
 }
 
-/// Emits a call or invoke instruction to the given nullary runtime
-/// function.
+/// Emits a call or invoke instruction to the given nullary runtime function.
 llvm::CallSite
 CodeGenFunction::EmitRuntimeCallOrInvoke(llvm::Value *callee,
                                          const Twine &name) {
@@ -3123,13 +3468,16 @@ CodeGenFunction::EmitCallOrInvoke(llvm::Value *Callee,
                                   ArrayRef<llvm::Value *> Args,
                                   const Twine &Name) {
   llvm::BasicBlock *InvokeDest = getInvokeDest();
+  SmallVector<llvm::OperandBundleDef, 1> BundleList;
+  getBundlesForFunclet(Callee, CurrentFuncletPad, BundleList);
 
   llvm::Instruction *Inst;
   if (!InvokeDest)
-    Inst = Builder.CreateCall(Callee, Args, Name);
+    Inst = Builder.CreateCall(Callee, Args, BundleList, Name);
   else {
     llvm::BasicBlock *ContBB = createBasicBlock("invoke.cont");
-    Inst = Builder.CreateInvoke(Callee, ContBB, InvokeDest, Args, Name);
+    Inst = Builder.CreateInvoke(Callee, ContBB, InvokeDest, Args, BundleList,
+                                Name);
     EmitBlock(ContBB);
   }
 
@@ -3208,7 +3556,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   // alloca to hold the result, unless one is given to us.
   Address SRetPtr = Address::invalid();
   size_t UnusedReturnSize = 0;
-  if (RetAI.isIndirect() || RetAI.isInAlloca()) {
+  if (RetAI.isIndirect() || RetAI.isInAlloca() || RetAI.isCoerceAndExpand()) {
     if (!ReturnValue.isNull()) {
       SRetPtr = ReturnValue.getValue();
     } else {
@@ -3222,12 +3570,15 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
     }
     if (IRFunctionArgs.hasSRetArg()) {
       IRCallArgs[IRFunctionArgs.getSRetArgNo()] = SRetPtr.getPointer();
-    } else {
+    } else if (RetAI.isInAlloca()) {
       Address Addr = createInAllocaStructGEP(RetAI.getInAllocaFieldIndex());
       Builder.CreateStore(SRetPtr.getPointer(), Addr);
     }
   }
 
+  Address swiftErrorTemp = Address::invalid();
+  Address swiftErrorArg = Address::invalid();
+
   assert(CallInfo.arg_size() == CallArgs.size() &&
          "Mismatch between function signature & arguments.");
   unsigned ArgNo = 0;
@@ -3334,6 +3685,25 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
         else
           V = Builder.CreateLoad(RV.getAggregateAddress());
 
+        // Implement swifterror by copying into a new swifterror argument.
+        // We'll write back in the normal path out of the call.
+        if (CallInfo.getExtParameterInfo(ArgNo).getABI()
+              == ParameterABI::SwiftErrorResult) {
+          assert(!swiftErrorTemp.isValid() && "multiple swifterror args");
+
+          QualType pointeeTy = I->Ty->getPointeeType();
+          swiftErrorArg =
+            Address(V, getContext().getTypeAlignInChars(pointeeTy));
+
+          swiftErrorTemp =
+            CreateMemTemp(pointeeTy, getPointerAlign(), "swifterror.temp");
+          V = swiftErrorTemp.getPointer();
+          cast<llvm::AllocaInst>(V)->setSwiftError(true);
+
+          llvm::Value *errorValue = Builder.CreateLoad(swiftErrorArg);
+          Builder.CreateStore(errorValue, swiftErrorTemp);
+        }
+
         // We might have to widen integers, but we should never truncate.
         if (ArgInfo.getCoerceToType() != V->getType() &&
             V->getType()->isIntegerTy())
@@ -3344,6 +3714,7 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
         if (FirstIRArg < IRFuncTy->getNumParams() &&
             V->getType() != IRFuncTy->getParamType(FirstIRArg))
           V = Builder.CreateBitCast(V, IRFuncTy->getParamType(FirstIRArg));
+
         IRCallArgs[FirstIRArg] = V;
         break;
       }
@@ -3402,6 +3773,51 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
       break;
     }
 
+    case ABIArgInfo::CoerceAndExpand: {
+      auto coercionType = ArgInfo.getCoerceAndExpandType();
+      auto layout = CGM.getDataLayout().getStructLayout(coercionType);
+
+      llvm::Value *tempSize = nullptr;
+      Address addr = Address::invalid();
+      if (RV.isAggregate()) {
+        addr = RV.getAggregateAddress();
+      } else {
+        assert(RV.isScalar()); // complex should always just be direct
+
+        llvm::Type *scalarType = RV.getScalarVal()->getType();
+        auto scalarSize = CGM.getDataLayout().getTypeAllocSize(scalarType);
+        auto scalarAlign = CGM.getDataLayout().getPrefTypeAlignment(scalarType);
+
+        tempSize = llvm::ConstantInt::get(CGM.Int64Ty, scalarSize);
+
+        // Materialize to a temporary.
+        addr = CreateTempAlloca(RV.getScalarVal()->getType(),
+                 CharUnits::fromQuantity(std::max(layout->getAlignment(),
+                                                  scalarAlign)));
+        EmitLifetimeStart(scalarSize, addr.getPointer());
+
+        Builder.CreateStore(RV.getScalarVal(), addr);
+      }
+
+      addr = Builder.CreateElementBitCast(addr, coercionType);
+
+      unsigned IRArgPos = FirstIRArg;
+      for (unsigned i = 0, e = coercionType->getNumElements(); i != e; ++i) {
+        llvm::Type *eltType = coercionType->getElementType(i);
+        if (ABIArgInfo::isPaddingForCoerceAndExpand(eltType)) continue;
+        Address eltAddr = Builder.CreateStructGEP(addr, i, layout);
+        llvm::Value *elt = Builder.CreateLoad(eltAddr);
+        IRCallArgs[IRArgPos++] = elt;
+      }
+      assert(IRArgPos == FirstIRArg + NumIRArgs);
+
+      if (tempSize) {
+        EmitLifetimeEnd(tempSize, addr.getPointer());
+      }
+
+      break;
+    }
+
     case ABIArgInfo::Expand:
       unsigned IRArgPos = FirstIRArg;
       ExpandTypeToArgs(I->Ty, RV, IRFuncTy, IRCallArgs, IRArgPos);
@@ -3541,6 +3957,13 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   CS.setAttributes(Attrs);
   CS.setCallingConv(static_cast<llvm::CallingConv::ID>(CallingConv));
 
+  // Insert instrumentation or attach profile metadata at indirect call sites.
+  // For more details, see the comment before the definition of
+  // IPVK_IndirectCallTarget in InstrProfData.inc.
+  if (!CS.getCalledFunction())
+    PGO.valueProfile(Builder, llvm::IPVK_IndirectCallTarget,
+                     CS.getInstruction(), Callee);
+
   // In ObjC ARC mode with no ObjC ARC exception safety, tell the ARC
   // optimizer it can aggressively ignore unwind edges.
   if (CGM.getLangOpts().ObjCAutoRefCount)
@@ -3567,9 +3990,15 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
   }
 
   llvm::Instruction *CI = CS.getInstruction();
-  if (Builder.isNamePreserving() && !CI->getType()->isVoidTy())
+  if (!CI->getType()->isVoidTy())
     CI->setName("call");
 
+  // Perform the swifterror writeback.
+  if (swiftErrorTemp.isValid()) {
+    llvm::Value *errorResult = Builder.CreateLoad(swiftErrorTemp);
+    Builder.CreateStore(errorResult, swiftErrorArg);
+  }
+
   // Emit any writebacks immediately.  Arguably this should happen
   // after any return-value munging.
   if (CallArgs.hasWritebacks())
@@ -3587,6 +4016,31 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
 
   RValue Ret = [&] {
     switch (RetAI.getKind()) {
+    case ABIArgInfo::CoerceAndExpand: {
+      auto coercionType = RetAI.getCoerceAndExpandType();
+      auto layout = CGM.getDataLayout().getStructLayout(coercionType);
+
+      Address addr = SRetPtr;
+      addr = Builder.CreateElementBitCast(addr, coercionType);
+
+      assert(CI->getType() == RetAI.getUnpaddedCoerceAndExpandType());
+      bool requiresExtract = isa<llvm::StructType>(CI->getType());
+
+      unsigned unpaddedIndex = 0;
+      for (unsigned i = 0, e = coercionType->getNumElements(); i != e; ++i) {
+        llvm::Type *eltType = coercionType->getElementType(i);
+        if (ABIArgInfo::isPaddingForCoerceAndExpand(eltType)) continue;
+        Address eltAddr = Builder.CreateStructGEP(addr, i, layout);
+        llvm::Value *elt = CI;
+        if (requiresExtract)
+          elt = Builder.CreateExtractValue(elt, unpaddedIndex++);
+        else
+          assert(unpaddedIndex == 0);
+        Builder.CreateStore(elt, eltAddr);
+      }
+      // FALLTHROUGH
+    }
+
     case ABIArgInfo::InAlloca:
     case ABIArgInfo::Indirect: {
       RValue ret = convertTempToRValue(SRetPtr, RetTy, SourceLocation());
diff --git a/lib/CodeGen/CGClass.cpp b/lib/CodeGen/CGClass.cpp
index 2e566de6d8ac..7ed891f426aa 100644
--- a/lib/CodeGen/CGClass.cpp
+++ b/lib/CodeGen/CGClass.cpp
@@ -26,6 +26,7 @@
 #include "clang/Frontend/CodeGenOptions.h"
 #include "llvm/IR/Intrinsics.h"
 #include "llvm/IR/Metadata.h"
+#include "llvm/Transforms/Utils/SanitizerStats.h"
 
 using namespace clang;
 using namespace CodeGen;
@@ -94,7 +95,7 @@ CodeGenModule::getDynamicOffsetAlignment(CharUnits actualBaseAlign,
   // unless we someday add some sort of attribute to change the
   // assumed alignment of 'this'.  So our goal here is pretty much
   // just to allow the user to explicitly say that a pointer is
-  // under-aligned and then safely access its fields and v-tables.
+  // under-aligned and then safely access its fields and vtables.
   if (actualBaseAlign >= expectedBaseAlign) {
     return expectedTargetAlign;
   }
@@ -745,7 +746,7 @@ static void EmitMemberInitializer(CodeGenFunction &CGF,
 
   ArrayRef<VarDecl *> ArrayIndexes;
   if (MemberInit->getNumArrayIndices())
-    ArrayIndexes = MemberInit->getArrayIndexes();
+    ArrayIndexes = MemberInit->getArrayIndices();
   CGF.EmitInitializerForField(Field, LHS, MemberInit->getInit(), ArrayIndexes);
 }
 
@@ -986,7 +987,7 @@ namespace {
     CodeGenFunction &CGF;
     SanitizerSet OldSanOpts;
   };
-}
+} // end anonymous namespace
  
 namespace {
   class FieldMemcpyizer {
@@ -1071,7 +1072,6 @@ namespace {
     const CXXRecordDecl *ClassDecl;
 
   private:
-
     void emitMemcpyIR(Address DestPtr, Address SrcPtr, CharUnits Size) {
       llvm::PointerType *DPT = DestPtr.getType();
       llvm::Type *DBP =
@@ -1087,13 +1087,12 @@ namespace {
     }
 
     void addInitialField(FieldDecl *F) {
-        FirstField = F;
-        LastField = F;
-        FirstFieldOffset = RecLayout.getFieldOffset(F->getFieldIndex());
-        LastFieldOffset = FirstFieldOffset;
-        LastAddedFieldIndex = F->getFieldIndex();
-        return;
-      }
+      FirstField = F;
+      LastField = F;
+      FirstFieldOffset = RecLayout.getFieldOffset(F->getFieldIndex());
+      LastFieldOffset = FirstFieldOffset;
+      LastAddedFieldIndex = F->getFieldIndex();
+    }
 
     void addNextField(FieldDecl *F) {
       // For the most part, the following invariant will hold:
@@ -1127,7 +1126,6 @@ namespace {
 
   class ConstructorMemcpyizer : public FieldMemcpyizer {
   private:
-
     /// Get source argument for copy constructor. Returns null if not a copy
     /// constructor.
     static const VarDecl *getTrivialCopySource(CodeGenFunction &CGF,
@@ -1232,7 +1230,6 @@ namespace {
 
   class AssignmentMemcpyizer : public FieldMemcpyizer {
   private:
-
     // Returns the memcpyable field copied by the given statement, if one
     // exists. Otherwise returns null.
     FieldDecl *getMemcpyableField(Stmt *S) {
@@ -1306,7 +1303,6 @@ namespace {
     SmallVector<Stmt*, 16> AggregatedStmts;
 
   public:
-
     AssignmentMemcpyizer(CodeGenFunction &CGF, const CXXMethodDecl *AD,
                          FunctionArgList &Args)
       : FieldMemcpyizer(CGF, AD->getParent(), Args[Args.size() - 1]),
@@ -1607,6 +1603,7 @@ void CodeGenFunction::emitImplicitAssignmentOperatorBody(FunctionArgList &Args)
 
   LexicalScope Scope(*this, RootCS->getSourceRange());
 
+  incrementProfileCounter(RootCS);
   AssignmentMemcpyizer AM(*this, AssignOp, Args);
   for (auto *I : RootCS->body())
     AM.emitAssignment(I);
@@ -1628,6 +1625,7 @@ namespace {
 
   struct CallDtorDeleteConditional final : EHScopeStack::Cleanup {
     llvm::Value *ShouldDeleteCondition;
+
   public:
     CallDtorDeleteConditional(llvm::Value *ShouldDeleteCondition)
         : ShouldDeleteCondition(ShouldDeleteCondition) {
@@ -1917,7 +1915,7 @@ void CodeGenFunction::EnterDtorCleanups(const CXXDestructorDecl *DD,
 /// \param zeroInitialize true if each element should be
 ///   zero-initialized before it is constructed
 void CodeGenFunction::EmitCXXAggrConstructorCall(
-    const CXXConstructorDecl *ctor, const ConstantArrayType *arrayType,
+    const CXXConstructorDecl *ctor, const ArrayType *arrayType,
     Address arrayBegin, const CXXConstructExpr *E, bool zeroInitialize) {
   QualType elementType;
   llvm::Value *numElements =
@@ -2050,6 +2048,62 @@ void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
                                              bool ForVirtualBase,
                                              bool Delegating, Address This,
                                              const CXXConstructExpr *E) {
+  CallArgList Args;
+
+  // Push the this ptr.
+  Args.add(RValue::get(This.getPointer()), D->getThisType(getContext()));
+
+  // If this is a trivial constructor, emit a memcpy now before we lose
+  // the alignment information on the argument.
+  // FIXME: It would be better to preserve alignment information into CallArg.
+  if (isMemcpyEquivalentSpecialMember(D)) {
+    assert(E->getNumArgs() == 1 && "unexpected argcount for trivial ctor");
+
+    const Expr *Arg = E->getArg(0);
+    QualType SrcTy = Arg->getType();
+    Address Src = EmitLValue(Arg).getAddress();
+    QualType DestTy = getContext().getTypeDeclType(D->getParent());
+    EmitAggregateCopyCtor(This, Src, DestTy, SrcTy);
+    return;
+  }
+
+  // Add the rest of the user-supplied arguments.
+  const FunctionProtoType *FPT = D->getType()->castAs<FunctionProtoType>();
+  EmitCallArgs(Args, FPT, E->arguments(), E->getConstructor());
+
+  EmitCXXConstructorCall(D, Type, ForVirtualBase, Delegating, This, Args);
+}
+
+static bool canEmitDelegateCallArgs(CodeGenFunction &CGF,
+                                    const CXXConstructorDecl *Ctor,
+                                    CXXCtorType Type, CallArgList &Args) {
+  // We can't forward a variadic call.
+  if (Ctor->isVariadic())
+    return false;
+
+  if (CGF.getTarget().getCXXABI().areArgsDestroyedLeftToRightInCallee()) {
+    // If the parameters are callee-cleanup, it's not safe to forward.
+    for (auto *P : Ctor->parameters())
+      if (P->getType().isDestructedType())
+        return false;
+
+    // Likewise if they're inalloca.
+    const CGFunctionInfo &Info =
+        CGF.CGM.getTypes().arrangeCXXConstructorCall(Args, Ctor, Type, 0);
+    if (Info.usesInAlloca())
+      return false;
+  }
+
+  // Anything else should be OK.
+  return true;
+}
+
+void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
+                                             CXXCtorType Type,
+                                             bool ForVirtualBase,
+                                             bool Delegating,
+                                             Address This,
+                                             CallArgList &Args) {
   const CXXRecordDecl *ClassDecl = D->getParent();
 
   // C++11 [class.mfct.non-static]p2:
@@ -2060,7 +2114,7 @@ void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
                 This.getPointer(), getContext().getRecordType(ClassDecl));
 
   if (D->isTrivial() && D->isDefaultConstructor()) {
-    assert(E->getNumArgs() == 0 && "trivial default ctor with args");
+    assert(Args.size() == 1 && "trivial default ctor with args");
     return;
   }
 
@@ -2068,24 +2122,24 @@ void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
   // union copy constructor, we must emit a memcpy, because the AST does not
   // model that copy.
   if (isMemcpyEquivalentSpecialMember(D)) {
-    assert(E->getNumArgs() == 1 && "unexpected argcount for trivial ctor");
+    assert(Args.size() == 2 && "unexpected argcount for trivial ctor");
 
-    const Expr *Arg = E->getArg(0);
-    QualType SrcTy = Arg->getType();
-    Address Src = EmitLValue(Arg).getAddress();
+    QualType SrcTy = D->getParamDecl(0)->getType().getNonReferenceType();
+    Address Src(Args[1].RV.getScalarVal(), getNaturalTypeAlignment(SrcTy));
     QualType DestTy = getContext().getTypeDeclType(ClassDecl);
     EmitAggregateCopyCtor(This, Src, DestTy, SrcTy);
     return;
   }
 
-  CallArgList Args;
-
-  // Push the this ptr.
-  Args.add(RValue::get(This.getPointer()), D->getThisType(getContext()));
-
-  // Add the rest of the user-supplied arguments.
-  const FunctionProtoType *FPT = D->getType()->castAs<FunctionProtoType>();
-  EmitCallArgs(Args, FPT, E->arguments(), E->getConstructor());
+  // Check whether we can actually emit the constructor before trying to do so.
+  if (auto Inherited = D->getInheritedConstructor()) {
+    if (getTypes().inheritingCtorHasParams(Inherited, Type) &&
+        !canEmitDelegateCallArgs(*this, D, Type, Args)) {
+      EmitInlinedInheritingCXXConstructorCall(D, Type, ForVirtualBase,
+                                              Delegating, Args);
+      return;
+    }
+  }
 
   // Insert any ABI-specific implicit constructor arguments.
   unsigned ExtraArgs = CGM.getCXXABI().addImplicitConstructorArgs(
@@ -2115,6 +2169,95 @@ void CodeGenFunction::EmitCXXConstructorCall(const CXXConstructorDecl *D,
     EmitVTableAssumptionLoads(ClassDecl, This);
 }
 
+void CodeGenFunction::EmitInheritedCXXConstructorCall(
+    const CXXConstructorDecl *D, bool ForVirtualBase, Address This,
+    bool InheritedFromVBase, const CXXInheritedCtorInitExpr *E) {
+  CallArgList Args;
+  CallArg ThisArg(RValue::get(This.getPointer()), D->getThisType(getContext()),
+                  /*NeedsCopy=*/false);
+
+  // Forward the parameters.
+  if (InheritedFromVBase &&
+      CGM.getTarget().getCXXABI().hasConstructorVariants()) {
+    // Nothing to do; this construction is not responsible for constructing
+    // the base class containing the inherited constructor.
+    // FIXME: Can we just pass undef's for the remaining arguments if we don't
+    // have constructor variants?
+    Args.push_back(ThisArg);
+  } else if (!CXXInheritedCtorInitExprArgs.empty()) {
+    // The inheriting constructor was inlined; just inject its arguments.
+    assert(CXXInheritedCtorInitExprArgs.size() >= D->getNumParams() &&
+           "wrong number of parameters for inherited constructor call");
+    Args = CXXInheritedCtorInitExprArgs;
+    Args[0] = ThisArg;
+  } else {
+    // The inheriting constructor was not inlined. Emit delegating arguments.
+    Args.push_back(ThisArg);
+    const auto *OuterCtor = cast<CXXConstructorDecl>(CurCodeDecl);
+    assert(OuterCtor->getNumParams() == D->getNumParams());
+    assert(!OuterCtor->isVariadic() && "should have been inlined");
+
+    for (const auto *Param : OuterCtor->parameters()) {
+      assert(getContext().hasSameUnqualifiedType(
+          OuterCtor->getParamDecl(Param->getFunctionScopeIndex())->getType(),
+          Param->getType()));
+      EmitDelegateCallArg(Args, Param, E->getLocation());
+
+      // Forward __attribute__(pass_object_size).
+      if (Param->hasAttr<PassObjectSizeAttr>()) {
+        auto *POSParam = SizeArguments[Param];
+        assert(POSParam && "missing pass_object_size value for forwarding");
+        EmitDelegateCallArg(Args, POSParam, E->getLocation());
+      }
+    }
+  }
+
+  EmitCXXConstructorCall(D, Ctor_Base, ForVirtualBase, /*Delegating*/false,
+                         This, Args);
+}
+
+void CodeGenFunction::EmitInlinedInheritingCXXConstructorCall(
+    const CXXConstructorDecl *Ctor, CXXCtorType CtorType, bool ForVirtualBase,
+    bool Delegating, CallArgList &Args) {
+  InlinedInheritingConstructorScope Scope(*this, GlobalDecl(Ctor, CtorType));
+
+  // Save the arguments to be passed to the inherited constructor.
+  CXXInheritedCtorInitExprArgs = Args;
+
+  FunctionArgList Params;
+  QualType RetType = BuildFunctionArgList(CurGD, Params);
+  FnRetTy = RetType;
+
+  // Insert any ABI-specific implicit constructor arguments.
+  CGM.getCXXABI().addImplicitConstructorArgs(*this, Ctor, CtorType,
+                                             ForVirtualBase, Delegating, Args);
+
+  // Emit a simplified prolog. We only need to emit the implicit params.
+  assert(Args.size() >= Params.size() && "too few arguments for call");
+  for (unsigned I = 0, N = Args.size(); I != N; ++I) {
+    if (I < Params.size() && isa<ImplicitParamDecl>(Params[I])) {
+      const RValue &RV = Args[I].RV;
+      assert(!RV.isComplex() && "complex indirect params not supported");
+      ParamValue Val = RV.isScalar()
+                           ? ParamValue::forDirect(RV.getScalarVal())
+                           : ParamValue::forIndirect(RV.getAggregateAddress());
+      EmitParmDecl(*Params[I], Val, I + 1);
+    }
+  }
+
+  // Create a return value slot if the ABI implementation wants one.
+  // FIXME: This is dumb, we should ask the ABI not to try to set the return
+  // value instead.
+  if (!RetType->isVoidType())
+    ReturnValue = CreateIRTemp(RetType, "retval.inhctor");
+
+  CGM.getCXXABI().EmitInstanceFunctionProlog(*this);
+  CXXThisValue = CXXABIThisValue;
+
+  // Directly emit the constructor initializers.
+  EmitCtorPrologue(Ctor, CtorType, Params);
+}
+
 void CodeGenFunction::EmitVTableAssumptionLoad(const VPtr &Vptr, Address This) {
   llvm::Value *VTableGlobal =
       CGM.getCXXABI().getVTableAddressPoint(Vptr.Base, Vptr.VTableClass);
@@ -2147,19 +2290,6 @@ void
 CodeGenFunction::EmitSynthesizedCXXCopyCtorCall(const CXXConstructorDecl *D,
                                                 Address This, Address Src,
                                                 const CXXConstructExpr *E) {
-  if (isMemcpyEquivalentSpecialMember(D)) {
-    assert(E->getNumArgs() == 1 && "unexpected argcount for trivial ctor");
-    assert(D->isCopyOrMoveConstructor() &&
-           "trivial 1-arg ctor not a copy/move ctor");
-    EmitAggregateCopyCtor(This, Src,
-                          getContext().getTypeDeclType(D->getParent()),
-                          (*E->arg_begin())->getType());
-    return;
-  }
-  llvm::Value *Callee = CGM.getAddrOfCXXStructor(D, StructorType::Complete);
-  assert(D->isInstance() &&
-         "Trying to emit a member call expr on a static method!");
-
   const FunctionProtoType *FPT = D->getType()->castAs<FunctionProtoType>();
 
   CallArgList Args;
@@ -2177,8 +2307,7 @@ CodeGenFunction::EmitSynthesizedCXXCopyCtorCall(const CXXConstructorDecl *D,
   EmitCallArgs(Args, FPT, drop_begin(E->arguments(), 1), E->getConstructor(),
                /*ParamsToSkip*/ 1);
 
-  EmitCall(CGM.getTypes().arrangeCXXMethodCall(Args, FPT, RequiredArgs::All),
-           Callee, ReturnValueSlot(), Args, D);
+  EmitCXXConstructorCall(D, Ctor_Complete, false, false, This, Args);
 }
 
 void
@@ -2192,21 +2321,17 @@ CodeGenFunction::EmitDelegateCXXConstructorCall(const CXXConstructorDecl *Ctor,
   assert(I != E && "no parameters to constructor");
 
   // this
-  DelegateArgs.add(RValue::get(LoadCXXThis()), (*I)->getType());
+  Address This = LoadCXXThisAddress();
+  DelegateArgs.add(RValue::get(This.getPointer()), (*I)->getType());
   ++I;
 
-  // vtt
-  if (llvm::Value *VTT = GetVTTParameter(GlobalDecl(Ctor, CtorType),
-                                         /*ForVirtualBase=*/false,
-                                         /*Delegating=*/true)) {
-    QualType VoidPP = getContext().getPointerType(getContext().VoidPtrTy);
-    DelegateArgs.add(RValue::get(VTT), VoidPP);
-
-    if (CGM.getCXXABI().NeedsVTTParameter(CurGD)) {
-      assert(I != E && "cannot skip vtt parameter, already done with args");
-      assert((*I)->getType() == VoidPP && "skipping parameter not of vtt type");
-      ++I;
-    }
+  // FIXME: The location of the VTT parameter in the parameter list is
+  // specific to the Itanium ABI and shouldn't be hardcoded here.
+  if (CGM.getCXXABI().NeedsVTTParameter(CurGD)) {
+    assert(I != E && "cannot skip vtt parameter, already done with args");
+    assert((*I)->getType()->isPointerType() &&
+           "skipping parameter not of vtt type");
+    ++I;
   }
 
   // Explicit arguments.
@@ -2216,11 +2341,8 @@ CodeGenFunction::EmitDelegateCXXConstructorCall(const CXXConstructorDecl *Ctor,
     EmitDelegateCallArg(DelegateArgs, param, Loc);
   }
 
-  llvm::Value *Callee =
-      CGM.getAddrOfCXXStructor(Ctor, getFromCtorType(CtorType));
-  EmitCall(CGM.getTypes()
-               .arrangeCXXStructorDeclaration(Ctor, getFromCtorType(CtorType)),
-           Callee, ReturnValueSlot(), DelegateArgs, Ctor);
+  EmitCXXConstructorCall(Ctor, CtorType, /*ForVirtualBase=*/false,
+                         /*Delegating=*/true, This, DelegateArgs);
 }
 
 namespace {
@@ -2289,7 +2411,7 @@ namespace {
                                 /*Delegating=*/false, Addr);
     }
   };
-}
+} // end anonymous namespace
 
 void CodeGenFunction::PushDestructorCleanup(const CXXDestructorDecl *D,
                                             Address Addr) {
@@ -2487,15 +2609,35 @@ LeastDerivedClassWithSameLayout(const CXXRecordDecl *RD) {
       RD->bases_begin()->getType()->getAsCXXRecordDecl());
 }
 
-void CodeGenFunction::EmitVTablePtrCheckForCall(const CXXMethodDecl *MD,
+void CodeGenFunction::EmitTypeMetadataCodeForVCall(const CXXRecordDecl *RD,
+                                                   llvm::Value *VTable,
+                                                   SourceLocation Loc) {
+  if (CGM.getCodeGenOpts().WholeProgramVTables &&
+      CGM.HasHiddenLTOVisibility(RD)) {
+    llvm::Metadata *MD =
+        CGM.CreateMetadataIdentifierForType(QualType(RD->getTypeForDecl(), 0));
+    llvm::Value *TypeId =
+        llvm::MetadataAsValue::get(CGM.getLLVMContext(), MD);
+
+    llvm::Value *CastedVTable = Builder.CreateBitCast(VTable, Int8PtrTy);
+    llvm::Value *TypeTest =
+        Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::type_test),
+                           {CastedVTable, TypeId});
+    Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::assume), TypeTest);
+  }
+
+  if (SanOpts.has(SanitizerKind::CFIVCall))
+    EmitVTablePtrCheckForCall(RD, VTable, CodeGenFunction::CFITCK_VCall, Loc);
+}
+
+void CodeGenFunction::EmitVTablePtrCheckForCall(const CXXRecordDecl *RD,
                                                 llvm::Value *VTable,
                                                 CFITypeCheckKind TCK,
                                                 SourceLocation Loc) {
-  const CXXRecordDecl *ClassDecl = MD->getParent();
   if (!SanOpts.has(SanitizerKind::CFICastStrict))
-    ClassDecl = LeastDerivedClassWithSameLayout(ClassDecl);
+    RD = LeastDerivedClassWithSameLayout(RD);
 
-  EmitVTablePtrCheck(ClassDecl, VTable, TCK, Loc);
+  EmitVTablePtrCheck(RD, VTable, TCK, Loc);
 }
 
 void CodeGenFunction::EmitVTablePtrCheckForCast(QualType T,
@@ -2547,26 +2689,41 @@ void CodeGenFunction::EmitVTablePtrCheck(const CXXRecordDecl *RD,
                                          llvm::Value *VTable,
                                          CFITypeCheckKind TCK,
                                          SourceLocation Loc) {
-  if (CGM.IsCFIBlacklistedRecord(RD))
+  if (!CGM.getCodeGenOpts().SanitizeCfiCrossDso &&
+      !CGM.HasHiddenLTOVisibility(RD))
+    return;
+
+  std::string TypeName = RD->getQualifiedNameAsString();
+  if (getContext().getSanitizerBlacklist().isBlacklistedType(TypeName))
     return;
 
   SanitizerScope SanScope(this);
+  llvm::SanitizerStatKind SSK;
+  switch (TCK) {
+  case CFITCK_VCall:
+    SSK = llvm::SanStat_CFI_VCall;
+    break;
+  case CFITCK_NVCall:
+    SSK = llvm::SanStat_CFI_NVCall;
+    break;
+  case CFITCK_DerivedCast:
+    SSK = llvm::SanStat_CFI_DerivedCast;
+    break;
+  case CFITCK_UnrelatedCast:
+    SSK = llvm::SanStat_CFI_UnrelatedCast;
+    break;
+  case CFITCK_ICall:
+    llvm_unreachable("not expecting CFITCK_ICall");
+  }
+  EmitSanitizerStatReport(SSK);
 
   llvm::Metadata *MD =
       CGM.CreateMetadataIdentifierForType(QualType(RD->getTypeForDecl(), 0));
-  llvm::Value *BitSetName = llvm::MetadataAsValue::get(getLLVMContext(), MD);
+  llvm::Value *TypeId = llvm::MetadataAsValue::get(getLLVMContext(), MD);
 
   llvm::Value *CastedVTable = Builder.CreateBitCast(VTable, Int8PtrTy);
-  llvm::Value *BitSetTest =
-      Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::bitset_test),
-                         {CastedVTable, BitSetName});
-
-  if (CGM.getCodeGenOpts().SanitizeCfiCrossDso) {
-    if (auto TypeId = CGM.CreateCfiIdForTypeMetadata(MD)) {
-      EmitCfiSlowPathCheck(BitSetTest, TypeId, CastedVTable);
-      return;
-    }
-  }
+  llvm::Value *TypeTest = Builder.CreateCall(
+      CGM.getIntrinsic(llvm::Intrinsic::type_test), {CastedVTable, TypeId});
 
   SanitizerMask M;
   switch (TCK) {
@@ -2582,15 +2739,70 @@ void CodeGenFunction::EmitVTablePtrCheck(const CXXRecordDecl *RD,
   case CFITCK_UnrelatedCast:
     M = SanitizerKind::CFIUnrelatedCast;
     break;
+  case CFITCK_ICall:
+    llvm_unreachable("not expecting CFITCK_ICall");
   }
 
   llvm::Constant *StaticData[] = {
+      llvm::ConstantInt::get(Int8Ty, TCK),
       EmitCheckSourceLocation(Loc),
       EmitCheckTypeDescriptor(QualType(RD->getTypeForDecl(), 0)),
-      llvm::ConstantInt::get(Int8Ty, TCK),
   };
-  EmitCheck(std::make_pair(BitSetTest, M), "cfi_bad_type", StaticData,
-            CastedVTable);
+
+  auto CrossDsoTypeId = CGM.CreateCrossDsoCfiTypeId(MD);
+  if (CGM.getCodeGenOpts().SanitizeCfiCrossDso && CrossDsoTypeId) {
+    EmitCfiSlowPathCheck(M, TypeTest, CrossDsoTypeId, CastedVTable, StaticData);
+    return;
+  }
+
+  if (CGM.getCodeGenOpts().SanitizeTrap.has(M)) {
+    EmitTrapCheck(TypeTest);
+    return;
+  }
+
+  llvm::Value *AllVtables = llvm::MetadataAsValue::get(
+      CGM.getLLVMContext(),
+      llvm::MDString::get(CGM.getLLVMContext(), "all-vtables"));
+  llvm::Value *ValidVtable = Builder.CreateCall(
+      CGM.getIntrinsic(llvm::Intrinsic::type_test), {CastedVTable, AllVtables});
+  EmitCheck(std::make_pair(TypeTest, M), "cfi_check_fail", StaticData,
+            {CastedVTable, ValidVtable});
+}
+
+bool CodeGenFunction::ShouldEmitVTableTypeCheckedLoad(const CXXRecordDecl *RD) {
+  if (!CGM.getCodeGenOpts().WholeProgramVTables ||
+      !SanOpts.has(SanitizerKind::CFIVCall) ||
+      !CGM.getCodeGenOpts().SanitizeTrap.has(SanitizerKind::CFIVCall) ||
+      !CGM.HasHiddenLTOVisibility(RD))
+    return false;
+
+  std::string TypeName = RD->getQualifiedNameAsString();
+  return !getContext().getSanitizerBlacklist().isBlacklistedType(TypeName);
+}
+
+llvm::Value *CodeGenFunction::EmitVTableTypeCheckedLoad(
+    const CXXRecordDecl *RD, llvm::Value *VTable, uint64_t VTableByteOffset) {
+  SanitizerScope SanScope(this);
+
+  EmitSanitizerStatReport(llvm::SanStat_CFI_VCall);
+
+  llvm::Metadata *MD =
+      CGM.CreateMetadataIdentifierForType(QualType(RD->getTypeForDecl(), 0));
+  llvm::Value *TypeId = llvm::MetadataAsValue::get(CGM.getLLVMContext(), MD);
+
+  llvm::Value *CastedVTable = Builder.CreateBitCast(VTable, Int8PtrTy);
+  llvm::Value *CheckedLoad = Builder.CreateCall(
+      CGM.getIntrinsic(llvm::Intrinsic::type_checked_load),
+      {CastedVTable, llvm::ConstantInt::get(Int32Ty, VTableByteOffset),
+       TypeId});
+  llvm::Value *CheckResult = Builder.CreateExtractValue(CheckedLoad, 1);
+
+  EmitCheck(std::make_pair(CheckResult, SanitizerKind::CFIVCall),
+            "cfi_check_fail", nullptr, nullptr);
+
+  return Builder.CreateBitCast(
+      Builder.CreateExtractValue(CheckedLoad, 0),
+      cast<llvm::PointerType>(VTable->getType())->getElementType());
 }
 
 // FIXME: Ideally Expr::IgnoreParenNoopCasts should do this, but it doesn't do
@@ -2731,7 +2943,7 @@ void CodeGenFunction::EmitLambdaBlockInvokeBody() {
   CallArgs.add(RValue::get(ThisPtr.getPointer()), ThisType);
 
   // Add the rest of the parameters.
-  for (auto param : BD->params())
+  for (auto param : BD->parameters())
     EmitDelegateCallArg(CallArgs, param, param->getLocStart());
 
   assert(!Lambda->isGenericLambda() &&
@@ -2761,7 +2973,7 @@ void CodeGenFunction::EmitLambdaDelegatingInvokeBody(const CXXMethodDecl *MD) {
   CallArgs.add(RValue::get(ThisPtr), ThisType);
 
   // Add the rest of the parameters.
-  for (auto Param : MD->params())
+  for (auto Param : MD->parameters())
     EmitDelegateCallArg(CallArgs, Param, Param->getLocStart());
 
   const CXXMethodDecl *CallOp = Lambda->getLambdaCallOperator();
diff --git a/lib/CodeGen/CGCleanup.cpp b/lib/CodeGen/CGCleanup.cpp
index ba7dcf7de6c7..b3278b3b4fef 100644
--- a/lib/CodeGen/CGCleanup.cpp
+++ b/lib/CodeGen/CGCleanup.cpp
@@ -112,7 +112,7 @@ RValue DominatingValue<RValue>::saved_type::restore(CodeGenFunction &CGF) {
 
 /// Push an entry of the given size onto this protected-scope stack.
 char *EHScopeStack::allocate(size_t Size) {
-  Size = llvm::RoundUpToAlignment(Size, ScopeStackAlignment);
+  Size = llvm::alignTo(Size, ScopeStackAlignment);
   if (!StartOfBuffer) {
     unsigned Capacity = 1024;
     while (Capacity < Size) Capacity *= 2;
@@ -143,7 +143,7 @@ char *EHScopeStack::allocate(size_t Size) {
 }
 
 void EHScopeStack::deallocate(size_t Size) {
-  StartOfData += llvm::RoundUpToAlignment(Size, ScopeStackAlignment);
+  StartOfData += llvm::alignTo(Size, ScopeStackAlignment);
 }
 
 bool EHScopeStack::containsOnlyLifetimeMarkers(
@@ -157,6 +157,20 @@ bool EHScopeStack::containsOnlyLifetimeMarkers(
   return true;
 }
 
+bool EHScopeStack::requiresLandingPad() const {
+  for (stable_iterator si = getInnermostEHScope(); si != stable_end(); ) {
+    // Skip lifetime markers.
+    if (auto *cleanup = dyn_cast<EHCleanupScope>(&*find(si)))
+      if (cleanup->isLifetimeMarker()) {
+        si = cleanup->getEnclosingEHScope();
+        continue;
+      }
+    return true;
+  }
+
+  return false;
+}
+
 EHScopeStack::stable_iterator
 EHScopeStack::getInnermostActiveNormalCleanup() const {
   for (stable_iterator si = getInnermostNormalCleanup(), se = stable_end();
@@ -174,6 +188,7 @@ void *EHScopeStack::pushCleanup(CleanupKind Kind, size_t Size) {
   bool IsNormalCleanup = Kind & NormalCleanup;
   bool IsEHCleanup = Kind & EHCleanup;
   bool IsActive = !(Kind & InactiveCleanup);
+  bool IsLifetimeMarker = Kind & LifetimeMarker;
   EHCleanupScope *Scope =
     new (Buffer) EHCleanupScope(IsNormalCleanup,
                                 IsEHCleanup,
@@ -186,6 +201,8 @@ void *EHScopeStack::pushCleanup(CleanupKind Kind, size_t Size) {
     InnermostNormalCleanup = stable_begin();
   if (IsEHCleanup)
     InnermostEHScope = stable_begin();
+  if (IsLifetimeMarker)
+    Scope->setLifetimeMarker();
 
   return Scope->getCleanupBuffer();
 }
diff --git a/lib/CodeGen/CGCleanup.h b/lib/CodeGen/CGCleanup.h
index 909f00b05925..98d01b1326c9 100644
--- a/lib/CodeGen/CGCleanup.h
+++ b/lib/CodeGen/CGCleanup.h
@@ -86,11 +86,6 @@ protected:
     /// The amount of extra storage needed by the Cleanup.
     /// Always a multiple of the scope-stack alignment.
     unsigned CleanupSize : 12;
-
-    /// The number of fixups required by enclosing scopes (not including
-    /// this one).  If this is the top cleanup scope, all the fixups
-    /// from this index onwards belong to this scope.
-    unsigned FixupDepth : 32 - 18 - NumCommonBits; // currently 12
   };
 
   class FilterBitFields {
@@ -188,6 +183,7 @@ public:
                EHScopeStack::stable_iterator enclosingEHScope)
     : EHScope(Catch, enclosingEHScope) {
     CatchBits.NumHandlers = numHandlers;
+    assert(CatchBits.NumHandlers == numHandlers && "NumHandlers overflow?");
   }
 
   unsigned getNumHandlers() const {
@@ -263,6 +259,11 @@ class LLVM_ALIGNAS(/*alignof(uint64_t)*/ 8) EHCleanupScope : public EHScope {
   };
   mutable struct ExtInfo *ExtInfo;
 
+  /// The number of fixups required by enclosing scopes (not including
+  /// this one).  If this is the top cleanup scope, all the fixups
+  /// from this index onwards belong to this scope.
+  unsigned FixupDepth;
+
   struct ExtInfo &getExtInfo() {
     if (!ExtInfo) ExtInfo = new struct ExtInfo();
     return *ExtInfo;
@@ -288,8 +289,9 @@ public:
                  unsigned cleanupSize, unsigned fixupDepth,
                  EHScopeStack::stable_iterator enclosingNormal,
                  EHScopeStack::stable_iterator enclosingEH)
-    : EHScope(EHScope::Cleanup, enclosingEH), EnclosingNormal(enclosingNormal),
-      NormalBlock(nullptr), ActiveFlag(nullptr), ExtInfo(nullptr) {
+      : EHScope(EHScope::Cleanup, enclosingEH),
+        EnclosingNormal(enclosingNormal), NormalBlock(nullptr),
+        ActiveFlag(nullptr), ExtInfo(nullptr), FixupDepth(fixupDepth) {
     CleanupBits.IsNormalCleanup = isNormal;
     CleanupBits.IsEHCleanup = isEH;
     CleanupBits.IsActive = isActive;
@@ -297,7 +299,6 @@ public:
     CleanupBits.TestFlagInNormalCleanup = false;
     CleanupBits.TestFlagInEHCleanup = false;
     CleanupBits.CleanupSize = cleanupSize;
-    CleanupBits.FixupDepth = fixupDepth;
 
     assert(CleanupBits.CleanupSize == cleanupSize && "cleanup size overflow");
   }
@@ -343,7 +344,7 @@ public:
     return CleanupBits.TestFlagInEHCleanup;
   }
 
-  unsigned getFixupDepth() const { return CleanupBits.FixupDepth; }
+  unsigned getFixupDepth() const { return FixupDepth; }
   EHScopeStack::stable_iterator getEnclosingNormalCleanup() const {
     return EnclosingNormal;
   }
@@ -451,6 +452,7 @@ public:
   EHFilterScope(unsigned numFilters)
     : EHScope(Filter, EHScopeStack::stable_end()) {
     FilterBits.NumFilters = numFilters;
+    assert(FilterBits.NumFilters == numFilters && "NumFilters overflow");
   }
 
   static size_t getSizeForNumFilters(unsigned numFilters) {
@@ -540,7 +542,7 @@ public:
       Size = EHPadEndScope::getSize();
       break;
     }
-    Ptr += llvm::RoundUpToAlignment(Size, ScopeStackAlignment);
+    Ptr += llvm::alignTo(Size, ScopeStackAlignment);
     return *this;
   }
 
diff --git a/lib/CodeGen/CGDebugInfo.cpp b/lib/CodeGen/CGDebugInfo.cpp
index 5df85194878d..5e9d73f082fc 100644
--- a/lib/CodeGen/CGDebugInfo.cpp
+++ b/lib/CodeGen/CGDebugInfo.cpp
@@ -13,6 +13,7 @@
 
 #include "CGDebugInfo.h"
 #include "CGBlocks.h"
+#include "CGRecordLayout.h"
 #include "CGCXXABI.h"
 #include "CGObjCRuntime.h"
 #include "CodeGenFunction.h"
@@ -168,10 +169,10 @@ llvm::DIScope *CGDebugInfo::getContextDescriptor(const Decl *Context,
   }
 
   // Check namespace.
-  if (const NamespaceDecl *NSDecl = dyn_cast<NamespaceDecl>(Context))
+  if (const auto *NSDecl = dyn_cast<NamespaceDecl>(Context))
     return getOrCreateNameSpace(NSDecl);
 
-  if (const RecordDecl *RDecl = dyn_cast<RecordDecl>(Context))
+  if (const auto *RDecl = dyn_cast<RecordDecl>(Context))
     if (!RDecl->isDependentType())
       return getOrCreateType(CGM.getContext().getTypeDeclType(RDecl),
                              getOrCreateMainFile());
@@ -184,30 +185,32 @@ StringRef CGDebugInfo::getFunctionName(const FunctionDecl *FD) {
   FunctionTemplateSpecializationInfo *Info =
       FD->getTemplateSpecializationInfo();
 
-  if (!Info && FII && !CGM.getCodeGenOpts().EmitCodeView)
+  // Emit the unqualified name in normal operation. LLVM and the debugger can
+  // compute the fully qualified name from the scope chain. If we're only
+  // emitting line table info, there won't be any scope chains, so emit the
+  // fully qualified name here so that stack traces are more accurate.
+  // FIXME: Do this when emitting DWARF as well as when emitting CodeView after
+  // evaluating the size impact.
+  bool UseQualifiedName = DebugKind == codegenoptions::DebugLineTablesOnly &&
+                          CGM.getCodeGenOpts().EmitCodeView;
+
+  if (!Info && FII && !UseQualifiedName)
     return FII->getName();
 
-  // Otherwise construct human readable name for debug info.
   SmallString<128> NS;
   llvm::raw_svector_ostream OS(NS);
   PrintingPolicy Policy(CGM.getLangOpts());
-
-  if (CGM.getCodeGenOpts().EmitCodeView) {
-    // Print a fully qualified name like MSVC would.
-    Policy.MSVCFormatting = true;
-    FD->printQualifiedName(OS, Policy);
-  } else {
-    // Print the unqualified name with some template arguments. This is what
-    // DWARF-based debuggers expect.
+  Policy.MSVCFormatting = CGM.getCodeGenOpts().EmitCodeView;
+  if (!UseQualifiedName)
     FD->printName(OS);
-    // Add any template specialization args.
-    if (Info) {
-      const TemplateArgumentList *TArgs = Info->TemplateArguments;
-      const TemplateArgument *Args = TArgs->data();
-      unsigned NumArgs = TArgs->size();
-      TemplateSpecializationType::PrintTemplateArgumentList(OS, Args, NumArgs,
-                                                            Policy);
-    }
+  else
+    FD->printQualifiedName(OS, Policy);
+
+  // Add any template specialization args.
+  if (Info) {
+    const TemplateArgumentList *TArgs = Info->TemplateArguments;
+    TemplateSpecializationType::PrintTemplateArgumentList(OS, TArgs->asArray(),
+                                                          Policy);
   }
 
   // Copy this name on the side and use its reference.
@@ -219,21 +222,18 @@ StringRef CGDebugInfo::getObjCMethodName(const ObjCMethodDecl *OMD) {
   llvm::raw_svector_ostream OS(MethodName);
   OS << (OMD->isInstanceMethod() ? '-' : '+') << '[';
   const DeclContext *DC = OMD->getDeclContext();
-  if (const ObjCImplementationDecl *OID =
-          dyn_cast<const ObjCImplementationDecl>(DC)) {
+  if (const auto *OID = dyn_cast<ObjCImplementationDecl>(DC)) {
     OS << OID->getName();
-  } else if (const ObjCInterfaceDecl *OID =
-                 dyn_cast<const ObjCInterfaceDecl>(DC)) {
+  } else if (const auto *OID = dyn_cast<ObjCInterfaceDecl>(DC)) {
     OS << OID->getName();
-  } else if (const ObjCCategoryDecl *OC = dyn_cast<ObjCCategoryDecl>(DC)) {
+  } else if (const auto *OC = dyn_cast<ObjCCategoryDecl>(DC)) {
     if (OC->IsClassExtension()) {
       OS << OC->getClassInterface()->getName();
     } else {
-      OS << ((const NamedDecl *)OC)->getIdentifier()->getNameStart() << '('
+      OS << OC->getIdentifier()->getNameStart() << '('
          << OC->getIdentifier()->getNameStart() << ')';
     }
-  } else if (const ObjCCategoryImplDecl *OCD =
-                 dyn_cast<const ObjCCategoryImplDecl>(DC)) {
+  } else if (const auto *OCD = dyn_cast<ObjCCategoryImplDecl>(DC)) {
     OS << ((const NamedDecl *)OCD)->getIdentifier()->getNameStart() << '('
        << OCD->getIdentifier()->getNameStart() << ')';
   } else if (isa<ObjCProtocolDecl>(DC)) {
@@ -254,20 +254,56 @@ StringRef CGDebugInfo::getSelectorName(Selector S) {
 }
 
 StringRef CGDebugInfo::getClassName(const RecordDecl *RD) {
-  // quick optimization to avoid having to intern strings that are already
-  // stored reliably elsewhere
-  if (!isa<ClassTemplateSpecializationDecl>(RD))
-    return RD->getName();
-
-  SmallString<128> Name;
-  {
+  if (isa<ClassTemplateSpecializationDecl>(RD)) {
+    SmallString<128> Name;
     llvm::raw_svector_ostream OS(Name);
     RD->getNameForDiagnostic(OS, CGM.getContext().getPrintingPolicy(),
                              /*Qualified*/ false);
+
+    // Copy this name on the side and use its reference.
+    return internString(Name);
   }
 
-  // Copy this name on the side and use its reference.
-  return internString(Name);
+  // quick optimization to avoid having to intern strings that are already
+  // stored reliably elsewhere
+  if (const IdentifierInfo *II = RD->getIdentifier())
+    return II->getName();
+
+  // The CodeView printer in LLVM wants to see the names of unnamed types: it is
+  // used to reconstruct the fully qualified type names.
+  if (CGM.getCodeGenOpts().EmitCodeView) {
+    if (const TypedefNameDecl *D = RD->getTypedefNameForAnonDecl()) {
+      assert(RD->getDeclContext() == D->getDeclContext() &&
+             "Typedef should not be in another decl context!");
+      assert(D->getDeclName().getAsIdentifierInfo() &&
+             "Typedef was not named!");
+      return D->getDeclName().getAsIdentifierInfo()->getName();
+    }
+
+    if (CGM.getLangOpts().CPlusPlus) {
+      StringRef Name;
+
+      ASTContext &Context = CGM.getContext();
+      if (const DeclaratorDecl *DD = Context.getDeclaratorForUnnamedTagDecl(RD))
+        // Anonymous types without a name for linkage purposes have their
+        // declarator mangled in if they have one.
+        Name = DD->getName();
+      else if (const TypedefNameDecl *TND =
+                   Context.getTypedefNameForUnnamedTagDecl(RD))
+        // Anonymous types without a name for linkage purposes have their
+        // associate typedef mangled in if they have one.
+        Name = TND->getName();
+
+      if (!Name.empty()) {
+        SmallString<256> UnnamedType("<unnamed-type-");
+        UnnamedType += Name;
+        UnnamedType += '>';
+        return internString(UnnamedType);
+      }
+    }
+  }
+
+  return StringRef();
 }
 
 llvm::DIFile *CGDebugInfo::getOrCreateFile(SourceLocation Loc) {
@@ -383,6 +419,8 @@ void CGDebugInfo::CreateCompileUnit() {
       LangTag = llvm::dwarf::DW_LANG_C_plus_plus;
   } else if (LO.ObjC1) {
     LangTag = llvm::dwarf::DW_LANG_ObjC;
+  } else if (LO.RenderScript) {
+    LangTag = llvm::dwarf::DW_LANG_GOOGLE_RenderScript;
   } else if (LO.C99) {
     LangTag = llvm::dwarf::DW_LANG_C99;
   } else {
@@ -396,16 +434,27 @@ void CGDebugInfo::CreateCompileUnit() {
   if (LO.ObjC1)
     RuntimeVers = LO.ObjCRuntime.isNonFragile() ? 2 : 1;
 
+  llvm::DICompileUnit::DebugEmissionKind EmissionKind;
+  switch (DebugKind) {
+  case codegenoptions::NoDebugInfo:
+  case codegenoptions::LocTrackingOnly:
+    EmissionKind = llvm::DICompileUnit::NoDebug;
+    break;
+  case codegenoptions::DebugLineTablesOnly:
+    EmissionKind = llvm::DICompileUnit::LineTablesOnly;
+    break;
+  case codegenoptions::LimitedDebugInfo:
+  case codegenoptions::FullDebugInfo:
+    EmissionKind = llvm::DICompileUnit::FullDebug;
+    break;
+  }
+
   // Create new compile unit.
   // FIXME - Eliminate TheCU.
   TheCU = DBuilder.createCompileUnit(
       LangTag, remapDIPath(MainFileName), remapDIPath(getCurrentDirname()),
       Producer, LO.Optimize, CGM.getCodeGenOpts().DwarfDebugFlags, RuntimeVers,
-      CGM.getCodeGenOpts().SplitDwarfFile,
-      DebugKind <= CodeGenOptions::DebugLineTablesOnly
-          ? llvm::DIBuilder::LineTablesOnly
-          : llvm::DIBuilder::FullDebug,
-      0 /* DWOid */, DebugKind != CodeGenOptions::LocTrackingOnly);
+      CGM.getCodeGenOpts().SplitDwarfFile, EmissionKind, 0 /* DWOid */);
 }
 
 llvm::DIType *CGDebugInfo::CreateType(const BuiltinType *BT) {
@@ -463,39 +512,11 @@ llvm::DIType *CGDebugInfo::CreateType(const BuiltinType *BT) {
     return SelTy;
   }
 
-  case BuiltinType::OCLImage1d:
-    return getOrCreateStructPtrType("opencl_image1d_t", OCLImage1dDITy);
-  case BuiltinType::OCLImage1dArray:
-    return getOrCreateStructPtrType("opencl_image1d_array_t",
-                                    OCLImage1dArrayDITy);
-  case BuiltinType::OCLImage1dBuffer:
-    return getOrCreateStructPtrType("opencl_image1d_buffer_t",
-                                    OCLImage1dBufferDITy);
-  case BuiltinType::OCLImage2d:
-    return getOrCreateStructPtrType("opencl_image2d_t", OCLImage2dDITy);
-  case BuiltinType::OCLImage2dArray:
-    return getOrCreateStructPtrType("opencl_image2d_array_t",
-                                    OCLImage2dArrayDITy);
-  case BuiltinType::OCLImage2dDepth:
-    return getOrCreateStructPtrType("opencl_image2d_depth_t",
-                                    OCLImage2dDepthDITy);
-  case BuiltinType::OCLImage2dArrayDepth:
-    return getOrCreateStructPtrType("opencl_image2d_array_depth_t",
-                                    OCLImage2dArrayDepthDITy);
-  case BuiltinType::OCLImage2dMSAA:
-    return getOrCreateStructPtrType("opencl_image2d_msaa_t",
-                                    OCLImage2dMSAADITy);
-  case BuiltinType::OCLImage2dArrayMSAA:
-    return getOrCreateStructPtrType("opencl_image2d_array_msaa_t",
-                                    OCLImage2dArrayMSAADITy);
-  case BuiltinType::OCLImage2dMSAADepth:
-    return getOrCreateStructPtrType("opencl_image2d_msaa_depth_t",
-                                    OCLImage2dMSAADepthDITy);
-  case BuiltinType::OCLImage2dArrayMSAADepth:
-    return getOrCreateStructPtrType("opencl_image2d_array_msaa_depth_t",
-                                    OCLImage2dArrayMSAADepthDITy);
-  case BuiltinType::OCLImage3d:
-    return getOrCreateStructPtrType("opencl_image3d_t", OCLImage3dDITy);
+#define IMAGE_TYPE(ImgType, Id, SingletonId, Access, Suffix) \
+  case BuiltinType::Id: \
+    return getOrCreateStructPtrType("opencl_" #ImgType "_" #Suffix "_t", \
+                                    SingletonId);
+#include "clang/Basic/OpenCLImageTypes.def"
   case BuiltinType::OCLSampler:
     return DBuilder.createBasicType(
         "opencl_sampler_t", CGM.getContext().getTypeSize(BT),
@@ -545,7 +566,13 @@ llvm::DIType *CGDebugInfo::CreateType(const BuiltinType *BT) {
   case BuiltinType::Half:
   case BuiltinType::Float:
   case BuiltinType::LongDouble:
+  case BuiltinType::Float128:
   case BuiltinType::Double:
+    // FIXME: For targets where long double and __float128 have the same size,
+    // they are currently indistinguishable in the debugger without some
+    // special treatment. However, there is currently no consensus on encoding
+    // and this should be updated once a DWARF encoding exists for distinct
+    // floating point types of the same size.
     Encoding = llvm::dwarf::DW_ATE_float;
     break;
   }
@@ -660,10 +687,6 @@ static SmallString<256> getUniqueTagTypeName(const TagType *Ty,
   if (!hasCXXMangling(TD, TheCU) || !TD->isExternallyVisible())
     return FullName;
 
-  // Microsoft Mangler does not have support for mangleCXXRTTIName yet.
-  if (CGM.getTarget().getCXXABI().isMicrosoft())
-    return FullName;
-
   // TODO: This is using the RTTI name. Is there a better way to get
   // a unique string for a type?
   llvm::raw_svector_ostream Out(FullName);
@@ -817,10 +840,10 @@ llvm::DIType *CGDebugInfo::CreateType(const TemplateSpecializationType *Ty,
                               /*qualified*/ false);
 
   TemplateSpecializationType::PrintTemplateArgumentList(
-      OS, Ty->getArgs(), Ty->getNumArgs(),
+      OS, Ty->template_arguments(),
       CGM.getContext().getPrintingPolicy());
 
-  TypeAliasDecl *AliasDecl = cast<TypeAliasTemplateDecl>(
+  auto *AliasDecl = cast<TypeAliasTemplateDecl>(
       Ty->getTemplateName().getAsTemplateDecl())->getTemplatedDecl();
 
   SourceLocation Loc = AliasDecl->getLocation();
@@ -842,6 +865,39 @@ llvm::DIType *CGDebugInfo::CreateType(const TypedefType *Ty,
       getDeclContextDescriptor(Ty->getDecl()));
 }
 
+static unsigned getDwarfCC(CallingConv CC) {
+  switch (CC) {
+  case CC_C:
+    // Avoid emitting DW_AT_calling_convention if the C convention was used.
+    return 0;
+
+  case CC_X86StdCall:
+    return llvm::dwarf::DW_CC_BORLAND_stdcall;
+  case CC_X86FastCall:
+    return llvm::dwarf::DW_CC_BORLAND_msfastcall;
+  case CC_X86ThisCall:
+    return llvm::dwarf::DW_CC_BORLAND_thiscall;
+  case CC_X86VectorCall:
+    return llvm::dwarf::DW_CC_LLVM_vectorcall;
+  case CC_X86Pascal:
+    return llvm::dwarf::DW_CC_BORLAND_pascal;
+
+  // FIXME: Create new DW_CC_ codes for these calling conventions.
+  case CC_X86_64Win64:
+  case CC_X86_64SysV:
+  case CC_AAPCS:
+  case CC_AAPCS_VFP:
+  case CC_IntelOclBicc:
+  case CC_SpirFunction:
+  case CC_OpenCLKernel:
+  case CC_Swift:
+  case CC_PreserveMost:
+  case CC_PreserveAll:
+    return 0;
+  }
+  return 0;
+}
+
 llvm::DIType *CGDebugInfo::CreateType(const FunctionType *Ty,
                                       llvm::DIFile *Unit) {
   SmallVector<llvm::Metadata *, 16> EltTys;
@@ -853,15 +909,16 @@ llvm::DIType *CGDebugInfo::CreateType(const FunctionType *Ty,
   // otherwise emit it as a variadic function.
   if (isa<FunctionNoProtoType>(Ty))
     EltTys.push_back(DBuilder.createUnspecifiedParameter());
-  else if (const FunctionProtoType *FPT = dyn_cast<FunctionProtoType>(Ty)) {
-    for (unsigned i = 0, e = FPT->getNumParams(); i != e; ++i)
-      EltTys.push_back(getOrCreateType(FPT->getParamType(i), Unit));
+  else if (const auto *FPT = dyn_cast<FunctionProtoType>(Ty)) {
+    for (const QualType &ParamType : FPT->param_types())
+      EltTys.push_back(getOrCreateType(ParamType, Unit));
     if (FPT->isVariadic())
       EltTys.push_back(DBuilder.createUnspecifiedParameter());
   }
 
   llvm::DITypeRefArray EltTypeArray = DBuilder.getOrCreateTypeArray(EltTys);
-  return DBuilder.createSubroutineType(EltTypeArray);
+  return DBuilder.createSubroutineType(EltTypeArray, 0,
+                                       getDwarfCC(Ty->getCallConv()));
 }
 
 /// Convert an AccessSpecifier into the corresponding DINode flag.
@@ -890,10 +947,38 @@ static unsigned getAccessFlag(AccessSpecifier Access, const RecordDecl *RD) {
   llvm_unreachable("unexpected access enumerator");
 }
 
-llvm::DIType *CGDebugInfo::createFieldType(
-    StringRef name, QualType type, uint64_t sizeInBitsOverride,
-    SourceLocation loc, AccessSpecifier AS, uint64_t offsetInBits,
-    llvm::DIFile *tunit, llvm::DIScope *scope, const RecordDecl *RD) {
+llvm::DIType *CGDebugInfo::createBitFieldType(const FieldDecl *BitFieldDecl,
+                                              llvm::DIScope *RecordTy,
+                                              const RecordDecl *RD) {
+  StringRef Name = BitFieldDecl->getName();
+  QualType Ty = BitFieldDecl->getType();
+  SourceLocation Loc = BitFieldDecl->getLocation();
+  llvm::DIFile *VUnit = getOrCreateFile(Loc);
+  llvm::DIType *DebugType = getOrCreateType(Ty, VUnit);
+
+  // Get the location for the field.
+  llvm::DIFile *File = getOrCreateFile(Loc);
+  unsigned Line = getLineNumber(Loc);
+
+  const CGBitFieldInfo &BitFieldInfo =
+      CGM.getTypes().getCGRecordLayout(RD).getBitFieldInfo(BitFieldDecl);
+  uint64_t SizeInBits = BitFieldInfo.Size;
+  assert(SizeInBits > 0 && "found named 0-width bitfield");
+  unsigned AlignInBits = CGM.getContext().getTypeAlign(Ty);
+  uint64_t StorageOffsetInBits =
+      CGM.getContext().toBits(BitFieldInfo.StorageOffset);
+  uint64_t OffsetInBits = StorageOffsetInBits + BitFieldInfo.Offset;
+  unsigned Flags = getAccessFlag(BitFieldDecl->getAccess(), RD);
+  return DBuilder.createBitFieldMemberType(
+      RecordTy, Name, File, Line, SizeInBits, AlignInBits, OffsetInBits,
+      StorageOffsetInBits, Flags, DebugType);
+}
+
+llvm::DIType *
+CGDebugInfo::createFieldType(StringRef name, QualType type, SourceLocation loc,
+                             AccessSpecifier AS, uint64_t offsetInBits,
+                             llvm::DIFile *tunit, llvm::DIScope *scope,
+                             const RecordDecl *RD) {
   llvm::DIType *debugType = getOrCreateType(type, tunit);
 
   // Get the location for the field.
@@ -906,9 +991,6 @@ llvm::DIType *CGDebugInfo::createFieldType(
     TypeInfo TI = CGM.getContext().getTypeInfo(type);
     SizeInBits = TI.Width;
     AlignInBits = TI.Align;
-
-    if (sizeInBitsOverride)
-      SizeInBits = sizeInBitsOverride;
   }
 
   unsigned flags = getAccessFlag(AS, RD);
@@ -930,19 +1012,15 @@ void CGDebugInfo::CollectRecordLambdaFields(
        I != E; ++I, ++Field, ++fieldno) {
     const LambdaCapture &C = *I;
     if (C.capturesVariable()) {
+      SourceLocation Loc = C.getLocation();
+      assert(!Field->isBitField() && "lambdas don't have bitfield members!");
       VarDecl *V = C.getCapturedVar();
-      llvm::DIFile *VUnit = getOrCreateFile(C.getLocation());
       StringRef VName = V->getName();
-      uint64_t SizeInBitsOverride = 0;
-      if (Field->isBitField()) {
-        SizeInBitsOverride = Field->getBitWidthValue(CGM.getContext());
-        assert(SizeInBitsOverride && "found named 0-width bitfield");
-      }
-      llvm::DIType *fieldType = createFieldType(
-          VName, Field->getType(), SizeInBitsOverride, C.getLocation(),
-          Field->getAccess(), layout.getFieldOffset(fieldno), VUnit, RecordTy,
-          CXXDecl);
-      elements.push_back(fieldType);
+      llvm::DIFile *VUnit = getOrCreateFile(Loc);
+      llvm::DIType *FieldType = createFieldType(
+          VName, Field->getType(), Loc, Field->getAccess(),
+          layout.getFieldOffset(fieldno), VUnit, RecordTy, CXXDecl);
+      elements.push_back(FieldType);
     } else if (C.capturesThis()) {
       // TODO: Need to handle 'this' in some way by probably renaming the
       // this of the lambda class and having a field member of 'this' or
@@ -952,7 +1030,7 @@ void CGDebugInfo::CollectRecordLambdaFields(
       llvm::DIFile *VUnit = getOrCreateFile(f->getLocation());
       QualType type = f->getType();
       llvm::DIType *fieldType = createFieldType(
-          "this", type, 0, f->getLocation(), f->getAccess(),
+          "this", type, f->getLocation(), f->getAccess(),
           layout.getFieldOffset(fieldno), VUnit, RecordTy, CXXDecl);
 
       elements.push_back(fieldType);
@@ -1000,24 +1078,23 @@ void CGDebugInfo::CollectRecordNormalField(
   if (name.empty() && !type->isRecordType())
     return;
 
-  uint64_t SizeInBitsOverride = 0;
+  llvm::DIType *FieldType;
   if (field->isBitField()) {
-    SizeInBitsOverride = field->getBitWidthValue(CGM.getContext());
-    assert(SizeInBitsOverride && "found named 0-width bitfield");
+    FieldType = createBitFieldType(field, RecordTy, RD);
+  } else {
+    FieldType =
+        createFieldType(name, type, field->getLocation(), field->getAccess(),
+                        OffsetInBits, tunit, RecordTy, RD);
   }
 
-  llvm::DIType *fieldType =
-      createFieldType(name, type, SizeInBitsOverride, field->getLocation(),
-                      field->getAccess(), OffsetInBits, tunit, RecordTy, RD);
-
-  elements.push_back(fieldType);
+  elements.push_back(FieldType);
 }
 
 void CGDebugInfo::CollectRecordFields(
     const RecordDecl *record, llvm::DIFile *tunit,
     SmallVectorImpl<llvm::Metadata *> &elements,
     llvm::DICompositeType *RecordTy) {
-  const CXXRecordDecl *CXXDecl = dyn_cast<CXXRecordDecl>(record);
+  const auto *CXXDecl = dyn_cast<CXXRecordDecl>(record);
 
   if (CXXDecl && CXXDecl->isLambda())
     CollectRecordLambdaFields(CXXDecl, elements, RecordTy);
@@ -1031,6 +1108,8 @@ void CGDebugInfo::CollectRecordFields(
     // the corresponding declarations in the source program.
     for (const auto *I : record->decls())
       if (const auto *V = dyn_cast<VarDecl>(I)) {
+        if (V->hasAttr<NoDebugAttr>())
+          continue;
         // Reuse the existing static member declaration if one exists
         auto MI = StaticDataMemberCache.find(V->getCanonicalDecl());
         if (MI != StaticDataMemberCache.end()) {
@@ -1112,13 +1191,14 @@ llvm::DISubroutineType *CGDebugInfo::getOrCreateInstanceMethodType(
   if (Func->getExtProtoInfo().RefQualifier == RQ_RValue)
     Flags |= llvm::DINode::FlagRValueReference;
 
-  return DBuilder.createSubroutineType(EltTypeArray, Flags);
+  return DBuilder.createSubroutineType(EltTypeArray, Flags,
+                                       getDwarfCC(Func->getCallConv()));
 }
 
 /// isFunctionLocalClass - Return true if CXXRecordDecl is defined
 /// inside a function.
 static bool isFunctionLocalClass(const CXXRecordDecl *RD) {
-  if (const CXXRecordDecl *NRD = dyn_cast<CXXRecordDecl>(RD->getDeclContext()))
+  if (const auto *NRD = dyn_cast<CXXRecordDecl>(RD->getDeclContext()))
     return isFunctionLocalClass(NRD);
   if (isa<FunctionDecl>(RD->getDeclContext()))
     return true;
@@ -1136,6 +1216,11 @@ llvm::DISubprogram *CGDebugInfo::CreateCXXMemberFunction(
   // Since a single ctor/dtor corresponds to multiple functions, it doesn't
   // make sense to give a single ctor/dtor a linkage name.
   StringRef MethodLinkageName;
+  // FIXME: 'isFunctionLocalClass' seems like an arbitrary/unintentional
+  // property to use here. It may've been intended to model "is non-external
+  // type" but misses cases of non-function-local but non-external classes such
+  // as those in anonymous namespaces as well as the reverse - external types
+  // that are function local, such as those in (non-local) inline functions.
   if (!IsCtorOrDtor && !isFunctionLocalClass(Method->getParent()))
     MethodLinkageName = CGM.getMangledName(Method);
 
@@ -1151,6 +1236,8 @@ llvm::DISubprogram *CGDebugInfo::CreateCXXMemberFunction(
   llvm::DIType *ContainingType = nullptr;
   unsigned Virtuality = 0;
   unsigned VIndex = 0;
+  unsigned Flags = 0;
+  int ThisAdjustment = 0;
 
   if (Method->isVirtual()) {
     if (Method->isPure())
@@ -1158,26 +1245,45 @@ llvm::DISubprogram *CGDebugInfo::CreateCXXMemberFunction(
     else
       Virtuality = llvm::dwarf::DW_VIRTUALITY_virtual;
 
-    // It doesn't make sense to give a virtual destructor a vtable index,
-    // since a single destructor has two entries in the vtable.
-    // FIXME: Add proper support for debug info for virtual calls in
-    // the Microsoft ABI, where we may use multiple vptrs to make a vftable
-    // lookup if we have multiple or virtual inheritance.
-    if (!isa<CXXDestructorDecl>(Method) &&
-        !CGM.getTarget().getCXXABI().isMicrosoft())
-      VIndex = CGM.getItaniumVTableContext().getMethodVTableIndex(Method);
+    if (CGM.getTarget().getCXXABI().isItaniumFamily()) {
+      // It doesn't make sense to give a virtual destructor a vtable index,
+      // since a single destructor has two entries in the vtable.
+      if (!isa<CXXDestructorDecl>(Method))
+        VIndex = CGM.getItaniumVTableContext().getMethodVTableIndex(Method);
+    } else {
+      // Emit MS ABI vftable information.  There is only one entry for the
+      // deleting dtor.
+      const auto *DD = dyn_cast<CXXDestructorDecl>(Method);
+      GlobalDecl GD = DD ? GlobalDecl(DD, Dtor_Deleting) : GlobalDecl(Method);
+      MicrosoftVTableContext::MethodVFTableLocation ML =
+          CGM.getMicrosoftVTableContext().getMethodVFTableLocation(GD);
+      VIndex = ML.Index;
+
+      // CodeView only records the vftable offset in the class that introduces
+      // the virtual method. This is possible because, unlike Itanium, the MS
+      // C++ ABI does not include all virtual methods from non-primary bases in
+      // the vtable for the most derived class. For example, if C inherits from
+      // A and B, C's primary vftable will not include B's virtual methods.
+      if (Method->begin_overridden_methods() == Method->end_overridden_methods())
+        Flags |= llvm::DINode::FlagIntroducedVirtual;
+
+      // The 'this' adjustment accounts for both the virtual and non-virtual
+      // portions of the adjustment. Presumably the debugger only uses it when
+      // it knows the dynamic type of an object.
+      ThisAdjustment = CGM.getCXXABI()
+                           .getVirtualFunctionPrologueThisAdjustment(GD)
+                           .getQuantity();
+    }
     ContainingType = RecordTy;
   }
 
-  unsigned Flags = 0;
   if (Method->isImplicit())
     Flags |= llvm::DINode::FlagArtificial;
   Flags |= getAccessFlag(Method->getAccess(), Method->getParent());
-  if (const CXXConstructorDecl *CXXC = dyn_cast<CXXConstructorDecl>(Method)) {
+  if (const auto *CXXC = dyn_cast<CXXConstructorDecl>(Method)) {
     if (CXXC->isExplicit())
       Flags |= llvm::DINode::FlagExplicit;
-  } else if (const CXXConversionDecl *CXXC =
-                 dyn_cast<CXXConversionDecl>(Method)) {
+  } else if (const auto *CXXC = dyn_cast<CXXConversionDecl>(Method)) {
     if (CXXC->isExplicit())
       Flags |= llvm::DINode::FlagExplicit;
   }
@@ -1191,9 +1297,9 @@ llvm::DISubprogram *CGDebugInfo::CreateCXXMemberFunction(
   llvm::DINodeArray TParamsArray = CollectFunctionTemplateParams(Method, Unit);
   llvm::DISubprogram *SP = DBuilder.createMethod(
       RecordTy, MethodName, MethodLinkageName, MethodDefUnit, MethodLine,
-      MethodTy, /*isLocalToUnit=*/false,
-      /* isDefinition=*/false, Virtuality, VIndex, ContainingType, Flags,
-      CGM.getLangOpts().Optimize, TParamsArray.get());
+      MethodTy, /*isLocalToUnit=*/false, /*isDefinition=*/false, Virtuality,
+      VIndex, ThisAdjustment, ContainingType, Flags, CGM.getLangOpts().Optimize,
+      TParamsArray.get());
 
   SPCache[Method->getCanonicalDecl()].reset(SP);
 
@@ -1246,7 +1352,7 @@ void CGDebugInfo::CollectCXXBases(const CXXRecordDecl *RD, llvm::DIFile *Unit,
     unsigned BFlags = 0;
     uint64_t BaseOffset;
 
-    const CXXRecordDecl *Base =
+    const auto *Base =
         cast<CXXRecordDecl>(BI.getType()->getAs<RecordType>()->getDecl());
 
     if (BI.isVirtual()) {
@@ -1334,8 +1440,7 @@ CGDebugInfo::CollectTemplateParams(const TemplateParameterList *TPList,
       llvm::Constant *V = nullptr;
       // Special case member data pointer null values since they're actually -1
       // instead of zero.
-      if (const MemberPointerType *MPT =
-              dyn_cast<MemberPointerType>(T.getTypePtr()))
+      if (const auto *MPT = dyn_cast<MemberPointerType>(T.getTypePtr()))
         // But treat member function pointers as simple zero integers because
         // it's easier than having a special case in LLVM's CodeGen. If LLVM
         // CodeGen grows handling for values of non-null member function
@@ -1346,7 +1451,7 @@ CGDebugInfo::CollectTemplateParams(const TemplateParameterList *TPList,
       if (!V)
         V = llvm::ConstantInt::get(CGM.Int8Ty, 0);
       TemplateParams.push_back(DBuilder.createTemplateValueParameter(
-          TheCU, Name, TTy, cast<llvm::Constant>(V)));
+          TheCU, Name, TTy, V));
     } break;
     case TemplateArgument::Template:
       TemplateParams.push_back(DBuilder.createTemplateTemplateParameter(
@@ -1367,7 +1472,7 @@ CGDebugInfo::CollectTemplateParams(const TemplateParameterList *TPList,
       assert(V && "Expression in template argument isn't constant");
       llvm::DIType *TTy = getOrCreateType(T, Unit);
       TemplateParams.push_back(DBuilder.createTemplateValueParameter(
-          TheCU, Name, TTy, cast<llvm::Constant>(V->stripPointerCasts())));
+          TheCU, Name, TTy, V->stripPointerCasts()));
     } break;
     // And the following should never occur:
     case TemplateArgument::TemplateExpansion:
@@ -1446,7 +1551,7 @@ void CGDebugInfo::CollectVTableInfo(const CXXRecordDecl *RD, llvm::DIFile *Unit,
 
 llvm::DIType *CGDebugInfo::getOrCreateRecordType(QualType RTy,
                                                  SourceLocation Loc) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
   llvm::DIType *T = getOrCreateType(RTy, getOrCreateFile(Loc));
   return T;
 }
@@ -1458,22 +1563,17 @@ llvm::DIType *CGDebugInfo::getOrCreateInterfaceType(QualType D,
 
 llvm::DIType *CGDebugInfo::getOrCreateStandaloneType(QualType D,
                                                      SourceLocation Loc) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
   assert(!D.isNull() && "null type");
   llvm::DIType *T = getOrCreateType(D, getOrCreateFile(Loc));
   assert(T && "could not create debug info for type");
 
-  // Composite types with UIDs were already retained by DIBuilder
-  // because they are only referenced by name in the IR.
-  if (auto *CTy = dyn_cast<llvm::DICompositeType>(T))
-    if (!CTy->getIdentifier().empty())
-      return T;
   RetainedTypes.push_back(D.getAsOpaquePtr());
   return T;
 }
 
 void CGDebugInfo::completeType(const EnumDecl *ED) {
-  if (DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (DebugKind <= codegenoptions::DebugLineTablesOnly)
     return;
   QualType Ty = CGM.getContext().getEnumType(ED);
   void *TyPtr = Ty.getAsOpaquePtr();
@@ -1486,16 +1586,16 @@ void CGDebugInfo::completeType(const EnumDecl *ED) {
 }
 
 void CGDebugInfo::completeType(const RecordDecl *RD) {
-  if (DebugKind > CodeGenOptions::LimitedDebugInfo ||
+  if (DebugKind > codegenoptions::LimitedDebugInfo ||
       !CGM.getLangOpts().CPlusPlus)
     completeRequiredType(RD);
 }
 
 void CGDebugInfo::completeRequiredType(const RecordDecl *RD) {
-  if (DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (DebugKind <= codegenoptions::DebugLineTablesOnly)
     return;
 
-  if (const CXXRecordDecl *CXXDecl = dyn_cast<CXXRecordDecl>(RD))
+  if (const auto *CXXDecl = dyn_cast<CXXRecordDecl>(RD))
     if (CXXDecl->isDynamicClass())
       return;
 
@@ -1509,7 +1609,7 @@ void CGDebugInfo::completeRequiredType(const RecordDecl *RD) {
 }
 
 void CGDebugInfo::completeClassData(const RecordDecl *RD) {
-  if (DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (DebugKind <= codegenoptions::DebugLineTablesOnly)
     return;
   QualType Ty = CGM.getContext().getRecordType(RD);
   void *TyPtr = Ty.getAsOpaquePtr();
@@ -1523,23 +1623,38 @@ void CGDebugInfo::completeClassData(const RecordDecl *RD) {
 
 static bool hasExplicitMemberDefinition(CXXRecordDecl::method_iterator I,
                                         CXXRecordDecl::method_iterator End) {
-  for (; I != End; ++I)
-    if (FunctionDecl *Tmpl = I->getInstantiatedFromMemberFunction())
+  for (CXXMethodDecl *MD : llvm::make_range(I, End))
+    if (FunctionDecl *Tmpl = MD->getInstantiatedFromMemberFunction())
       if (!Tmpl->isImplicit() && Tmpl->isThisDeclarationADefinition() &&
-          !I->getMemberSpecializationInfo()->isExplicitSpecialization())
+          !MD->getMemberSpecializationInfo()->isExplicitSpecialization())
         return true;
   return false;
 }
 
-static bool shouldOmitDefinition(CodeGenOptions::DebugInfoKind DebugKind,
-                                 bool DebugTypeExtRefs,
-                                 const RecordDecl *RD,
+/// Does a type definition exist in an imported clang module?
+static bool isDefinedInClangModule(const RecordDecl *RD) {
+  if (!RD || !RD->isFromASTFile())
+    return false;
+  if (!RD->isExternallyVisible() && RD->getName().empty())
+    return false;
+  if (auto *CXXDecl = dyn_cast<CXXRecordDecl>(RD)) {
+    assert(CXXDecl->isCompleteDefinition() && "incomplete record definition");
+    if (CXXDecl->getTemplateSpecializationKind() != TSK_Undeclared)
+      // Make sure the instantiation is actually in a module.
+      if (CXXDecl->field_begin() != CXXDecl->field_end())
+        return CXXDecl->field_begin()->isFromASTFile();
+  }
+
+  return true;
+}
+
+static bool shouldOmitDefinition(codegenoptions::DebugInfoKind DebugKind,
+                                 bool DebugTypeExtRefs, const RecordDecl *RD,
                                  const LangOptions &LangOpts) {
-  // Does the type exist in an imported clang module?
-  if (DebugTypeExtRefs && RD->isFromASTFile() && RD->getDefinition())
-      return true;
+  if (DebugTypeExtRefs && isDefinedInClangModule(RD->getDefinition()))
+    return true;
 
-  if (DebugKind > CodeGenOptions::LimitedDebugInfo)
+  if (DebugKind > codegenoptions::LimitedDebugInfo)
     return false;
 
   if (!LangOpts.CPlusPlus)
@@ -1548,7 +1663,7 @@ static bool shouldOmitDefinition(CodeGenOptions::DebugInfoKind DebugKind,
   if (!RD->isCompleteDefinitionRequired())
     return true;
 
-  const CXXRecordDecl *CXXDecl = dyn_cast<CXXRecordDecl>(RD);
+  const auto *CXXDecl = dyn_cast<CXXRecordDecl>(RD);
 
   if (!CXXDecl)
     return false;
@@ -1557,8 +1672,7 @@ static bool shouldOmitDefinition(CodeGenOptions::DebugInfoKind DebugKind,
     return true;
 
   TemplateSpecializationKind Spec = TSK_Undeclared;
-  if (const ClassTemplateSpecializationDecl *SD =
-          dyn_cast<ClassTemplateSpecializationDecl>(RD))
+  if (const auto *SD = dyn_cast<ClassTemplateSpecializationDecl>(RD))
     Spec = SD->getSpecializationKind();
 
   if (Spec == TSK_ExplicitInstantiationDeclaration &&
@@ -1600,7 +1714,7 @@ llvm::DIType *CGDebugInfo::CreateTypeDefinition(const RecordType *Ty) {
   if (!D || !D->isCompleteDefinition())
     return FwdDecl;
 
-  if (const CXXRecordDecl *CXXDecl = dyn_cast<CXXRecordDecl>(RD))
+  if (const auto *CXXDecl = dyn_cast<CXXRecordDecl>(RD))
     CollectContainingType(CXXDecl, FwdDecl);
 
   // Push the struct on region stack.
@@ -1615,7 +1729,7 @@ llvm::DIType *CGDebugInfo::CreateTypeDefinition(const RecordType *Ty) {
   // gdb tests will depend on a certain ordering at printout. The debug
   // information offsets are still correct if we merge them all together
   // though.
-  const CXXRecordDecl *CXXDecl = dyn_cast<CXXRecordDecl>(RD);
+  const auto *CXXDecl = dyn_cast<CXXRecordDecl>(RD);
   if (CXXDecl) {
     CollectCXXBases(CXXDecl, DefUnit, EltTys, FwdDecl);
     CollectVTableInfo(CXXDecl, DefUnit, EltTys);
@@ -1676,8 +1790,11 @@ llvm::DIType *CGDebugInfo::CreateType(const ObjCInterfaceType *Ty,
   if (!ID)
     return nullptr;
 
-  // Return a forward declaration if this type was imported from a clang module.
-  if (DebugTypeExtRefs && ID->isFromASTFile() && ID->getDefinition())
+  // Return a forward declaration if this type was imported from a clang module,
+  // and this is not the compile unit with the implementation of the type (which
+  // may contain hidden ivars).
+  if (DebugTypeExtRefs && ID->isFromASTFile() && ID->getDefinition() &&
+      !ID->getImplementation())
     return DBuilder.createForwardDecl(llvm::dwarf::DW_TAG_structure_type,
                                       ID->getName(),
                                       getDeclContextDescriptor(ID), Unit, 0);
@@ -1739,11 +1856,14 @@ CGDebugInfo::getOrCreateModuleRef(ExternalASTSource::ASTSourceDescriptor Mod,
 
   bool IsRootModule = M ? !M->Parent : true;
   if (CreateSkeletonCU && IsRootModule) {
+    // PCH files don't have a signature field in the control block,
+    // but LLVM detects skeleton CUs by looking for a non-zero DWO id.
+    uint64_t Signature = Mod.getSignature() ? Mod.getSignature() : ~1ULL;
     llvm::DIBuilder DIB(CGM.getModule());
     DIB.createCompileUnit(TheCU->getSourceLanguage(), Mod.getModuleName(),
                           Mod.getPath(), TheCU->getProducer(), true,
                           StringRef(), 0, Mod.getASTFile(),
-                          llvm::DIBuilder::FullDebug, Mod.getSignature());
+                          llvm::DICompileUnit::FullDebug, Signature);
     DIB.finalize();
   }
   llvm::DIModule *Parent =
@@ -1942,7 +2062,7 @@ llvm::DIType *CGDebugInfo::CreateType(const ArrayType *Ty, llvm::DIFile *Unit) {
   uint64_t Align;
 
   // FIXME: make getTypeAlign() aware of VLAs and incomplete array types
-  if (const VariableArrayType *VAT = dyn_cast<VariableArrayType>(Ty)) {
+  if (const auto *VAT = dyn_cast<VariableArrayType>(Ty)) {
     Size = 0;
     Align =
         CGM.getContext().getTypeAlign(CGM.getContext().getBaseElementType(VAT));
@@ -1975,7 +2095,7 @@ llvm::DIType *CGDebugInfo::CreateType(const ArrayType *Ty, llvm::DIFile *Unit) {
     //     int x[0];
     //   };
     int64_t Count = -1; // Count == -1 is an unbounded array.
-    if (const ConstantArrayType *CAT = dyn_cast<ConstantArrayType>(Ty))
+    if (const auto *CAT = dyn_cast<ConstantArrayType>(Ty))
       Count = CAT->getSize().getZExtValue();
 
     // FIXME: Verify this is right for VLAs.
@@ -2003,12 +2123,35 @@ llvm::DIType *CGDebugInfo::CreateType(const RValueReferenceType *Ty,
 
 llvm::DIType *CGDebugInfo::CreateType(const MemberPointerType *Ty,
                                       llvm::DIFile *U) {
-  uint64_t Size =
-      !Ty->isIncompleteType() ? CGM.getContext().getTypeSize(Ty) : 0;
+  unsigned Flags = 0;
+  uint64_t Size = 0;
+
+  if (!Ty->isIncompleteType()) {
+    Size = CGM.getContext().getTypeSize(Ty);
+
+    // Set the MS inheritance model. There is no flag for the unspecified model.
+    if (CGM.getTarget().getCXXABI().isMicrosoft()) {
+      switch (Ty->getMostRecentCXXRecordDecl()->getMSInheritanceModel()) {
+      case MSInheritanceAttr::Keyword_single_inheritance:
+        Flags |= llvm::DINode::FlagSingleInheritance;
+        break;
+      case MSInheritanceAttr::Keyword_multiple_inheritance:
+        Flags |= llvm::DINode::FlagMultipleInheritance;
+        break;
+      case MSInheritanceAttr::Keyword_virtual_inheritance:
+        Flags |= llvm::DINode::FlagVirtualInheritance;
+        break;
+      case MSInheritanceAttr::Keyword_unspecified_inheritance:
+        break;
+      }
+    }
+  }
+
   llvm::DIType *ClassType = getOrCreateType(QualType(Ty->getClass(), 0), U);
   if (Ty->isMemberDataPointerType())
     return DBuilder.createMemberPointerType(
-        getOrCreateType(Ty->getPointeeType(), U), ClassType, Size);
+        getOrCreateType(Ty->getPointeeType(), U), ClassType, Size, /*Align=*/0,
+        Flags);
 
   const FunctionProtoType *FPT =
       Ty->getPointeeType()->getAs<FunctionProtoType>();
@@ -2016,7 +2159,7 @@ llvm::DIType *CGDebugInfo::CreateType(const MemberPointerType *Ty,
       getOrCreateInstanceMethodType(CGM.getContext().getPointerType(QualType(
                                         Ty->getClass(), FPT->getTypeQuals())),
                                     FPT, U),
-      ClassType, Size);
+      ClassType, Size, /*Align=*/0, Flags);
 }
 
 llvm::DIType *CGDebugInfo::CreateType(const AtomicType *Ty, llvm::DIFile *U) {
@@ -2048,13 +2191,23 @@ llvm::DIType *CGDebugInfo::CreateEnumType(const EnumType *Ty) {
   // If this is just a forward declaration, construct an appropriately
   // marked node and just return it.
   if (isImportedFromModule || !ED->getDefinition()) {
+    // Note that it is possible for enums to be created as part of
+    // their own declcontext. In this case a FwdDecl will be created
+    // twice. This doesn't cause a problem because both FwdDecls are
+    // entered into the ReplaceMap: finalize() will replace the first
+    // FwdDecl with the second and then replace the second with
+    // complete type.
     llvm::DIScope *EDContext = getDeclContextDescriptor(ED);
     llvm::DIFile *DefUnit = getOrCreateFile(ED->getLocation());
+    llvm::TempDIScope TmpContext(DBuilder.createReplaceableCompositeType(
+        llvm::dwarf::DW_TAG_enumeration_type, "", TheCU, DefUnit, 0));
+
     unsigned Line = getLineNumber(ED->getLocation());
     StringRef EDName = ED->getName();
     llvm::DIType *RetTy = DBuilder.createReplaceableCompositeType(
         llvm::dwarf::DW_TAG_enumeration_type, EDName, EDContext, DefUnit, Line,
         0, Size, Align, llvm::DINode::FlagFwdDecl, FullName);
+
     ReplaceMap.emplace_back(
         std::piecewise_construct, std::make_tuple(Ty),
         std::make_tuple(static_cast<llvm::Metadata *>(RetTy)));
@@ -2168,7 +2321,7 @@ llvm::DIType *CGDebugInfo::getTypeOrNull(QualType Ty) {
 
 void CGDebugInfo::completeTemplateDefinition(
     const ClassTemplateSpecializationDecl &SD) {
-  if (DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (DebugKind <= codegenoptions::DebugLineTablesOnly)
     return;
 
   completeClassData(&SD);
@@ -2220,8 +2373,12 @@ llvm::DIModule *CGDebugInfo::getParentModuleOrNull(const Decl *D) {
     // option.
     FullSourceLoc Loc(D->getLocation(), CGM.getContext().getSourceManager());
     if (Module *M = ClangModuleMap->inferModuleFromLocation(Loc)) {
+      // This is a (sub-)module.
       auto Info = ExternalASTSource::ASTSourceDescriptor(*M);
       return getOrCreateModuleRef(Info, /*SkeletonCU=*/false);
+    } else {
+      // This the precompiled header being built.
+      return getOrCreateModuleRef(PCHDescriptor, /*SkeletonCU=*/false);
     }
   }
 
@@ -2369,11 +2526,34 @@ llvm::DICompositeType *CGDebugInfo::CreateLimitedType(const RecordType *Ty) {
       getTagForRecord(RD), RDName, RDContext, DefUnit, Line, 0, Size, Align, 0,
       FullName);
 
+  // Elements of composite types usually have back to the type, creating
+  // uniquing cycles.  Distinct nodes are more efficient.
+  switch (RealDecl->getTag()) {
+  default:
+    llvm_unreachable("invalid composite type tag");
+
+  case llvm::dwarf::DW_TAG_array_type:
+  case llvm::dwarf::DW_TAG_enumeration_type:
+    // Array elements and most enumeration elements don't have back references,
+    // so they don't tend to be involved in uniquing cycles and there is some
+    // chance of merging them when linking together two modules.  Only make
+    // them distinct if they are ODR-uniqued.
+    if (FullName.empty())
+      break;
+
+  case llvm::dwarf::DW_TAG_structure_type:
+  case llvm::dwarf::DW_TAG_union_type:
+  case llvm::dwarf::DW_TAG_class_type:
+    // Immediatley resolve to a distinct node.
+    RealDecl =
+        llvm::MDNode::replaceWithDistinct(llvm::TempDICompositeType(RealDecl));
+    break;
+  }
+
   RegionMap[Ty->getDecl()].reset(RealDecl);
   TypeCache[QualType(Ty, 0).getAsOpaquePtr()].reset(RealDecl);
 
-  if (const ClassTemplateSpecializationDecl *TSpecial =
-          dyn_cast<ClassTemplateSpecializationDecl>(RD))
+  if (const auto *TSpecial = dyn_cast<ClassTemplateSpecializationDecl>(RD))
     DBuilder.replaceArrays(RealDecl, llvm::DINodeArray(),
                            CollectCXXTemplateParams(TSpecial, DefUnit));
   return RealDecl;
@@ -2420,7 +2600,7 @@ void CGDebugInfo::collectFunctionDeclProps(GlobalDecl GD, llvm::DIFile *Unit,
                                            llvm::DIScope *&FDContext,
                                            llvm::DINodeArray &TParamsArray,
                                            unsigned &Flags) {
-  const FunctionDecl *FD = cast<FunctionDecl>(GD.getDecl());
+  const auto *FD = cast<FunctionDecl>(GD.getDecl());
   Name = getFunctionName(FD);
   // Use mangled name as linkage name for C/C++ functions.
   if (FD->hasPrototype()) {
@@ -2430,13 +2610,12 @@ void CGDebugInfo::collectFunctionDeclProps(GlobalDecl GD, llvm::DIFile *Unit,
   // No need to replicate the linkage name if it isn't different from the
   // subprogram name, no need to have it at all unless coverage is enabled or
   // debug is set to more than just line tables.
-  if (LinkageName == Name ||
-      (!CGM.getCodeGenOpts().EmitGcovArcs &&
-       !CGM.getCodeGenOpts().EmitGcovNotes &&
-       DebugKind <= CodeGenOptions::DebugLineTablesOnly))
+  if (LinkageName == Name || (!CGM.getCodeGenOpts().EmitGcovArcs &&
+                              !CGM.getCodeGenOpts().EmitGcovNotes &&
+                              DebugKind <= codegenoptions::DebugLineTablesOnly))
     LinkageName = StringRef();
 
-  if (DebugKind >= CodeGenOptions::LimitedDebugInfo) {
+  if (DebugKind >= codegenoptions::LimitedDebugInfo) {
     if (const NamespaceDecl *NSDecl =
         dyn_cast_or_null<NamespaceDecl>(FD->getDeclContext()))
       FDContext = getOrCreateNameSpace(NSDecl);
@@ -2513,15 +2692,15 @@ CGDebugInfo::getFunctionForwardDeclaration(const FunctionDecl *FD) {
   SmallVector<QualType, 16> ArgTypes;
   for (const ParmVarDecl *Parm: FD->parameters())
     ArgTypes.push_back(Parm->getType());
-  QualType FnType =
-    CGM.getContext().getFunctionType(FD->getReturnType(), ArgTypes,
-                                     FunctionProtoType::ExtProtoInfo());
+  CallingConv CC = FD->getType()->castAs<FunctionType>()->getCallConv();
+  QualType FnType = CGM.getContext().getFunctionType(
+      FD->getReturnType(), ArgTypes, FunctionProtoType::ExtProtoInfo(CC));
   llvm::DISubprogram *SP = DBuilder.createTempFunctionFwdDecl(
       DContext, Name, LinkageName, Unit, Line,
       getOrCreateFunctionType(FD, FnType, Unit), !FD->isExternallyVisible(),
       /* isDefinition = */ false, 0, Flags, CGM.getLangOpts().Optimize,
       TParamsArray.get(), getFunctionDeclaration(FD));
-  const FunctionDecl *CanonDecl = cast<FunctionDecl>(FD->getCanonicalDecl());
+  const auto *CanonDecl = cast<FunctionDecl>(FD->getCanonicalDecl());
   FwdDeclReplaceMap.emplace_back(std::piecewise_construct,
                                  std::make_tuple(CanonDecl),
                                  std::make_tuple(SP));
@@ -2553,7 +2732,7 @@ llvm::DINode *CGDebugInfo::getDeclarationOrDefinition(const Decl *D) {
   // we would otherwise do to get a type for a pointee. (forward declarations in
   // limited debug info, full definitions (if the type definition is available)
   // in unlimited debug info)
-  if (const TypeDecl *TD = dyn_cast<TypeDecl>(D))
+  if (const auto *TD = dyn_cast<TypeDecl>(D))
     return getOrCreateType(CGM.getContext().getTypeDeclType(TD),
                            getOrCreateFile(TD->getLocation()));
   auto I = DeclCache.find(D->getCanonicalDecl());
@@ -2563,7 +2742,7 @@ llvm::DINode *CGDebugInfo::getDeclarationOrDefinition(const Decl *D) {
 
   // No definition for now. Emit a forward definition that might be
   // merged with a potential upcoming definition.
-  if (const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(D))
+  if (const auto *FD = dyn_cast<FunctionDecl>(D))
     return getFunctionForwardDeclaration(FD);
   else if (const auto *VD = dyn_cast<VarDecl>(D))
     return getGlobalVariableForwardDeclaration(VD);
@@ -2572,10 +2751,10 @@ llvm::DINode *CGDebugInfo::getDeclarationOrDefinition(const Decl *D) {
 }
 
 llvm::DISubprogram *CGDebugInfo::getFunctionDeclaration(const Decl *D) {
-  if (!D || DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (!D || DebugKind <= codegenoptions::DebugLineTablesOnly)
     return nullptr;
 
-  const FunctionDecl *FD = dyn_cast<FunctionDecl>(D);
+  const auto *FD = dyn_cast<FunctionDecl>(D);
   if (!FD)
     return nullptr;
 
@@ -2584,8 +2763,7 @@ llvm::DISubprogram *CGDebugInfo::getFunctionDeclaration(const Decl *D) {
 
   auto MI = SPCache.find(FD->getCanonicalDecl());
   if (MI == SPCache.end()) {
-    if (const CXXMethodDecl *MD =
-            dyn_cast<CXXMethodDecl>(FD->getCanonicalDecl())) {
+    if (const auto *MD = dyn_cast<CXXMethodDecl>(FD->getCanonicalDecl())) {
       return CreateCXXMemberFunction(MD, getOrCreateFile(MD->getLocation()),
                                      cast<llvm::DICompositeType>(S));
     }
@@ -2612,14 +2790,18 @@ llvm::DISubprogram *CGDebugInfo::getFunctionDeclaration(const Decl *D) {
 llvm::DISubroutineType *CGDebugInfo::getOrCreateFunctionType(const Decl *D,
                                                              QualType FnType,
                                                              llvm::DIFile *F) {
-  if (!D || DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (!D || DebugKind <= codegenoptions::DebugLineTablesOnly)
     // Create fake but valid subroutine type. Otherwise -verify would fail, and
     // subprogram DIE will miss DW_AT_decl_file and DW_AT_decl_line fields.
     return DBuilder.createSubroutineType(DBuilder.getOrCreateTypeArray(None));
 
-  if (const CXXMethodDecl *Method = dyn_cast<CXXMethodDecl>(D))
+  if (const auto *Method = dyn_cast<CXXMethodDecl>(D))
     return getOrCreateMethodType(Method, F);
-  if (const ObjCMethodDecl *OMethod = dyn_cast<ObjCMethodDecl>(D)) {
+
+  const auto *FTy = FnType->getAs<FunctionType>();
+  CallingConv CC = FTy ? FTy->getCallConv() : CallingConv::CC_C;
+
+  if (const auto *OMethod = dyn_cast<ObjCMethodDecl>(D)) {
     // Add "self" and "_cmd"
     SmallVector<llvm::Metadata *, 16> Elts;
 
@@ -2645,28 +2827,28 @@ llvm::DISubroutineType *CGDebugInfo::getOrCreateFunctionType(const Decl *D,
     Elts.push_back(DBuilder.createArtificialType(
         getOrCreateType(CGM.getContext().getObjCSelType(), F)));
     // Get rest of the arguments.
-    for (const auto *PI : OMethod->params())
+    for (const auto *PI : OMethod->parameters())
       Elts.push_back(getOrCreateType(PI->getType(), F));
     // Variadic methods need a special marker at the end of the type list.
     if (OMethod->isVariadic())
       Elts.push_back(DBuilder.createUnspecifiedParameter());
 
     llvm::DITypeRefArray EltTypeArray = DBuilder.getOrCreateTypeArray(Elts);
-    return DBuilder.createSubroutineType(EltTypeArray);
+    return DBuilder.createSubroutineType(EltTypeArray, 0, getDwarfCC(CC));
   }
 
   // Handle variadic function types; they need an additional
   // unspecified parameter.
-  if (const FunctionDecl *FD = dyn_cast<FunctionDecl>(D))
+  if (const auto *FD = dyn_cast<FunctionDecl>(D))
     if (FD->isVariadic()) {
       SmallVector<llvm::Metadata *, 16> EltTys;
       EltTys.push_back(getOrCreateType(FD->getReturnType(), F));
-      if (const FunctionProtoType *FPT = dyn_cast<FunctionProtoType>(FnType))
-        for (unsigned i = 0, e = FPT->getNumParams(); i != e; ++i)
-          EltTys.push_back(getOrCreateType(FPT->getParamType(i), F));
+      if (const auto *FPT = dyn_cast<FunctionProtoType>(FnType))
+        for (QualType ParamType : FPT->param_types())
+          EltTys.push_back(getOrCreateType(ParamType, F));
       EltTys.push_back(DBuilder.createUnspecifiedParameter());
       llvm::DITypeRefArray EltTypeArray = DBuilder.getOrCreateTypeArray(EltTys);
-      return DBuilder.createSubroutineType(EltTypeArray);
+      return DBuilder.createSubroutineType(EltTypeArray, 0, getDwarfCC(CC));
     }
 
   return cast<llvm::DISubroutineType>(getOrCreateType(FnType, F));
@@ -2691,7 +2873,7 @@ void CGDebugInfo::EmitFunctionStart(GlobalDecl GD, SourceLocation Loc,
   if (!HasDecl) {
     // Use llvm function name.
     LinkageName = Fn->getName();
-  } else if (const FunctionDecl *FD = dyn_cast<FunctionDecl>(D)) {
+  } else if (const auto *FD = dyn_cast<FunctionDecl>(D)) {
     // If there is a subprogram for this function available then use it.
     auto FI = SPCache.find(FD->getCanonicalDecl());
     if (FI != SPCache.end()) {
@@ -2704,7 +2886,7 @@ void CGDebugInfo::EmitFunctionStart(GlobalDecl GD, SourceLocation Loc,
     }
     collectFunctionDeclProps(GD, Unit, Name, LinkageName, FDContext,
                              TParamsArray, Flags);
-  } else if (const ObjCMethodDecl *OMD = dyn_cast<ObjCMethodDecl>(D)) {
+  } else if (const auto *OMD = dyn_cast<ObjCMethodDecl>(D)) {
     Name = getObjCMethodName(OMD);
     Flags |= llvm::DINode::FlagPrototyped;
   } else {
@@ -2712,7 +2894,7 @@ void CGDebugInfo::EmitFunctionStart(GlobalDecl GD, SourceLocation Loc,
     Name = Fn->getName();
     Flags |= llvm::DINode::FlagPrototyped;
   }
-  if (!Name.empty() && Name[0] == '\01')
+  if (Name.startswith("\01"))
     Name = Name.substr(1);
 
   if (!HasDecl || D->isImplicit()) {
@@ -2731,7 +2913,7 @@ void CGDebugInfo::EmitFunctionStart(GlobalDecl GD, SourceLocation Loc,
   // are emitted as CU level entities by the backend.
   llvm::DISubprogram *SP = DBuilder.createFunction(
       FDContext, Name, LinkageName, Unit, LineNo,
-      getOrCreateFunctionType(D, FnType, Unit), Fn->hasInternalLinkage(),
+      getOrCreateFunctionType(D, FnType, Unit), Fn->hasLocalLinkage(),
       true /*definition*/, ScopeLine, Flags, CGM.getLangOpts().Optimize,
       TParamsArray.get(), getFunctionDeclaration(D));
   Fn->setSubprogram(SP);
@@ -2739,7 +2921,7 @@ void CGDebugInfo::EmitFunctionStart(GlobalDecl GD, SourceLocation Loc,
   // code for the initialization of globals. Do not record these decls
   // as they will overwrite the actual VarDecl Decl in the cache.
   if (HasDecl && isa<FunctionDecl>(D))
-    DeclCache[D->getCanonicalDecl()].reset(static_cast<llvm::Metadata *>(SP));
+    DeclCache[D->getCanonicalDecl()].reset(SP);
 
   // Push the function onto the lexical block stack.
   LexicalBlockStack.emplace_back(SP);
@@ -2765,7 +2947,7 @@ void CGDebugInfo::EmitFunctionDecl(GlobalDecl GD, SourceLocation Loc,
     // If there is a DISubprogram for this function available then use it.
     collectFunctionDeclProps(GD, Unit, Name, LinkageName, FDContext,
                              TParamsArray, Flags);
-  } else if (const ObjCMethodDecl *OMD = dyn_cast<ObjCMethodDecl>(D)) {
+  } else if (const auto *OMD = dyn_cast<ObjCMethodDecl>(D)) {
     Name = getObjCMethodName(OMD);
     Flags |= llvm::DINode::FlagPrototyped;
   } else {
@@ -2783,11 +2965,11 @@ void CGDebugInfo::EmitFunctionDecl(GlobalDecl GD, SourceLocation Loc,
   unsigned LineNo = getLineNumber(Loc);
   unsigned ScopeLine = 0;
 
-  DBuilder.createFunction(FDContext, Name, LinkageName, Unit, LineNo,
-                          getOrCreateFunctionType(D, FnType, Unit),
-                          false /*internalLinkage*/, true /*definition*/,
-                          ScopeLine, Flags, CGM.getLangOpts().Optimize,
-                          TParamsArray.get(), getFunctionDeclaration(D));
+  DBuilder.retainType(DBuilder.createFunction(
+      FDContext, Name, LinkageName, Unit, LineNo,
+      getOrCreateFunctionType(D, FnType, Unit), false /*internalLinkage*/,
+      false /*definition*/, ScopeLine, Flags, CGM.getLangOpts().Optimize,
+      TParamsArray.get(), getFunctionDeclaration(D)));
 }
 
 void CGDebugInfo::EmitLocation(CGBuilderTy &Builder, SourceLocation Loc) {
@@ -2820,7 +3002,7 @@ void CGDebugInfo::EmitLexicalBlockStart(CGBuilderTy &Builder,
   Builder.SetCurrentDebugLocation(llvm::DebugLoc::get(
       getLineNumber(Loc), getColumnNumber(Loc), LexicalBlockStack.back()));
 
-  if (DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (DebugKind <= codegenoptions::DebugLineTablesOnly)
     return;
 
   // Create a new lexical block and push it on the stack.
@@ -2834,7 +3016,7 @@ void CGDebugInfo::EmitLexicalBlockEnd(CGBuilderTy &Builder,
   // Provide an entry in the line table for the end of the block.
   EmitLocation(Builder, Loc);
 
-  if (DebugKind <= CodeGenOptions::DebugLineTablesOnly)
+  if (DebugKind <= codegenoptions::DebugLineTablesOnly)
     return;
 
   LexicalBlockStack.pop_back();
@@ -2896,8 +3078,7 @@ llvm::DIType *CGDebugInfo::EmitTypeForVarWithBlocksAttr(const VarDecl *VD,
                   CGM.getTarget().getPointerAlign(0))) {
     CharUnits FieldOffsetInBytes =
         CGM.getContext().toCharUnitsFromBits(FieldOffset);
-    CharUnits AlignedOffsetInBytes =
-        FieldOffsetInBytes.RoundUpToAlignment(Align);
+    CharUnits AlignedOffsetInBytes = FieldOffsetInBytes.alignTo(Align);
     CharUnits NumPaddingBytes = AlignedOffsetInBytes - FieldOffsetInBytes;
 
     if (NumPaddingBytes.isPositive()) {
@@ -2930,8 +3111,10 @@ llvm::DIType *CGDebugInfo::EmitTypeForVarWithBlocksAttr(const VarDecl *VD,
 void CGDebugInfo::EmitDeclare(const VarDecl *VD, llvm::Value *Storage,
                               llvm::Optional<unsigned> ArgNo,
                               CGBuilderTy &Builder) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
   assert(!LexicalBlockStack.empty() && "Region stack mismatch, stack empty!");
+  if (VD->hasAttr<NoDebugAttr>())
+    return;
 
   bool Unwritten =
       VD->isImplicit() || (isa<Decl>(VD->getDeclContext()) &&
@@ -2969,7 +3152,7 @@ void CGDebugInfo::EmitDeclare(const VarDecl *VD, llvm::Value *Storage,
   // otherwise it is 'self' or 'this'.
   if (isa<ImplicitParamDecl>(VD) && ArgNo && *ArgNo == 1)
     Flags |= llvm::DINode::FlagObjectPointer;
-  if (llvm::Argument *Arg = dyn_cast<llvm::Argument>(Storage))
+  if (auto *Arg = dyn_cast<llvm::Argument>(Storage))
     if (Arg->getType()->isPointerTy() && !Arg->hasByValAttr() &&
         !VD->getType()->isPointerType())
       Expr.push_back(llvm::dwarf::DW_OP_deref);
@@ -3005,10 +3188,10 @@ void CGDebugInfo::EmitDeclare(const VarDecl *VD, llvm::Value *Storage,
       return;
     } else if (isa<VariableArrayType>(VD->getType()))
       Expr.push_back(llvm::dwarf::DW_OP_deref);
-  } else if (const RecordType *RT = dyn_cast<RecordType>(VD->getType())) {
+  } else if (const auto *RT = dyn_cast<RecordType>(VD->getType())) {
     // If VD is an anonymous union then Storage represents value for
     // all union fields.
-    const RecordDecl *RD = cast<RecordDecl>(RT->getDecl());
+    const auto *RD = cast<RecordDecl>(RT->getDecl());
     if (RD->isUnion() && RD->isAnonymousStructOrUnion()) {
       // GDB has trouble finding local variables in anonymous unions, so we emit
       // artifical local variables for each of the members.
@@ -3056,7 +3239,7 @@ void CGDebugInfo::EmitDeclare(const VarDecl *VD, llvm::Value *Storage,
 void CGDebugInfo::EmitDeclareOfAutoVariable(const VarDecl *VD,
                                             llvm::Value *Storage,
                                             CGBuilderTy &Builder) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
   EmitDeclare(VD, Storage, llvm::None, Builder);
 }
 
@@ -3071,11 +3254,13 @@ llvm::DIType *CGDebugInfo::CreateSelfType(const QualType &QualTy,
 void CGDebugInfo::EmitDeclareOfBlockDeclRefVariable(
     const VarDecl *VD, llvm::Value *Storage, CGBuilderTy &Builder,
     const CGBlockInfo &blockInfo, llvm::Instruction *InsertPoint) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
   assert(!LexicalBlockStack.empty() && "Region stack mismatch, stack empty!");
 
   if (Builder.GetInsertBlock() == nullptr)
     return;
+  if (VD->hasAttr<NoDebugAttr>())
+    return;
 
   bool isByRef = VD->hasAttr<BlocksAttr>();
 
@@ -3139,7 +3324,7 @@ void CGDebugInfo::EmitDeclareOfBlockDeclRefVariable(
 void CGDebugInfo::EmitDeclareOfArgVariable(const VarDecl *VD, llvm::Value *AI,
                                            unsigned ArgNo,
                                            CGBuilderTy &Builder) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
   EmitDeclare(VD, AI, ArgNo, Builder);
 }
 
@@ -3158,7 +3343,7 @@ void CGDebugInfo::EmitDeclareOfBlockLiteralArgVariable(const CGBlockInfo &block,
                                                        unsigned ArgNo,
                                                        llvm::Value *LocalAddr,
                                                        CGBuilderTy &Builder) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
   ASTContext &C = CGM.getContext();
   const BlockDecl *blockDecl = block.getBlockDecl();
 
@@ -3175,25 +3360,25 @@ void CGDebugInfo::EmitDeclareOfBlockLiteralArgVariable(const CGBlockInfo &block,
       CGM.getDataLayout().getStructLayout(block.StructureType);
 
   SmallVector<llvm::Metadata *, 16> fields;
-  fields.push_back(createFieldType("__isa", C.VoidPtrTy, 0, loc, AS_public,
+  fields.push_back(createFieldType("__isa", C.VoidPtrTy, loc, AS_public,
                                    blockLayout->getElementOffsetInBits(0),
                                    tunit, tunit));
-  fields.push_back(createFieldType("__flags", C.IntTy, 0, loc, AS_public,
+  fields.push_back(createFieldType("__flags", C.IntTy, loc, AS_public,
                                    blockLayout->getElementOffsetInBits(1),
                                    tunit, tunit));
-  fields.push_back(createFieldType("__reserved", C.IntTy, 0, loc, AS_public,
+  fields.push_back(createFieldType("__reserved", C.IntTy, loc, AS_public,
                                    blockLayout->getElementOffsetInBits(2),
                                    tunit, tunit));
   auto *FnTy = block.getBlockExpr()->getFunctionType();
   auto FnPtrType = CGM.getContext().getPointerType(FnTy->desugar());
-  fields.push_back(createFieldType("__FuncPtr", FnPtrType, 0, loc, AS_public,
+  fields.push_back(createFieldType("__FuncPtr", FnPtrType, loc, AS_public,
                                    blockLayout->getElementOffsetInBits(3),
                                    tunit, tunit));
   fields.push_back(createFieldType(
       "__descriptor", C.getPointerType(block.NeedsCopyDispose
                                            ? C.getBlockDescriptorExtendedType()
                                            : C.getBlockDescriptorType()),
-      0, loc, AS_public, blockLayout->getElementOffsetInBits(4), tunit, tunit));
+      loc, AS_public, blockLayout->getElementOffsetInBits(4), tunit, tunit));
 
   // We want to sort the captures by offset, not because DWARF
   // requires this, but because we're paranoid about debuggers.
@@ -3227,19 +3412,22 @@ void CGDebugInfo::EmitDeclareOfBlockLiteralArgVariable(const CGBlockInfo &block,
   // Sort by offset.
   llvm::array_pod_sort(chunks.begin(), chunks.end());
 
-  for (SmallVectorImpl<BlockLayoutChunk>::iterator i = chunks.begin(),
-                                                   e = chunks.end();
-       i != e; ++i) {
-    uint64_t offsetInBits = i->OffsetInBits;
-    const BlockDecl::Capture *capture = i->Capture;
+  for (const BlockLayoutChunk &Chunk : chunks) {
+    uint64_t offsetInBits = Chunk.OffsetInBits;
+    const BlockDecl::Capture *capture = Chunk.Capture;
 
     // If we have a null capture, this must be the C++ 'this' capture.
     if (!capture) {
-      const CXXMethodDecl *method =
-          cast<CXXMethodDecl>(blockDecl->getNonClosureContext());
-      QualType type = method->getThisType(C);
+      QualType type;
+      if (auto *Method =
+              cast_or_null<CXXMethodDecl>(blockDecl->getNonClosureContext()))
+        type = Method->getThisType(C);
+      else if (auto *RDecl = dyn_cast<CXXRecordDecl>(blockDecl->getParent()))
+        type = QualType(RDecl->getTypeForDecl(), 0);
+      else
+        llvm_unreachable("unexpected block declcontext");
 
-      fields.push_back(createFieldType("this", type, 0, loc, AS_public,
+      fields.push_back(createFieldType("this", type, loc, AS_public,
                                        offsetInBits, tunit, tunit));
       continue;
     }
@@ -3259,7 +3447,7 @@ void CGDebugInfo::EmitDeclareOfBlockLiteralArgVariable(const CGBlockInfo &block,
           DBuilder.createMemberType(tunit, name, tunit, line, PtrInfo.Width,
                                     PtrInfo.Align, offsetInBits, 0, fieldType);
     } else {
-      fieldType = createFieldType(name, variable->getType(), 0, loc, AS_public,
+      fieldType = createFieldType(name, variable->getType(), loc, AS_public,
                                   offsetInBits, tunit, tunit);
     }
     fields.push_back(fieldType);
@@ -3328,8 +3516,7 @@ llvm::DIGlobalVariable *CGDebugInfo::CollectAnonRecordDecls(
 
     // Ignore unnamed fields, but recurse into anonymous records.
     if (FieldName.empty()) {
-      const RecordType *RT = dyn_cast<RecordType>(Field->getType());
-      if (RT)
+      if (const auto *RT = dyn_cast<RecordType>(Field->getType()))
         GV = CollectAnonRecordDecls(RT->getDecl(), Unit, LineNo, LinkageName,
                                     Var, DContext);
       continue;
@@ -3337,14 +3524,16 @@ llvm::DIGlobalVariable *CGDebugInfo::CollectAnonRecordDecls(
     // Use VarDecl's Tag, Scope and Line number.
     GV = DBuilder.createGlobalVariable(DContext, FieldName, LinkageName, Unit,
                                        LineNo, FieldTy,
-                                       Var->hasInternalLinkage(), Var, nullptr);
+                                       Var->hasLocalLinkage(), Var, nullptr);
   }
   return GV;
 }
 
 void CGDebugInfo::EmitGlobalVariable(llvm::GlobalVariable *Var,
                                      const VarDecl *D) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
+  if (D->hasAttr<NoDebugAttr>())
+    return;
   // Create global variable debug descriptor.
   llvm::DIFile *Unit = nullptr;
   llvm::DIScope *DContext = nullptr;
@@ -3368,21 +3557,23 @@ void CGDebugInfo::EmitGlobalVariable(llvm::GlobalVariable *Var,
   } else {
     GV = DBuilder.createGlobalVariable(
         DContext, DeclName, LinkageName, Unit, LineNo, getOrCreateType(T, Unit),
-        Var->hasInternalLinkage(), Var,
+        Var->hasLocalLinkage(), Var,
         getOrCreateStaticDataMemberDeclarationOrNull(D));
   }
-  DeclCache[D->getCanonicalDecl()].reset(static_cast<llvm::Metadata *>(GV));
+  DeclCache[D->getCanonicalDecl()].reset(GV);
 }
 
 void CGDebugInfo::EmitGlobalVariable(const ValueDecl *VD,
                                      llvm::Constant *Init) {
-  assert(DebugKind >= CodeGenOptions::LimitedDebugInfo);
+  assert(DebugKind >= codegenoptions::LimitedDebugInfo);
+  if (VD->hasAttr<NoDebugAttr>())
+    return;
   // Create the descriptor for the variable.
   llvm::DIFile *Unit = getOrCreateFile(VD->getLocation());
   StringRef Name = VD->getName();
   llvm::DIType *Ty = getOrCreateType(VD->getType(), Unit);
-  if (const EnumConstantDecl *ECD = dyn_cast<EnumConstantDecl>(VD)) {
-    const EnumDecl *ED = cast<EnumDecl>(ECD->getDeclContext());
+  if (const auto *ECD = dyn_cast<EnumConstantDecl>(VD)) {
+    const auto *ED = cast<EnumDecl>(ECD->getDeclContext());
     assert(isa<EnumType>(ED->getTypeForDecl()) && "Enum without EnumType?");
     Ty = getOrCreateType(QualType(ED->getTypeForDecl(), 0), Unit);
   }
@@ -3400,6 +3591,9 @@ void CGDebugInfo::EmitGlobalVariable(const ValueDecl *VD,
     auto *RD = cast<RecordDecl>(VarD->getDeclContext());
     getDeclContextDescriptor(VarD);
     // Ensure that the type is retained even though it's otherwise unreferenced.
+    //
+    // FIXME: This is probably unnecessary, since Ty should reference RD
+    // through its scope.
     RetainedTypes.push_back(
         CGM.getContext().getRecordType(RD).getAsOpaquePtr());
     return;
@@ -3423,7 +3617,7 @@ llvm::DIScope *CGDebugInfo::getCurrentContextDescriptor(const Decl *D) {
 }
 
 void CGDebugInfo::EmitUsingDirective(const UsingDirectiveDecl &UD) {
-  if (CGM.getCodeGenOpts().getDebugInfo() < CodeGenOptions::LimitedDebugInfo)
+  if (CGM.getCodeGenOpts().getDebugInfo() < codegenoptions::LimitedDebugInfo)
     return;
   const NamespaceDecl *NSDecl = UD.getNominatedNamespace();
   if (!NSDecl->isAnonymousNamespace() || 
@@ -3436,7 +3630,7 @@ void CGDebugInfo::EmitUsingDirective(const UsingDirectiveDecl &UD) {
 }
 
 void CGDebugInfo::EmitUsingDecl(const UsingDecl &UD) {
-  if (CGM.getCodeGenOpts().getDebugInfo() < CodeGenOptions::LimitedDebugInfo)
+  if (CGM.getCodeGenOpts().getDebugInfo() < codegenoptions::LimitedDebugInfo)
     return;
   assert(UD.shadow_size() &&
          "We shouldn't be codegening an invalid UsingDecl containing no decls");
@@ -3451,6 +3645,8 @@ void CGDebugInfo::EmitUsingDecl(const UsingDecl &UD) {
 }
 
 void CGDebugInfo::EmitImportDecl(const ImportDecl &ID) {
+  if (CGM.getCodeGenOpts().getDebuggerTuning() != llvm::DebuggerKind::LLDB)
+    return;
   if (Module *M = ID.getImportedModule()) {
     auto Info = ExternalASTSource::ASTSourceDescriptor(*M);
     DBuilder.createImportedDeclaration(
@@ -3462,13 +3658,13 @@ void CGDebugInfo::EmitImportDecl(const ImportDecl &ID) {
 
 llvm::DIImportedEntity *
 CGDebugInfo::EmitNamespaceAlias(const NamespaceAliasDecl &NA) {
-  if (CGM.getCodeGenOpts().getDebugInfo() < CodeGenOptions::LimitedDebugInfo)
+  if (CGM.getCodeGenOpts().getDebugInfo() < codegenoptions::LimitedDebugInfo)
     return nullptr;
   auto &VH = NamespaceAliasCache[&NA];
   if (VH)
     return cast<llvm::DIImportedEntity>(VH);
   llvm::DIImportedEntity *R;
-  if (const NamespaceAliasDecl *Underlying =
+  if (const auto *Underlying =
           dyn_cast<NamespaceAliasDecl>(NA.getAliasedNamespace()))
     // This could cache & dedup here rather than relying on metadata deduping.
     R = DBuilder.createImportedDeclaration(
@@ -3557,7 +3753,7 @@ void CGDebugInfo::finalize() {
 }
 
 void CGDebugInfo::EmitExplicitCastType(QualType Ty) {
-  if (CGM.getCodeGenOpts().getDebugInfo() < CodeGenOptions::LimitedDebugInfo)
+  if (CGM.getCodeGenOpts().getDebugInfo() < codegenoptions::LimitedDebugInfo)
     return;
 
   if (auto *DieTy = getOrCreateType(Ty, getOrCreateMainFile()))
diff --git a/lib/CodeGen/CGDebugInfo.h b/lib/CodeGen/CGDebugInfo.h
index a68dd33fa5fe..366dd81ac812 100644
--- a/lib/CodeGen/CGDebugInfo.h
+++ b/lib/CodeGen/CGDebugInfo.h
@@ -16,6 +16,7 @@
 
 #include "CGBuilder.h"
 #include "clang/AST/Expr.h"
+#include "clang/AST/ExternalASTSource.h"
 #include "clang/AST/Type.h"
 #include "clang/Basic/SourceLocation.h"
 #include "clang/Frontend/CodeGenOptions.h"
@@ -52,28 +53,20 @@ class CGDebugInfo {
   friend class ApplyDebugLocation;
   friend class SaveAndRestoreLocation;
   CodeGenModule &CGM;
-  const CodeGenOptions::DebugInfoKind DebugKind;
+  const codegenoptions::DebugInfoKind DebugKind;
   bool DebugTypeExtRefs;
   llvm::DIBuilder DBuilder;
   llvm::DICompileUnit *TheCU = nullptr;
   ModuleMap *ClangModuleMap = nullptr;
+  ExternalASTSource::ASTSourceDescriptor PCHDescriptor;
   SourceLocation CurLoc;
   llvm::DIType *VTablePtrType = nullptr;
   llvm::DIType *ClassTy = nullptr;
   llvm::DICompositeType *ObjTy = nullptr;
   llvm::DIType *SelTy = nullptr;
-  llvm::DIType *OCLImage1dDITy = nullptr;
-  llvm::DIType *OCLImage1dArrayDITy = nullptr;
-  llvm::DIType *OCLImage1dBufferDITy = nullptr;
-  llvm::DIType *OCLImage2dDITy = nullptr;
-  llvm::DIType *OCLImage2dArrayDITy = nullptr;
-  llvm::DIType *OCLImage2dDepthDITy = nullptr;
-  llvm::DIType *OCLImage2dArrayDepthDITy = nullptr;
-  llvm::DIType *OCLImage2dMSAADITy = nullptr;
-  llvm::DIType *OCLImage2dArrayMSAADITy = nullptr;
-  llvm::DIType *OCLImage2dMSAADepthDITy = nullptr;
-  llvm::DIType *OCLImage2dArrayMSAADepthDITy = nullptr;
-  llvm::DIType *OCLImage3dDITy = nullptr;
+#define IMAGE_TYPE(ImgType, Id, SingletonId, Access, Suffix) \
+  llvm::DIType *SingletonId = nullptr;
+#include "clang/Basic/OpenCLImageTypes.def"
   llvm::DIType *OCLEventDITy = nullptr;
   llvm::DIType *OCLClkEventDITy = nullptr;
   llvm::DIType *OCLQueueDITy = nullptr;
@@ -107,7 +100,7 @@ class CGDebugInfo {
   /// compilation.
   std::vector<std::pair<const TagType *, llvm::TrackingMDRef>> ReplaceMap;
 
-  /// Cache of replaceable forward declarartions (functions and
+  /// Cache of replaceable forward declarations (functions and
   /// variables) to RAUW at the end of compilation.
   std::vector<std::pair<const DeclaratorDecl *, llvm::TrackingMDRef>>
       FwdDeclReplaceMap;
@@ -239,11 +232,16 @@ class CGDebugInfo {
                            llvm::DIFile *F);
 
   llvm::DIType *createFieldType(StringRef name, QualType type,
-                                uint64_t sizeInBitsOverride, SourceLocation loc,
-                                AccessSpecifier AS, uint64_t offsetInBits,
-                                llvm::DIFile *tunit, llvm::DIScope *scope,
+                                SourceLocation loc, AccessSpecifier AS,
+                                uint64_t offsetInBits, llvm::DIFile *tunit,
+                                llvm::DIScope *scope,
                                 const RecordDecl *RD = nullptr);
 
+  /// Create new bit field member.
+  llvm::DIType *createBitFieldType(const FieldDecl *BitFieldDecl,
+                                   llvm::DIScope *RecordTy,
+                                   const RecordDecl *RD);
+
   /// Helpers for collecting fields of a record.
   /// @{
   void CollectRecordLambdaFields(const CXXRecordDecl *CXXDecl,
@@ -275,6 +273,8 @@ public:
 
   void finalize();
 
+  /// Module debugging: Support for building PCMs.
+  /// @{
   /// Set the main CU's DwoId field to \p Signature.
   void setDwoId(uint64_t Signature);
 
@@ -283,6 +283,14 @@ public:
   /// the module of origin of each Decl.
   void setModuleMap(ModuleMap &MMap) { ClangModuleMap = &MMap; }
 
+  /// When generating debug information for a clang module or
+  /// precompiled header, this module map will be used to determine
+  /// the module of origin of each Decl.
+  void setPCHDescriptor(ExternalASTSource::ASTSourceDescriptor PCH) {
+    PCHDescriptor = PCH;
+  }
+  /// @}
+
   /// Update the current source location. If \arg loc is invalid it is
   /// ignored.
   void setLocation(SourceLocation Loc);
diff --git a/lib/CodeGen/CGDecl.cpp b/lib/CodeGen/CGDecl.cpp
index b78e80d79ddd..89407cd70c3d 100644
--- a/lib/CodeGen/CGDecl.cpp
+++ b/lib/CodeGen/CGDecl.cpp
@@ -16,11 +16,13 @@
 #include "CGCleanup.h"
 #include "CGDebugInfo.h"
 #include "CGOpenCLRuntime.h"
+#include "CGOpenMPRuntime.h"
 #include "CodeGenModule.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/AST/CharUnits.h"
 #include "clang/AST/Decl.h"
 #include "clang/AST/DeclObjC.h"
+#include "clang/AST/DeclOpenMP.h"
 #include "clang/Basic/SourceManager.h"
 #include "clang/Basic/TargetInfo.h"
 #include "clang/CodeGen/CGFunctionInfo.h"
@@ -29,10 +31,10 @@
 #include "llvm/IR/GlobalVariable.h"
 #include "llvm/IR/Intrinsics.h"
 #include "llvm/IR/Type.h"
+
 using namespace clang;
 using namespace CodeGen;
 
-
 void CodeGenFunction::EmitDecl(const Decl &D) {
   switch (D.getKind()) {
   case Decl::BuiltinTemplate:
@@ -71,6 +73,8 @@ void CodeGenFunction::EmitDecl(const Decl &D) {
   case Decl::ObjCImplementation:
   case Decl::ObjCProperty:
   case Decl::ObjCCompatibleAlias:
+  case Decl::PragmaComment:
+  case Decl::PragmaDetectMismatch:
   case Decl::AccessSpec:
   case Decl::LinkageSpec:
   case Decl::ObjCPropertyImpl:
@@ -81,6 +85,7 @@ void CodeGenFunction::EmitDecl(const Decl &D) {
   case Decl::Captured:
   case Decl::ClassScopeFunctionSpecialization:
   case Decl::UsingShadow:
+  case Decl::ConstructorUsingShadow:
   case Decl::ObjCTypeParam:
     llvm_unreachable("Declaration should not be in declstmts!");
   case Decl::Function:  // void X();
@@ -92,6 +97,7 @@ void CodeGenFunction::EmitDecl(const Decl &D) {
   case Decl::Label:        // __label__ x;
   case Decl::Import:
   case Decl::OMPThreadPrivate:
+  case Decl::OMPCapturedExpr:
   case Decl::Empty:
     // None of these decls require codegen support.
     return;
@@ -115,6 +121,9 @@ void CodeGenFunction::EmitDecl(const Decl &D) {
     return EmitVarDecl(VD);
   }
 
+  case Decl::OMPDeclareReduction:
+    return CGM.EmitOMPDeclareReduction(cast<OMPDeclareReductionDecl>(&D), this);
+
   case Decl::Typedef:      // typedef int X;
   case Decl::TypeAlias: {  // using X = int; [C++0x]
     const TypedefNameDecl &TD = cast<TypedefNameDecl>(D);
@@ -363,8 +372,15 @@ void CodeGenFunction::EmitStaticVarDecl(const VarDecl &D,
 
   llvm::GlobalVariable *var =
     cast<llvm::GlobalVariable>(addr->stripPointerCasts());
+
+  // CUDA's local and local static __shared__ variables should not
+  // have any non-empty initializers. This is ensured by Sema.
+  // Whatever initializer such variable may have when it gets here is
+  // a no-op and should not be emitted.
+  bool isCudaSharedVar = getLangOpts().CUDA && getLangOpts().CUDAIsDevice &&
+                         D.hasAttr<CUDASharedAttr>();
   // If this value has an initializer, emit it.
-  if (D.getInit())
+  if (D.getInit() && !isCudaSharedVar)
     var = AddInitializerToStaticVarDecl(D, var);
 
   var->setAlignment(alignment.getQuantity());
@@ -394,7 +410,7 @@ void CodeGenFunction::EmitStaticVarDecl(const VarDecl &D,
   // Emit global variable debug descriptor for static vars.
   CGDebugInfo *DI = getDebugInfo();
   if (DI &&
-      CGM.getCodeGenOpts().getDebugInfo() >= CodeGenOptions::LimitedDebugInfo) {
+      CGM.getCodeGenOpts().getDebugInfo() >= codegenoptions::LimitedDebugInfo) {
     DI->setLocation(D.getLocation());
     DI->EmitGlobalVariable(var, &D);
   }
@@ -513,20 +529,7 @@ namespace {
       CGF.EmitCall(FnInfo, CleanupFn, ReturnValueSlot(), Args);
     }
   };
-
-  /// A cleanup to call @llvm.lifetime.end.
-  class CallLifetimeEnd final : public EHScopeStack::Cleanup {
-    llvm::Value *Addr;
-    llvm::Value *Size;
-  public:
-    CallLifetimeEnd(Address addr, llvm::Value *size)
-      : Addr(addr.getPointer()), Size(size) {}
-
-    void Emit(CodeGenFunction &CGF, Flags flags) override {
-      CGF.EmitLifetimeEnd(Size, Addr);
-    }
-  };
-}
+} // end anonymous namespace
 
 /// EmitAutoVarWithLifetime - Does the setup required for an automatic
 /// variable with lifetime.
@@ -644,7 +647,6 @@ static bool tryEmitARCCopyWeakInit(CodeGenFunction &CGF,
     }
 
     init = castExpr->getSubExpr();
-    continue;
   }
   return false;
 }
@@ -665,10 +667,10 @@ void CodeGenFunction::EmitScalarInit(const Expr *init, const ValueDecl *D,
     EmitStoreThroughLValue(RValue::get(value), lvalue, true);
     return;
   }
-  
+
   if (const CXXDefaultInitExpr *DIE = dyn_cast<CXXDefaultInitExpr>(init))
     init = DIE->getExpr();
-    
+
   // If we're emitting a value with lifetime, we have to do the
   // initialization *before* we leave the cleanup scopes.
   if (const ExprWithCleanups *ewc = dyn_cast<ExprWithCleanups>(init)) {
@@ -715,8 +717,7 @@ void CodeGenFunction::EmitScalarInit(const Expr *init, const ValueDecl *D,
     llvm_unreachable("present but none");
 
   case Qualifiers::OCL_ExplicitNone:
-    // nothing to do
-    value = EmitScalarExpr(init);
+    value = EmitARCUnsafeUnretainedScalarExpr(init);
     break;
 
   case Qualifiers::OCL_Strong: {
@@ -819,7 +820,7 @@ static bool canEmitInitWithFewStoresAfterMemset(llvm::Constant *Init,
     }
     return true;
   }
-  
+
   if (llvm::ConstantDataSequential *CDS =
         dyn_cast<llvm::ConstantDataSequential>(Init)) {
     for (unsigned i = 0, e = CDS->getNumElements(); i != e; ++i) {
@@ -848,9 +849,9 @@ static void emitStoresForInitAfterMemset(llvm::Constant *Init, llvm::Value *Loc,
     Builder.CreateDefaultAlignedStore(Init, Loc, isVolatile);
     return;
   }
-  
-  if (llvm::ConstantDataSequential *CDS = 
-        dyn_cast<llvm::ConstantDataSequential>(Init)) {
+
+  if (llvm::ConstantDataSequential *CDS =
+          dyn_cast<llvm::ConstantDataSequential>(Init)) {
     for (unsigned i = 0, e = CDS->getNumElements(); i != e; ++i) {
       llvm::Constant *Elt = CDS->getElementAsConstant(i);
 
@@ -877,7 +878,6 @@ static void emitStoresForInitAfterMemset(llvm::Constant *Init, llvm::Value *Loc,
   }
 }
 
-
 /// shouldUseMemSetPlusStoresToInitialize - Decide whether we should use memset
 /// plus some stores to initialize a local variable instead of using a memcpy
 /// from a constant global.  It is beneficial to use memset if the global is all
@@ -907,18 +907,29 @@ void CodeGenFunction::EmitAutoVarDecl(const VarDecl &D) {
   EmitAutoVarCleanups(emission);
 }
 
+/// shouldEmitLifetimeMarkers - Decide whether we need emit the life-time
+/// markers.
+static bool shouldEmitLifetimeMarkers(const CodeGenOptions &CGOpts,
+                                      const LangOptions &LangOpts) {
+  // Asan uses markers for use-after-scope checks.
+  if (CGOpts.SanitizeAddressUseAfterScope)
+    return true;
+
+  // Disable lifetime markers in msan builds.
+  // FIXME: Remove this when msan works with lifetime markers.
+  if (LangOpts.Sanitize.has(SanitizerKind::Memory))
+    return false;
+
+  // For now, only in optimized builds.
+  return CGOpts.OptimizationLevel != 0;
+}
+
 /// Emit a lifetime.begin marker if some criteria are satisfied.
 /// \return a pointer to the temporary size Value if a marker was emitted, null
 /// otherwise
 llvm::Value *CodeGenFunction::EmitLifetimeStart(uint64_t Size,
                                                 llvm::Value *Addr) {
-  // For now, only in optimized builds.
-  if (CGM.getCodeGenOpts().OptimizationLevel == 0)
-    return nullptr;
-
-  // Disable lifetime markers in msan builds.
-  // FIXME: Remove this when msan works with lifetime markers.
-  if (getLangOpts().Sanitize.has(SanitizerKind::Memory))
+  if (!shouldEmitLifetimeMarkers(CGM.getCodeGenOpts(), getLangOpts()))
     return nullptr;
 
   llvm::Value *SizeV = llvm::ConstantInt::get(Int64Ty, Size);
@@ -1086,8 +1097,8 @@ CodeGenFunction::EmitAutoVarAlloca(const VarDecl &D) {
   // Emit debug info for local var declaration.
   if (HaveInsertPoint())
     if (CGDebugInfo *DI = getDebugInfo()) {
-      if (CGM.getCodeGenOpts().getDebugInfo()
-            >= CodeGenOptions::LimitedDebugInfo) {
+      if (CGM.getCodeGenOpts().getDebugInfo() >=
+          codegenoptions::LimitedDebugInfo) {
         DI->setLocation(D.getLocation());
         DI->EmitDeclareOfAutoVariable(&D, address.getPointer(), Builder);
       }
@@ -1163,6 +1174,7 @@ bool CodeGenFunction::isTrivialInitializer(const Expr *Init) {
 
   return false;
 }
+
 void CodeGenFunction::EmitAutoVarInit(const AutoVarEmission &emission) {
   assert(emission.Variable && "emission was not valid!");
 
@@ -1250,7 +1262,7 @@ void CodeGenFunction::EmitAutoVarInit(const AutoVarEmission &emission) {
                                llvm::GlobalValue::PrivateLinkage,
                                constant, Name);
     GV->setAlignment(Loc.getAlignment().getQuantity());
-    GV->setUnnamedAddr(true);
+    GV->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
 
     Address SrcPtr = Address(GV, Loc.getAlignment());
     if (SrcPtr.getType() != BP)
@@ -1381,13 +1393,10 @@ void CodeGenFunction::EmitAutoVarCleanups(const AutoVarEmission &emission) {
 
   // Make sure we call @llvm.lifetime.end.  This needs to happen
   // *last*, so the cleanup needs to be pushed *first*.
-  if (emission.useLifetimeMarkers()) {
-    EHStack.pushCleanup<CallLifetimeEnd>(NormalCleanup,
+  if (emission.useLifetimeMarkers())
+    EHStack.pushCleanup<CallLifetimeEnd>(NormalEHLifetimeMarker,
                                          emission.getAllocatedAddress(),
                                          emission.getSizeForLifetimeMarkers());
-    EHCleanupScope &cleanup = cast<EHCleanupScope>(*EHStack.begin());
-    cleanup.setLifetimeMarker();
-  }
 
   // Check the type for a cleanup.
   if (QualType::DestructionKind dtorKind = D.getType().isDestructedType())
@@ -1662,7 +1671,7 @@ namespace {
                               ElementType, ElementAlign, Destroyer);
     }
   };
-}
+} // end anonymous namespace
 
 /// pushIrregularPartialArrayCleanup - Push an EH cleanup to destroy
 /// already-constructed elements of the given array.  The cleanup
@@ -1731,7 +1740,7 @@ namespace {
       CGF.EmitARCRelease(Param, Precise);
     }
   };
-}
+} // end anonymous namespace
 
 /// Emit an alloca (or GlobalValue depending on target)
 /// for the specified parameter and set up LocalDeclMap.
@@ -1852,8 +1861,8 @@ void CodeGenFunction::EmitParmDecl(const VarDecl &D, ParamValue Arg,
 
   // Emit debug info for param declaration.
   if (CGDebugInfo *DI = getDebugInfo()) {
-    if (CGM.getCodeGenOpts().getDebugInfo()
-          >= CodeGenOptions::LimitedDebugInfo) {
+    if (CGM.getCodeGenOpts().getDebugInfo() >=
+        codegenoptions::LimitedDebugInfo) {
       DI->EmitDeclareOfArgVariable(&D, DeclPtr.getPointer(), ArgNo, Builder);
     }
   }
@@ -1861,3 +1870,10 @@ void CodeGenFunction::EmitParmDecl(const VarDecl &D, ParamValue Arg,
   if (D.hasAttr<AnnotateAttr>())
     EmitVarAnnotations(&D, DeclPtr.getPointer());
 }
+
+void CodeGenModule::EmitOMPDeclareReduction(const OMPDeclareReductionDecl *D,
+                                            CodeGenFunction *CGF) {
+  if (!LangOpts.OpenMP || (!LangOpts.EmitAllDecls && !D->isUsed()))
+    return;
+  getOpenMPRuntime().emitUserDefinedReduction(CGF, D);
+}
diff --git a/lib/CodeGen/CGDeclCXX.cpp b/lib/CodeGen/CGDeclCXX.cpp
index adba73168797..89d142e44b49 100644
--- a/lib/CodeGen/CGDeclCXX.cpp
+++ b/lib/CodeGen/CGDeclCXX.cpp
@@ -86,13 +86,21 @@ static void EmitDeclDestroy(CodeGenFunction &CGF, const VarDecl &D,
   llvm::Constant *function;
   llvm::Constant *argument;
 
-  // Special-case non-array C++ destructors, where there's a function
-  // with the right signature that we can just call.
-  const CXXRecordDecl *record = nullptr;
-  if (dtorKind == QualType::DK_cxx_destructor &&
-      (record = type->getAsCXXRecordDecl())) {
-    assert(!record->hasTrivialDestructor());
-    CXXDestructorDecl *dtor = record->getDestructor();
+  // Special-case non-array C++ destructors, if they have the right signature.
+  // Under some ABIs, destructors return this instead of void, and cannot be
+  // passed directly to __cxa_atexit if the target does not allow this mismatch.
+  const CXXRecordDecl *Record = type->getAsCXXRecordDecl();
+  bool CanRegisterDestructor =
+      Record && (!CGM.getCXXABI().HasThisReturn(
+                     GlobalDecl(Record->getDestructor(), Dtor_Complete)) ||
+                 CGM.getCXXABI().canCallMismatchedFunctionType());
+  // If __cxa_atexit is disabled via a flag, a different helper function is
+  // generated elsewhere which uses atexit instead, and it takes the destructor
+  // directly.
+  bool UsingExternalHelper = !CGM.getCodeGenOpts().CXAAtExit;
+  if (Record && (CanRegisterDestructor || UsingExternalHelper)) {
+    assert(!Record->hasTrivialDestructor());
+    CXXDestructorDecl *dtor = Record->getDestructor();
 
     function = CGM.getAddrOfCXXStructor(dtor, StructorType::Complete);
     argument = llvm::ConstantExpr::getBitCast(
@@ -304,6 +312,17 @@ void
 CodeGenModule::EmitCXXGlobalVarDeclInitFunc(const VarDecl *D,
                                             llvm::GlobalVariable *Addr,
                                             bool PerformInit) {
+
+  // According to E.2.3.1 in CUDA-7.5 Programming guide: __device__,
+  // __constant__ and __shared__ variables defined in namespace scope,
+  // that are of class type, cannot have a non-empty constructor. All
+  // the checks have been done in Sema by now. Whatever initializers
+  // are allowed are empty and we just need to ignore them here.
+  if (getLangOpts().CUDA && getLangOpts().CUDAIsDevice &&
+      (D->hasAttr<CUDADeviceAttr>() || D->hasAttr<CUDAConstantAttr>() ||
+       D->hasAttr<CUDASharedAttr>()))
+    return;
+
   // Check if we've already initialized this decl.
   auto I = DelayedCXXInitPosition.find(D);
   if (I != DelayedCXXInitPosition.end() && I->second == ~0U)
@@ -587,8 +606,8 @@ llvm::Function *CodeGenFunction::generateDestroyHelper(
                         getContext().VoidPtrTy);
   args.push_back(&dst);
 
-  const CGFunctionInfo &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      getContext().VoidTy, args, FunctionType::ExtInfo(), /*variadic=*/false);
+  const CGFunctionInfo &FI =
+    CGM.getTypes().arrangeBuiltinFunctionDeclaration(getContext().VoidTy, args);
   llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
   llvm::Function *fn = CGM.CreateGlobalInitOrDestructFunction(
       FTy, "__cxx_global_array_dtor", FI, VD->getLocation());
diff --git a/lib/CodeGen/CGException.cpp b/lib/CodeGen/CGException.cpp
index fce2e7581962..4a7dc4205e09 100644
--- a/lib/CodeGen/CGException.cpp
+++ b/lib/CodeGen/CGException.cpp
@@ -686,8 +686,10 @@ llvm::BasicBlock *CodeGenFunction::getInvokeDestImpl() {
   assert(EHStack.requiresLandingPad());
   assert(!EHStack.empty());
 
-  // If exceptions are disabled, there are usually no landingpads. However, when
-  // SEH is enabled, functions using SEH still get landingpads.
+  // If exceptions are disabled and SEH is not in use, then there is no invoke
+  // destination. SEH "works" even if exceptions are off. In practice, this
+  // means that C++ destructors and other EH cleanups don't run, which is
+  // consistent with MSVC's behavior.
   const LangOptions &LO = CGM.getLangOpts();
   if (!LO.Exceptions) {
     if (!LO.Borland && !LO.MicrosoftExt)
@@ -1326,11 +1328,13 @@ llvm::BasicBlock *CodeGenFunction::getTerminateHandler() {
   TerminateHandler = createBasicBlock("terminate.handler");
   Builder.SetInsertPoint(TerminateHandler);
   llvm::Value *Exn = nullptr;
+  SaveAndRestore<llvm::Instruction *> RestoreCurrentFuncletPad(
+      CurrentFuncletPad);
   if (EHPersonality::get(*this).usesFuncletPads()) {
     llvm::Value *ParentPad = CurrentFuncletPad;
     if (!ParentPad)
       ParentPad = llvm::ConstantTokenNone::get(CGM.getLLVMContext());
-    Builder.CreateCleanupPad(ParentPad);
+    CurrentFuncletPad = Builder.CreateCleanupPad(ParentPad);
   } else {
     if (getLangOpts().CPlusPlus)
       Exn = getExceptionFromSlot();
@@ -1422,12 +1426,8 @@ struct PerformSEHFinally final : EHScopeStack::Cleanup {
     Args.add(RValue::get(FP), ArgTys[1]);
 
     // Arrange a two-arg function info and type.
-    FunctionProtoType::ExtProtoInfo EPI;
-    const auto *FPT = cast<FunctionProtoType>(
-        Context.getFunctionType(Context.VoidTy, ArgTys, EPI));
     const CGFunctionInfo &FnInfo =
-        CGM.getTypes().arrangeFreeFunctionCall(Args, FPT,
-                                               /*chainCall=*/false);
+        CGM.getTypes().arrangeBuiltinFunctionCall(Context.VoidTy, Args);
 
     CGF.EmitCall(FnInfo, OutlinedFinally, ReturnValueSlot(), Args);
   }
@@ -1623,14 +1623,13 @@ void CodeGenFunction::startOutlinedSEHHelper(CodeGenFunction &ParentCGF,
   SmallString<128> Name;
   {
     llvm::raw_svector_ostream OS(Name);
-    const Decl *ParentCodeDecl = ParentCGF.CurCodeDecl;
-    const NamedDecl *Parent = dyn_cast_or_null<NamedDecl>(ParentCodeDecl);
-    assert(Parent && "FIXME: handle unnamed decls (lambdas, blocks) with SEH");
+    const FunctionDecl *ParentSEHFn = ParentCGF.CurSEHParent;
+    assert(ParentSEHFn && "No CurSEHParent!");
     MangleContext &Mangler = CGM.getCXXABI().getMangleContext();
     if (IsFilter)
-      Mangler.mangleSEHFilterExpression(Parent, OS);
+      Mangler.mangleSEHFilterExpression(ParentSEHFn, OS);
     else
-      Mangler.mangleSEHFinallyBlock(Parent, OS);
+      Mangler.mangleSEHFinallyBlock(ParentSEHFn, OS);
   }
 
   FunctionArgList Args;
@@ -1656,8 +1655,8 @@ void CodeGenFunction::startOutlinedSEHHelper(CodeGenFunction &ParentCGF,
   QualType RetTy = IsFilter ? getContext().LongTy : getContext().VoidTy;
 
   llvm::Function *ParentFn = ParentCGF.CurFn;
-  const CGFunctionInfo &FnInfo = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      RetTy, Args, FunctionType::ExtInfo(), /*isVariadic=*/false);
+  const CGFunctionInfo &FnInfo =
+    CGM.getTypes().arrangeBuiltinFunctionDeclaration(RetTy, Args);
 
   llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
   llvm::Function *Fn = llvm::Function::Create(
@@ -1677,6 +1676,7 @@ void CodeGenFunction::startOutlinedSEHHelper(CodeGenFunction &ParentCGF,
 
   StartFunction(GlobalDecl(), RetTy, Fn, FnInfo, Args,
                 OutlinedStmt->getLocStart(), OutlinedStmt->getLocStart());
+  CurSEHParent = ParentCGF.CurSEHParent;
 
   CGM.SetLLVMFunctionAttributes(nullptr, FnInfo, CurFn);
   EmitCapturedLocals(ParentCGF, OutlinedStmt, IsFilter);
@@ -1708,12 +1708,6 @@ CodeGenFunction::GenerateSEHFinallyFunction(CodeGenFunction &ParentCGF,
   const Stmt *FinallyBlock = Finally.getBlock();
   startOutlinedSEHHelper(ParentCGF, false, FinallyBlock);
 
-  // Mark finally block calls as nounwind and noinline to make LLVM's job a
-  // little easier.
-  // FIXME: Remove these restrictions in the future.
-  CurFn->addFnAttr(llvm::Attribute::NoUnwind);
-  CurFn->addFnAttr(llvm::Attribute::NoInline);
-
   // Emit the original filter expression, convert to i32, and return.
   EmitStmt(FinallyBlock);
 
diff --git a/lib/CodeGen/CGExpr.cpp b/lib/CodeGen/CGExpr.cpp
index 507ce3d7d0ce..3e1ae3604f94 100644
--- a/lib/CodeGen/CGExpr.cpp
+++ b/lib/CodeGen/CGExpr.cpp
@@ -11,13 +11,14 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenFunction.h"
 #include "CGCXXABI.h"
 #include "CGCall.h"
+#include "CGCleanup.h"
 #include "CGDebugInfo.h"
 #include "CGObjCRuntime.h"
 #include "CGOpenMPRuntime.h"
 #include "CGRecordLayout.h"
+#include "CodeGenFunction.h"
 #include "CodeGenModule.h"
 #include "TargetInfo.h"
 #include "clang/AST/ASTContext.h"
@@ -32,6 +33,8 @@
 #include "llvm/IR/MDBuilder.h"
 #include "llvm/Support/ConvertUTF.h"
 #include "llvm/Support/MathExtras.h"
+#include "llvm/Support/Path.h"
+#include "llvm/Transforms/Utils/SanitizerStats.h"
 
 using namespace clang;
 using namespace CodeGen;
@@ -65,8 +68,6 @@ Address CodeGenFunction::CreateTempAlloca(llvm::Type *Ty, CharUnits Align,
 /// block.
 llvm::AllocaInst *CodeGenFunction::CreateTempAlloca(llvm::Type *Ty,
                                                     const Twine &Name) {
-  if (!Builder.isNamePreserving())
-    return new llvm::AllocaInst(Ty, nullptr, "", AllocaInsertPt);
   return new llvm::AllocaInst(Ty, nullptr, Name, AllocaInsertPt);
 }
 
@@ -361,9 +362,16 @@ EmitMaterializeTemporaryExpr(const MaterializeTemporaryExpr *M) {
                            ConvertTypeForMem(E->getType())
                              ->getPointerTo(Object.getAddressSpace())),
                        Object.getAlignment());
-      // We should not have emitted the initializer for this temporary as a
-      // constant.
-      assert(!Var->hasInitializer());
+
+      // createReferenceTemporary will promote the temporary to a global with a
+      // constant initializer if it can.  It can only do this to a value of
+      // ARC-manageable type if the value is global and therefore "immune" to
+      // ref-counting operations.  Therefore we have no need to emit either a
+      // dynamic initialization or a cleanup and we can just return the address
+      // of the temporary.
+      if (Var->hasInitializer())
+        return MakeAddrLValue(Object, M->getType(), AlignmentSource::Decl);
+
       Var->setInitializer(CGM.EmitNullConstant(E->getType()));
     }
     LValue RefTempDst = MakeAddrLValue(Object, M->getType(),
@@ -416,6 +424,23 @@ EmitMaterializeTemporaryExpr(const MaterializeTemporaryExpr *M) {
       EmitAnyExprToMem(E, Object, Qualifiers(), /*IsInit*/true);
     }
   } else {
+    switch (M->getStorageDuration()) {
+    case SD_Automatic:
+    case SD_FullExpression:
+      if (auto *Size = EmitLifetimeStart(
+              CGM.getDataLayout().getTypeAllocSize(Object.getElementType()),
+              Object.getPointer())) {
+        if (M->getStorageDuration() == SD_Automatic)
+          pushCleanupAfterFullExpr<CallLifetimeEnd>(NormalEHLifetimeMarker,
+                                                    Object, Size);
+        else
+          pushFullExprCleanup<CallLifetimeEnd>(NormalEHLifetimeMarker, Object,
+                                               Size);
+      }
+      break;
+    default:
+      break;
+    }
     EmitAnyExprToMem(E, Object, Qualifiers(), /*IsInit*/true);
   }
   pushTemporaryCleanup(*this, M, E, Object);
@@ -577,7 +602,7 @@ void CodeGenFunction::EmitTypeCheck(TypeCheckKind TCK, SourceLocation Loc,
 
   if (Checks.size() > 0) {
     llvm::Constant *StaticData[] = {
-      EmitCheckSourceLocation(Loc),
+     EmitCheckSourceLocation(Loc),
       EmitCheckTypeDescriptor(Ty),
       llvm::ConstantInt::get(SizeTy, AlignVal),
       llvm::ConstantInt::get(Int8Ty, TCK)
@@ -824,7 +849,8 @@ Address CodeGenFunction::EmitPointerWithAlignment(const Expr *E,
                          getNaturalPointeeTypeAlignment(E->getType(), Source));
         }
 
-        if (SanOpts.has(SanitizerKind::CFIUnrelatedCast)) {
+        if (SanOpts.has(SanitizerKind::CFIUnrelatedCast) &&
+            CE->getCastKind() == CK_BitCast) {
           if (auto PT = E->getType()->getAs<PointerType>())
             EmitVTablePtrCheckForCast(PT->getPointeeType(), Addr.getPointer(),
                                       /*MayBeNull=*/true,
@@ -1265,10 +1291,10 @@ llvm::Value *CodeGenFunction::EmitLoadOfScalar(Address Addr, bool Volatile,
   }
 
   // Atomic operations have to be done on integral types.
-  if (Ty->isAtomicType() || typeIsSuitableForInlineAtomic(Ty, Volatile)) {
-    LValue lvalue =
+  LValue AtomicLValue =
       LValue::MakeAddr(Addr, Ty, getContext(), AlignSource, TBAAInfo);
-    return EmitAtomicLoad(lvalue, Loc).getScalarVal();
+  if (Ty->isAtomicType() || LValueIsSuitableForInlineAtomic(AtomicLValue)) {
+    return EmitAtomicLoad(AtomicLValue, Loc).getScalarVal();
   }
 
   llvm::LoadInst *Load = Builder.CreateLoad(Addr, Volatile);
@@ -1376,12 +1402,11 @@ void CodeGenFunction::EmitStoreOfScalar(llvm::Value *Value, Address Addr,
 
   Value = EmitToMemory(Value, Ty);
 
+  LValue AtomicLValue =
+      LValue::MakeAddr(Addr, Ty, getContext(), AlignSource, TBAAInfo);
   if (Ty->isAtomicType() ||
-      (!isInit && typeIsSuitableForInlineAtomic(Ty, Volatile))) {
-    EmitAtomicStore(RValue::get(Value),
-                    LValue::MakeAddr(Addr, Ty, getContext(),
-                                     AlignSource, TBAAInfo),
-                    isInit);
+      (!isInit && LValueIsSuitableForInlineAtomic(AtomicLValue))) {
+    EmitAtomicStore(RValue::get(Value), AtomicLValue, isInit);
     return;
   }
 
@@ -1733,8 +1758,7 @@ void CodeGenFunction::EmitStoreThroughExtVectorComponentLValue(RValue Src,
 
   if (const VectorType *VTy = Dst.getType()->getAs<VectorType>()) {
     unsigned NumSrcElts = VTy->getNumElements();
-    unsigned NumDstElts =
-       cast<llvm::VectorType>(Vec->getType())->getNumElements();
+    unsigned NumDstElts = Vec->getType()->getVectorNumElements();
     if (NumDstElts == NumSrcElts) {
       // Use shuffle vector is the src and destination are the same number of
       // elements and restore the vector mask since it is on the side it will be
@@ -1947,6 +1971,21 @@ LValue CodeGenFunction::EmitLoadOfReferenceLValue(Address RefAddr,
   return MakeAddrLValue(Addr, RefTy->getPointeeType(), Source);
 }
 
+Address CodeGenFunction::EmitLoadOfPointer(Address Ptr,
+                                           const PointerType *PtrTy,
+                                           AlignmentSource *Source) {
+  llvm::Value *Addr = Builder.CreateLoad(Ptr);
+  return Address(Addr, getNaturalTypeAlignment(PtrTy->getPointeeType(), Source,
+                                               /*forPointeeType=*/true));
+}
+
+LValue CodeGenFunction::EmitLoadOfPointerLValue(Address PtrAddr,
+                                                const PointerType *PtrTy) {
+  AlignmentSource Source;
+  Address Addr = EmitLoadOfPointer(PtrAddr, PtrTy, &Source);
+  return MakeAddrLValue(Addr, PtrTy->getPointeeType(), Source);
+}
+
 static LValue EmitGlobalVarDeclLValue(CodeGenFunction &CGF,
                                       const Expr *E, const VarDecl *VD) {
   QualType T = E->getType();
@@ -2302,7 +2341,7 @@ llvm::Constant *CodeGenFunction::EmitCheckTypeDescriptor(QualType T) {
   auto *GV = new llvm::GlobalVariable(
       CGM.getModule(), Descriptor->getType(),
       /*isConstant=*/true, llvm::GlobalVariable::PrivateLinkage, Descriptor);
-  GV->setUnnamedAddr(true);
+  GV->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
   CGM.getSanitizerMetadata()->disableSanitizerForGlobal(GV);
 
   // Remember the descriptor for this type.
@@ -2352,7 +2391,33 @@ llvm::Constant *CodeGenFunction::EmitCheckSourceLocation(SourceLocation Loc) {
 
   PresumedLoc PLoc = getContext().getSourceManager().getPresumedLoc(Loc);
   if (PLoc.isValid()) {
-    auto FilenameGV = CGM.GetAddrOfConstantCString(PLoc.getFilename(), ".src");
+    StringRef FilenameString = PLoc.getFilename();
+
+    int PathComponentsToStrip =
+        CGM.getCodeGenOpts().EmitCheckPathComponentsToStrip;
+    if (PathComponentsToStrip < 0) {
+      assert(PathComponentsToStrip != INT_MIN);
+      int PathComponentsToKeep = -PathComponentsToStrip;
+      auto I = llvm::sys::path::rbegin(FilenameString);
+      auto E = llvm::sys::path::rend(FilenameString);
+      while (I != E && --PathComponentsToKeep)
+        ++I;
+
+      FilenameString = FilenameString.substr(I - E);
+    } else if (PathComponentsToStrip > 0) {
+      auto I = llvm::sys::path::begin(FilenameString);
+      auto E = llvm::sys::path::end(FilenameString);
+      while (I != E && PathComponentsToStrip--)
+        ++I;
+
+      if (I != E)
+        FilenameString =
+            FilenameString.substr(I - llvm::sys::path::begin(FilenameString));
+      else
+        FilenameString = llvm::sys::path::filename(FilenameString);
+    }
+
+    auto FilenameGV = CGM.GetAddrOfConstantCString(FilenameString, ".src");
     CGM.getSanitizerMetadata()->disableSanitizerForGlobal(
                           cast<llvm::GlobalVariable>(FilenameGV.getPointer()));
     Filename = FilenameGV.getPointer();
@@ -2483,24 +2548,26 @@ void CodeGenFunction::EmitCheck(
   Branch->setMetadata(llvm::LLVMContext::MD_prof, Node);
   EmitBlock(Handlers);
 
-  // Emit handler arguments and create handler function type.
-  llvm::Constant *Info = llvm::ConstantStruct::getAnon(StaticArgs);
-  auto *InfoPtr =
-      new llvm::GlobalVariable(CGM.getModule(), Info->getType(), false,
-                               llvm::GlobalVariable::PrivateLinkage, Info);
-  InfoPtr->setUnnamedAddr(true);
-  CGM.getSanitizerMetadata()->disableSanitizerForGlobal(InfoPtr);
-
+  // Handler functions take an i8* pointing to the (handler-specific) static
+  // information block, followed by a sequence of intptr_t arguments
+  // representing operand values.
   SmallVector<llvm::Value *, 4> Args;
   SmallVector<llvm::Type *, 4> ArgTypes;
   Args.reserve(DynamicArgs.size() + 1);
   ArgTypes.reserve(DynamicArgs.size() + 1);
 
-  // Handler functions take an i8* pointing to the (handler-specific) static
-  // information block, followed by a sequence of intptr_t arguments
-  // representing operand values.
-  Args.push_back(Builder.CreateBitCast(InfoPtr, Int8PtrTy));
-  ArgTypes.push_back(Int8PtrTy);
+  // Emit handler arguments and create handler function type.
+  if (!StaticArgs.empty()) {
+    llvm::Constant *Info = llvm::ConstantStruct::getAnon(StaticArgs);
+    auto *InfoPtr =
+        new llvm::GlobalVariable(CGM.getModule(), Info->getType(), false,
+                                 llvm::GlobalVariable::PrivateLinkage, Info);
+    InfoPtr->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
+    CGM.getSanitizerMetadata()->disableSanitizerForGlobal(InfoPtr);
+    Args.push_back(Builder.CreateBitCast(InfoPtr, Int8PtrTy));
+    ArgTypes.push_back(Int8PtrTy);
+  }
+
   for (size_t i = 0, n = DynamicArgs.size(); i != n; ++i) {
     Args.push_back(EmitCheckValue(DynamicArgs[i]));
     ArgTypes.push_back(IntPtrTy);
@@ -2532,10 +2599,9 @@ void CodeGenFunction::EmitCheck(
   EmitBlock(Cont);
 }
 
-void CodeGenFunction::EmitCfiSlowPathCheck(llvm::Value *Cond,
-                                           llvm::ConstantInt *TypeId,
-                                           llvm::Value *Ptr) {
-  auto &Ctx = getLLVMContext();
+void CodeGenFunction::EmitCfiSlowPathCheck(
+    SanitizerMask Kind, llvm::Value *Cond, llvm::ConstantInt *TypeId,
+    llvm::Value *Ptr, ArrayRef<llvm::Constant *> StaticArgs) {
   llvm::BasicBlock *Cont = createBasicBlock("cfi.cont");
 
   llvm::BasicBlock *CheckBB = createBasicBlock("cfi.slowpath");
@@ -2547,19 +2613,122 @@ void CodeGenFunction::EmitCfiSlowPathCheck(llvm::Value *Cond,
 
   EmitBlock(CheckBB);
 
-  llvm::Constant *SlowPathFn = CGM.getModule().getOrInsertFunction(
-      "__cfi_slowpath",
-      llvm::FunctionType::get(
-          llvm::Type::getVoidTy(Ctx),
-          {llvm::Type::getInt64Ty(Ctx),
-           llvm::PointerType::getUnqual(llvm::Type::getInt8Ty(Ctx))},
-          false));
-  llvm::CallInst *CheckCall = Builder.CreateCall(SlowPathFn, {TypeId, Ptr});
+  bool WithDiag = !CGM.getCodeGenOpts().SanitizeTrap.has(Kind);
+
+  llvm::CallInst *CheckCall;
+  if (WithDiag) {
+    llvm::Constant *Info = llvm::ConstantStruct::getAnon(StaticArgs);
+    auto *InfoPtr =
+        new llvm::GlobalVariable(CGM.getModule(), Info->getType(), false,
+                                 llvm::GlobalVariable::PrivateLinkage, Info);
+    InfoPtr->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
+    CGM.getSanitizerMetadata()->disableSanitizerForGlobal(InfoPtr);
+
+    llvm::Constant *SlowPathDiagFn = CGM.getModule().getOrInsertFunction(
+        "__cfi_slowpath_diag",
+        llvm::FunctionType::get(VoidTy, {Int64Ty, Int8PtrTy, Int8PtrTy},
+                                false));
+    CheckCall = Builder.CreateCall(
+        SlowPathDiagFn,
+        {TypeId, Ptr, Builder.CreateBitCast(InfoPtr, Int8PtrTy)});
+  } else {
+    llvm::Constant *SlowPathFn = CGM.getModule().getOrInsertFunction(
+        "__cfi_slowpath",
+        llvm::FunctionType::get(VoidTy, {Int64Ty, Int8PtrTy}, false));
+    CheckCall = Builder.CreateCall(SlowPathFn, {TypeId, Ptr});
+  }
+
   CheckCall->setDoesNotThrow();
 
   EmitBlock(Cont);
 }
 
+// This function is basically a switch over the CFI failure kind, which is
+// extracted from CFICheckFailData (1st function argument). Each case is either
+// llvm.trap or a call to one of the two runtime handlers, based on
+// -fsanitize-trap and -fsanitize-recover settings.  Default case (invalid
+// failure kind) traps, but this should really never happen.  CFICheckFailData
+// can be nullptr if the calling module has -fsanitize-trap behavior for this
+// check kind; in this case __cfi_check_fail traps as well.
+void CodeGenFunction::EmitCfiCheckFail() {
+  SanitizerScope SanScope(this);
+  FunctionArgList Args;
+  ImplicitParamDecl ArgData(getContext(), nullptr, SourceLocation(), nullptr,
+                            getContext().VoidPtrTy);
+  ImplicitParamDecl ArgAddr(getContext(), nullptr, SourceLocation(), nullptr,
+                            getContext().VoidPtrTy);
+  Args.push_back(&ArgData);
+  Args.push_back(&ArgAddr);
+
+  const CGFunctionInfo &FI =
+    CGM.getTypes().arrangeBuiltinFunctionDeclaration(getContext().VoidTy, Args);
+
+  llvm::Function *F = llvm::Function::Create(
+      llvm::FunctionType::get(VoidTy, {VoidPtrTy, VoidPtrTy}, false),
+      llvm::GlobalValue::WeakODRLinkage, "__cfi_check_fail", &CGM.getModule());
+  F->setVisibility(llvm::GlobalValue::HiddenVisibility);
+
+  StartFunction(GlobalDecl(), CGM.getContext().VoidTy, F, FI, Args,
+                SourceLocation());
+
+  llvm::Value *Data =
+      EmitLoadOfScalar(GetAddrOfLocalVar(&ArgData), /*Volatile=*/false,
+                       CGM.getContext().VoidPtrTy, ArgData.getLocation());
+  llvm::Value *Addr =
+      EmitLoadOfScalar(GetAddrOfLocalVar(&ArgAddr), /*Volatile=*/false,
+                       CGM.getContext().VoidPtrTy, ArgAddr.getLocation());
+
+  // Data == nullptr means the calling module has trap behaviour for this check.
+  llvm::Value *DataIsNotNullPtr =
+      Builder.CreateICmpNE(Data, llvm::ConstantPointerNull::get(Int8PtrTy));
+  EmitTrapCheck(DataIsNotNullPtr);
+
+  llvm::StructType *SourceLocationTy =
+      llvm::StructType::get(VoidPtrTy, Int32Ty, Int32Ty, nullptr);
+  llvm::StructType *CfiCheckFailDataTy =
+      llvm::StructType::get(Int8Ty, SourceLocationTy, VoidPtrTy, nullptr);
+
+  llvm::Value *V = Builder.CreateConstGEP2_32(
+      CfiCheckFailDataTy,
+      Builder.CreatePointerCast(Data, CfiCheckFailDataTy->getPointerTo(0)), 0,
+      0);
+  Address CheckKindAddr(V, getIntAlign());
+  llvm::Value *CheckKind = Builder.CreateLoad(CheckKindAddr);
+
+  llvm::Value *AllVtables = llvm::MetadataAsValue::get(
+      CGM.getLLVMContext(),
+      llvm::MDString::get(CGM.getLLVMContext(), "all-vtables"));
+  llvm::Value *ValidVtable = Builder.CreateZExt(
+      Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::type_test),
+                         {Addr, AllVtables}),
+      IntPtrTy);
+
+  const std::pair<int, SanitizerMask> CheckKinds[] = {
+      {CFITCK_VCall, SanitizerKind::CFIVCall},
+      {CFITCK_NVCall, SanitizerKind::CFINVCall},
+      {CFITCK_DerivedCast, SanitizerKind::CFIDerivedCast},
+      {CFITCK_UnrelatedCast, SanitizerKind::CFIUnrelatedCast},
+      {CFITCK_ICall, SanitizerKind::CFIICall}};
+
+  SmallVector<std::pair<llvm::Value *, SanitizerMask>, 5> Checks;
+  for (auto CheckKindMaskPair : CheckKinds) {
+    int Kind = CheckKindMaskPair.first;
+    SanitizerMask Mask = CheckKindMaskPair.second;
+    llvm::Value *Cond =
+        Builder.CreateICmpNE(CheckKind, llvm::ConstantInt::get(Int8Ty, Kind));
+    if (CGM.getLangOpts().Sanitize.has(Mask))
+      EmitCheck(std::make_pair(Cond, Mask), "cfi_check_fail", {},
+                {Data, Addr, ValidVtable});
+    else
+      EmitTrapCheck(Cond);
+  }
+
+  FinishFunction();
+  // The only reference to this function will be created during LTO link.
+  // Make sure it survives until then.
+  CGM.addUsedGlobal(F);
+}
+
 void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked) {
   llvm::BasicBlock *Cont = createBasicBlock("cont");
 
@@ -2827,22 +2996,55 @@ LValue CodeGenFunction::EmitArraySubscriptExpr(const ArraySubscriptExpr *E,
   return LV;
 }
 
+static Address emitOMPArraySectionBase(CodeGenFunction &CGF, const Expr *Base,
+                                       AlignmentSource &AlignSource,
+                                       QualType BaseTy, QualType ElTy,
+                                       bool IsLowerBound) {
+  LValue BaseLVal;
+  if (auto *ASE = dyn_cast<OMPArraySectionExpr>(Base->IgnoreParenImpCasts())) {
+    BaseLVal = CGF.EmitOMPArraySectionExpr(ASE, IsLowerBound);
+    if (BaseTy->isArrayType()) {
+      Address Addr = BaseLVal.getAddress();
+      AlignSource = BaseLVal.getAlignmentSource();
+
+      // If the array type was an incomplete type, we need to make sure
+      // the decay ends up being the right type.
+      llvm::Type *NewTy = CGF.ConvertType(BaseTy);
+      Addr = CGF.Builder.CreateElementBitCast(Addr, NewTy);
+
+      // Note that VLA pointers are always decayed, so we don't need to do
+      // anything here.
+      if (!BaseTy->isVariableArrayType()) {
+        assert(isa<llvm::ArrayType>(Addr.getElementType()) &&
+               "Expected pointer to array");
+        Addr = CGF.Builder.CreateStructGEP(Addr, 0, CharUnits::Zero(),
+                                           "arraydecay");
+      }
+
+      return CGF.Builder.CreateElementBitCast(Addr,
+                                              CGF.ConvertTypeForMem(ElTy));
+    }
+    CharUnits Align = CGF.getNaturalTypeAlignment(ElTy, &AlignSource);
+    return Address(CGF.Builder.CreateLoad(BaseLVal.getAddress()), Align);
+  }
+  return CGF.EmitPointerWithAlignment(Base, &AlignSource);
+}
+
 LValue CodeGenFunction::EmitOMPArraySectionExpr(const OMPArraySectionExpr *E,
                                                 bool IsLowerBound) {
-  LValue Base;
+  QualType BaseTy;
   if (auto *ASE =
           dyn_cast<OMPArraySectionExpr>(E->getBase()->IgnoreParenImpCasts()))
-    Base = EmitOMPArraySectionExpr(ASE, IsLowerBound);
+    BaseTy = OMPArraySectionExpr::getBaseOriginalType(ASE);
   else
-    Base = EmitLValue(E->getBase());
-  QualType BaseTy = Base.getType();
-  llvm::Value *Idx = nullptr;
+    BaseTy = E->getBase()->getType();
   QualType ResultExprTy;
   if (auto *AT = getContext().getAsArrayType(BaseTy))
     ResultExprTy = AT->getElementType();
   else
     ResultExprTy = BaseTy->getPointeeType();
-  if (IsLowerBound || (!IsLowerBound && E->getColonLoc().isInvalid())) {
+  llvm::Value *Idx = nullptr;
+  if (IsLowerBound || E->getColonLoc().isInvalid()) {
     // Requesting lower bound or upper bound, but without provided length and
     // without ':' symbol for the default length -> length = 1.
     // Idx = LowerBound ?: 0;
@@ -2853,9 +3055,9 @@ LValue CodeGenFunction::EmitOMPArraySectionExpr(const OMPArraySectionExpr *E,
     } else
       Idx = llvm::ConstantInt::getNullValue(IntPtrTy);
   } else {
-    // Try to emit length or lower bound as constant. If this is possible, 1 is
-    // subtracted from constant length or lower bound. Otherwise, emit LLVM IR
-    // (LB + Len) - 1.
+    // Try to emit length or lower bound as constant. If this is possible, 1
+    // is subtracted from constant length or lower bound. Otherwise, emit LLVM
+    // IR (LB + Len) - 1.
     auto &C = CGM.getContext();
     auto *Length = E->getLength();
     llvm::APSInt ConstLength;
@@ -2901,12 +3103,15 @@ LValue CodeGenFunction::EmitOMPArraySectionExpr(const OMPArraySectionExpr *E,
         Idx = llvm::ConstantInt::get(IntPtrTy, ConstLength + ConstLowerBound);
     } else {
       // Idx = ArraySize - 1;
-      if (auto *VAT = C.getAsVariableArrayType(BaseTy)) {
+      QualType ArrayTy = BaseTy->isPointerType()
+                             ? E->getBase()->IgnoreParenImpCasts()->getType()
+                             : BaseTy;
+      if (auto *VAT = C.getAsVariableArrayType(ArrayTy)) {
         Length = VAT->getSizeExpr();
         if (Length->isIntegerConstantExpr(ConstLength, C))
           Length = nullptr;
       } else {
-        auto *CAT = C.getAsConstantArrayType(BaseTy);
+        auto *CAT = C.getAsConstantArrayType(ArrayTy);
         ConstLength = CAT->getSize();
       }
       if (Length) {
@@ -2925,52 +3130,56 @@ LValue CodeGenFunction::EmitOMPArraySectionExpr(const OMPArraySectionExpr *E,
   }
   assert(Idx);
 
-  llvm::Value *EltPtr;
-  QualType FixedSizeEltType = ResultExprTy;
+  Address EltPtr = Address::invalid();
+  AlignmentSource AlignSource;
   if (auto *VLA = getContext().getAsVariableArrayType(ResultExprTy)) {
+    // The base must be a pointer, which is not an aggregate.  Emit
+    // it.  It needs to be emitted first in case it's what captures
+    // the VLA bounds.
+    Address Base =
+        emitOMPArraySectionBase(*this, E->getBase(), AlignSource, BaseTy,
+                                VLA->getElementType(), IsLowerBound);
     // The element count here is the total number of non-VLA elements.
-    llvm::Value *numElements = getVLASize(VLA).first;
-    FixedSizeEltType = getFixedSizeElementType(getContext(), VLA);
+    llvm::Value *NumElements = getVLASize(VLA).first;
 
     // Effectively, the multiply by the VLA size is part of the GEP.
     // GEP indexes are signed, and scaling an index isn't permitted to
     // signed-overflow, so we use the same semantics for our explicit
     // multiply.  We suppress this if overflow is not undefined behavior.
-    if (getLangOpts().isSignedOverflowDefined()) {
-      Idx = Builder.CreateMul(Idx, numElements);
-      EltPtr = Builder.CreateGEP(Base.getPointer(), Idx, "arrayidx");
-    } else {
-      Idx = Builder.CreateNSWMul(Idx, numElements);
-      EltPtr = Builder.CreateInBoundsGEP(Base.getPointer(), Idx, "arrayidx");
-    }
-  } else if (BaseTy->isConstantArrayType()) {
-    llvm::Value *ArrayPtr = Base.getPointer();
-    llvm::Value *Zero = llvm::ConstantInt::getNullValue(IntPtrTy);
-    llvm::Value *Args[] = {Zero, Idx};
-
     if (getLangOpts().isSignedOverflowDefined())
-      EltPtr = Builder.CreateGEP(ArrayPtr, Args, "arrayidx");
+      Idx = Builder.CreateMul(Idx, NumElements);
     else
-      EltPtr = Builder.CreateInBoundsGEP(ArrayPtr, Args, "arrayidx");
-  } else {
-    // The base must be a pointer, which is not an aggregate.  Emit it.
-    if (getLangOpts().isSignedOverflowDefined())
-      EltPtr = Builder.CreateGEP(Base.getPointer(), Idx, "arrayidx");
+      Idx = Builder.CreateNSWMul(Idx, NumElements);
+    EltPtr = emitArraySubscriptGEP(*this, Base, Idx, VLA->getElementType(),
+                                   !getLangOpts().isSignedOverflowDefined());
+  } else if (const Expr *Array = isSimpleArrayDecayOperand(E->getBase())) {
+    // If this is A[i] where A is an array, the frontend will have decayed the
+    // base to be a ArrayToPointerDecay implicit cast.  While correct, it is
+    // inefficient at -O0 to emit a "gep A, 0, 0" when codegen'ing it, then a
+    // "gep x, i" here.  Emit one "gep A, 0, i".
+    assert(Array->getType()->isArrayType() &&
+           "Array to pointer decay must have array source type!");
+    LValue ArrayLV;
+    // For simple multidimensional array indexing, set the 'accessed' flag for
+    // better bounds-checking of the base expression.
+    if (const auto *ASE = dyn_cast<ArraySubscriptExpr>(Array))
+      ArrayLV = EmitArraySubscriptExpr(ASE, /*Accessed*/ true);
     else
-      EltPtr = Builder.CreateInBoundsGEP(Base.getPointer(), Idx, "arrayidx");
-  }
-
-  CharUnits EltAlign =
-    Base.getAlignment().alignmentOfArrayElement(
-                          getContext().getTypeSizeInChars(FixedSizeEltType));
-
-  // Limit the alignment to that of the result type.
-  LValue LV = MakeAddrLValue(Address(EltPtr, EltAlign), ResultExprTy,
-                             Base.getAlignmentSource());
+      ArrayLV = EmitLValue(Array);
 
-  LV.getQuals().setAddressSpace(BaseTy.getAddressSpace());
+    // Propagate the alignment from the array itself to the result.
+    EltPtr = emitArraySubscriptGEP(
+        *this, ArrayLV.getAddress(), {CGM.getSize(CharUnits::Zero()), Idx},
+        ResultExprTy, !getLangOpts().isSignedOverflowDefined());
+    AlignSource = ArrayLV.getAlignmentSource();
+  } else {
+    Address Base = emitOMPArraySectionBase(*this, E->getBase(), AlignSource,
+                                           BaseTy, ResultExprTy, IsLowerBound);
+    EltPtr = emitArraySubscriptGEP(*this, Base, Idx, ResultExprTy,
+                                   !getLangOpts().isSignedOverflowDefined());
+  }
 
-  return LV;
+  return MakeAddrLValue(EltPtr, ResultExprTy, AlignSource);
 }
 
 LValue CodeGenFunction::
@@ -3508,6 +3717,10 @@ RValue CodeGenFunction::EmitRValueForField(LValue LV,
   case TEK_Aggregate:
     return FieldLV.asAggregateRValue();
   case TEK_Scalar:
+    // This routine is used to load fields one-by-one to perform a copy, so
+    // don't load reference fields.
+    if (FD->getType()->isReferenceType())
+      return RValue::get(FieldLV.getPointer());
     return EmitLoadOfLValue(FieldLV, Loc);
   }
   llvm_unreachable("bad evaluation kind");
@@ -3851,25 +4064,28 @@ RValue CodeGenFunction::EmitCall(QualType CalleeType, llvm::Value *Callee,
   if (SanOpts.has(SanitizerKind::CFIICall) &&
       (!TargetDecl || !isa<FunctionDecl>(TargetDecl))) {
     SanitizerScope SanScope(this);
+    EmitSanitizerStatReport(llvm::SanStat_CFI_ICall);
 
     llvm::Metadata *MD = CGM.CreateMetadataIdentifierForType(QualType(FnType, 0));
-    llvm::Value *BitSetName = llvm::MetadataAsValue::get(getLLVMContext(), MD);
+    llvm::Value *TypeId = llvm::MetadataAsValue::get(getLLVMContext(), MD);
 
     llvm::Value *CastedCallee = Builder.CreateBitCast(Callee, Int8PtrTy);
-    llvm::Value *BitSetTest =
-        Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::bitset_test),
-                           {CastedCallee, BitSetName});
+    llvm::Value *TypeTest = Builder.CreateCall(
+        CGM.getIntrinsic(llvm::Intrinsic::type_test), {CastedCallee, TypeId});
 
-    auto TypeId = CGM.CreateCfiIdForTypeMetadata(MD);
-    if (CGM.getCodeGenOpts().SanitizeCfiCrossDso && TypeId) {
-      EmitCfiSlowPathCheck(BitSetTest, TypeId, CastedCallee);
+    auto CrossDsoTypeId = CGM.CreateCrossDsoCfiTypeId(MD);
+    llvm::Constant *StaticData[] = {
+        llvm::ConstantInt::get(Int8Ty, CFITCK_ICall),
+        EmitCheckSourceLocation(E->getLocStart()),
+        EmitCheckTypeDescriptor(QualType(FnType, 0)),
+    };
+    if (CGM.getCodeGenOpts().SanitizeCfiCrossDso && CrossDsoTypeId) {
+      EmitCfiSlowPathCheck(SanitizerKind::CFIICall, TypeTest, CrossDsoTypeId,
+                           CastedCallee, StaticData);
     } else {
-      llvm::Constant *StaticData[] = {
-          EmitCheckSourceLocation(E->getLocStart()),
-          EmitCheckTypeDescriptor(QualType(FnType, 0)),
-      };
-      EmitCheck(std::make_pair(BitSetTest, SanitizerKind::CFIICall),
-                "cfi_bad_icall", StaticData, CastedCallee);
+      EmitCheck(std::make_pair(TypeTest, SanitizerKind::CFIICall),
+                "cfi_check_fail", StaticData,
+                {CastedCallee, llvm::UndefValue::get(IntPtrTy)});
     }
   }
 
diff --git a/lib/CodeGen/CGExprAgg.cpp b/lib/CodeGen/CGExprAgg.cpp
index a4547a9982be..6d18843591f3 100644
--- a/lib/CodeGen/CGExprAgg.cpp
+++ b/lib/CodeGen/CGExprAgg.cpp
@@ -175,6 +175,7 @@ public:
   }
   void VisitCXXBindTemporaryExpr(CXXBindTemporaryExpr *E);
   void VisitCXXConstructExpr(const CXXConstructExpr *E);
+  void VisitCXXInheritedCtorInitExpr(const CXXInheritedCtorInitExpr *E);
   void VisitLambdaExpr(LambdaExpr *E);
   void VisitCXXStdInitializerListExpr(CXXStdInitializerListExpr *E);
   void VisitExprWithCleanups(ExprWithCleanups *E);
@@ -967,12 +968,9 @@ void AggExprEmitter::VisitVAArgExpr(VAArgExpr *VE) {
   Address ArgValue = Address::invalid();
   Address ArgPtr = CGF.EmitVAArg(VE, ArgValue);
 
+  // If EmitVAArg fails, emit an error.
   if (!ArgPtr.isValid()) {
-    // If EmitVAArg fails, we fall back to the LLVM instruction.
-    llvm::Value *Val = Builder.CreateVAArg(ArgValue.getPointer(),
-                                           CGF.ConvertType(VE->getType()));
-    if (!Dest.isIgnored())
-      Builder.CreateStore(Val, Dest.getAddress());
+    CGF.ErrorUnsupported(VE, "aggregate va_arg expression");
     return;
   }
 
@@ -1001,6 +999,14 @@ AggExprEmitter::VisitCXXConstructExpr(const CXXConstructExpr *E) {
   CGF.EmitCXXConstructExpr(E, Slot);
 }
 
+void AggExprEmitter::VisitCXXInheritedCtorInitExpr(
+    const CXXInheritedCtorInitExpr *E) {
+  AggValueSlot Slot = EnsureSlot(E->getType());
+  CGF.EmitInheritedCXXConstructorCall(
+      E->getConstructor(), E->constructsVBase(), Slot.getAddress(),
+      E->inheritedFromVBase(), E);
+}
+
 void
 AggExprEmitter::VisitLambdaExpr(LambdaExpr *E) {
   AggValueSlot Slot = EnsureSlot(E->getType());
@@ -1174,6 +1180,38 @@ void AggExprEmitter::VisitInitListExpr(InitListExpr *E) {
   unsigned NumInitElements = E->getNumInits();
   RecordDecl *record = E->getType()->castAs<RecordType>()->getDecl();
 
+  // We'll need to enter cleanup scopes in case any of the element
+  // initializers throws an exception.
+  SmallVector<EHScopeStack::stable_iterator, 16> cleanups;
+  llvm::Instruction *cleanupDominator = nullptr;
+
+  unsigned curInitIndex = 0;
+
+  // Emit initialization of base classes.
+  if (auto *CXXRD = dyn_cast<CXXRecordDecl>(record)) {
+    assert(E->getNumInits() >= CXXRD->getNumBases() &&
+           "missing initializer for base class");
+    for (auto &Base : CXXRD->bases()) {
+      assert(!Base.isVirtual() && "should not see vbases here");
+      auto *BaseRD = Base.getType()->getAsCXXRecordDecl();
+      Address V = CGF.GetAddressOfDirectBaseInCompleteClass(
+          Dest.getAddress(), CXXRD, BaseRD,
+          /*isBaseVirtual*/ false);
+      AggValueSlot AggSlot =
+        AggValueSlot::forAddr(V, Qualifiers(),
+                              AggValueSlot::IsDestructed,
+                              AggValueSlot::DoesNotNeedGCBarriers,
+                              AggValueSlot::IsNotAliased);
+      CGF.EmitAggExpr(E->getInit(curInitIndex++), AggSlot);
+
+      if (QualType::DestructionKind dtorKind =
+              Base.getType().isDestructedType()) {
+        CGF.pushDestroy(dtorKind, V, Base.getType());
+        cleanups.push_back(CGF.EHStack.stable_begin());
+      }
+    }
+  }
+
   // Prepare a 'this' for CXXDefaultInitExprs.
   CodeGenFunction::FieldConstructionScope FCS(CGF, Dest.getAddress());
 
@@ -1207,14 +1245,8 @@ void AggExprEmitter::VisitInitListExpr(InitListExpr *E) {
     return;
   }
 
-  // We'll need to enter cleanup scopes in case any of the member
-  // initializers throw an exception.
-  SmallVector<EHScopeStack::stable_iterator, 16> cleanups;
-  llvm::Instruction *cleanupDominator = nullptr;
-
   // Here we iterate over the fields; this makes it simpler to both
   // default-initialize fields and skip over unnamed fields.
-  unsigned curInitIndex = 0;
   for (const auto *field : record->fields()) {
     // We're done once we hit the flexible array member.
     if (field->getType()->isIncompleteArrayType())
@@ -1320,6 +1352,10 @@ static CharUnits GetNumNonZeroBytesInInit(const Expr *E, CodeGenFunction &CGF) {
       CharUnits NumNonZeroBytes = CharUnits::Zero();
       
       unsigned ILEElement = 0;
+      if (auto *CXXRD = dyn_cast<CXXRecordDecl>(SD))
+        while (ILEElement != CXXRD->getNumBases())
+          NumNonZeroBytes +=
+              GetNumNonZeroBytesInInit(ILE->getInit(ILEElement++), CGF);
       for (const auto *Field : SD->fields()) {
         // We're done once we hit the flexible array member or run out of
         // InitListExpr elements.
diff --git a/lib/CodeGen/CGExprCXX.cpp b/lib/CodeGen/CGExprCXX.cpp
index 604cde76a7b1..eec2aceb88a2 100644
--- a/lib/CodeGen/CGExprCXX.cpp
+++ b/lib/CodeGen/CGExprCXX.cpp
@@ -24,10 +24,11 @@
 using namespace clang;
 using namespace CodeGen;
 
-static RequiredArgs commonEmitCXXMemberOrOperatorCall(
-    CodeGenFunction &CGF, const CXXMethodDecl *MD, llvm::Value *Callee,
-    ReturnValueSlot ReturnValue, llvm::Value *This, llvm::Value *ImplicitParam,
-    QualType ImplicitParamTy, const CallExpr *CE, CallArgList &Args) {
+static RequiredArgs
+commonEmitCXXMemberOrOperatorCall(CodeGenFunction &CGF, const CXXMethodDecl *MD,
+                                  llvm::Value *This, llvm::Value *ImplicitParam,
+                                  QualType ImplicitParamTy, const CallExpr *CE,
+                                  CallArgList &Args) {
   assert(CE == nullptr || isa<CXXMemberCallExpr>(CE) ||
          isa<CXXOperatorCallExpr>(CE));
   assert(MD->isInstance() &&
@@ -53,7 +54,7 @@ static RequiredArgs commonEmitCXXMemberOrOperatorCall(
   }
 
   const FunctionProtoType *FPT = MD->getType()->castAs<FunctionProtoType>();
-  RequiredArgs required = RequiredArgs::forPrototypePlus(FPT, Args.size());
+  RequiredArgs required = RequiredArgs::forPrototypePlus(FPT, Args.size(), MD);
 
   // And the rest of the call args.
   if (CE) {
@@ -76,21 +77,20 @@ RValue CodeGenFunction::EmitCXXMemberOrOperatorCall(
   const FunctionProtoType *FPT = MD->getType()->castAs<FunctionProtoType>();
   CallArgList Args;
   RequiredArgs required = commonEmitCXXMemberOrOperatorCall(
-      *this, MD, Callee, ReturnValue, This, ImplicitParam, ImplicitParamTy, CE,
-      Args);
+      *this, MD, This, ImplicitParam, ImplicitParamTy, CE, Args);
   return EmitCall(CGM.getTypes().arrangeCXXMethodCall(Args, FPT, required),
                   Callee, ReturnValue, Args, MD);
 }
 
-RValue CodeGenFunction::EmitCXXStructorCall(
-    const CXXMethodDecl *MD, llvm::Value *Callee, ReturnValueSlot ReturnValue,
-    llvm::Value *This, llvm::Value *ImplicitParam, QualType ImplicitParamTy,
-    const CallExpr *CE, StructorType Type) {
+RValue CodeGenFunction::EmitCXXDestructorCall(
+    const CXXDestructorDecl *DD, llvm::Value *Callee, llvm::Value *This,
+    llvm::Value *ImplicitParam, QualType ImplicitParamTy, const CallExpr *CE,
+    StructorType Type) {
   CallArgList Args;
-  commonEmitCXXMemberOrOperatorCall(*this, MD, Callee, ReturnValue, This,
-                                    ImplicitParam, ImplicitParamTy, CE, Args);
-  return EmitCall(CGM.getTypes().arrangeCXXStructorDeclaration(MD, Type),
-                  Callee, ReturnValue, Args, MD);
+  commonEmitCXXMemberOrOperatorCall(*this, DD, This, ImplicitParam,
+                                    ImplicitParamTy, CE, Args);
+  return EmitCall(CGM.getTypes().arrangeCXXStructorDeclaration(DD, Type),
+                  Callee, ReturnValueSlot(), Args, DD);
 }
 
 static CXXRecordDecl *getCXXRecord(const Expr *E) {
@@ -259,7 +259,8 @@ RValue CodeGenFunction::EmitCXXMemberOrOperatorMemberCallExpr(
     if (SanOpts.has(SanitizerKind::CFINVCall) &&
         MD->getParent()->isDynamicClass()) {
       llvm::Value *VTable = GetVTablePtr(This, Int8PtrTy, MD->getParent());
-      EmitVTablePtrCheckForCall(MD, VTable, CFITCK_NVCall, CE->getLocStart());
+      EmitVTablePtrCheckForCall(MD->getParent(), VTable, CFITCK_NVCall,
+                                CE->getLocStart());
     }
 
     if (getLangOpts().AppleKext && MD->isVirtual() && HasQualifier)
@@ -273,7 +274,7 @@ RValue CodeGenFunction::EmitCXXMemberOrOperatorMemberCallExpr(
 
   if (MD->isVirtual()) {
     This = CGM.getCXXABI().adjustThisArgumentForVirtualFunctionCall(
-        *this, MD, This, UseVirtualCall);
+        *this, CalleeDecl, This, UseVirtualCall);
   }
 
   return EmitCXXMemberOrOperatorCall(MD, Callee, ReturnValue, This.getPointer(),
@@ -323,10 +324,11 @@ CodeGenFunction::EmitCXXMemberPointerCallExpr(const CXXMemberCallExpr *E,
   // Push the this ptr.
   Args.add(RValue::get(ThisPtrForCall), ThisType);
 
-  RequiredArgs required = RequiredArgs::forPrototypePlus(FPT, 1);
-  
+  RequiredArgs required =
+      RequiredArgs::forPrototypePlus(FPT, 1, /*FD=*/nullptr);
+
   // And the rest of the call args
-  EmitCallArgs(Args, FPT, E->arguments(), E->getDirectCallee());
+  EmitCallArgs(Args, FPT, E->arguments());
   return EmitCall(CGM.getTypes().arrangeCXXMethodCall(Args, FPT, required),
                   Callee, ReturnValue, Args);
 }
@@ -369,6 +371,9 @@ static void EmitNullBaseClassInitialization(CodeGenFunction &CGF,
   std::vector<CharUnits> VBPtrOffsets =
       CGF.CGM.getCXXABI().getVBPtrOffsets(Base);
   for (CharUnits VBPtrOffset : VBPtrOffsets) {
+    // Stop before we hit any virtual base pointers located in virtual bases.
+    if (VBPtrOffset >= NVSize)
+      break;
     std::pair<CharUnits, CharUnits> LastStore = Stores.pop_back_val();
     CharUnits LastStoreOffset = LastStore.first;
     CharUnits LastStoreSize = LastStore.second;
@@ -471,8 +476,8 @@ CodeGenFunction::EmitCXXConstructExpr(const CXXConstructExpr *E,
     }
   }
   
-  if (const ConstantArrayType *arrayType 
-        = getContext().getAsConstantArrayType(E->getType())) {
+  if (const ArrayType *arrayType
+        = getContext().getAsArrayType(E->getType())) {
     EmitCXXAggrConstructorCall(CD, arrayType, Dest.getAddress(), E);
   } else {
     CXXCtorType Type = Ctor_Complete;
@@ -1010,15 +1015,18 @@ void CodeGenFunction::EmitNewArrayInitializer(
   if (auto *ILE = dyn_cast<InitListExpr>(Init)) {
     if (const RecordType *RType = ILE->getType()->getAs<RecordType>()) {
       if (RType->getDecl()->isStruct()) {
-        unsigned NumFields = 0;
+        unsigned NumElements = 0;
+        if (auto *CXXRD = dyn_cast<CXXRecordDecl>(RType->getDecl()))
+          NumElements = CXXRD->getNumBases();
         for (auto *Field : RType->getDecl()->fields())
           if (!Field->isUnnamedBitfield())
-            ++NumFields;
-        if (ILE->getNumInits() == NumFields)
+            ++NumElements;
+        // FIXME: Recurse into nested InitListExprs.
+        if (ILE->getNumInits() == NumElements)
           for (unsigned i = 0, e = ILE->getNumInits(); i != e; ++i)
             if (!isa<ImplicitValueInitExpr>(ILE->getInit(i)))
-              --NumFields;
-        if (ILE->getNumInits() == NumFields && TryMemsetInitialization())
+              --NumElements;
+        if (ILE->getNumInits() == NumElements && TryMemsetInitialization())
           return;
       }
     }
diff --git a/lib/CodeGen/CGExprConstant.cpp b/lib/CodeGen/CGExprConstant.cpp
index ee049f1810a2..803b39907dd7 100644
--- a/lib/CodeGen/CGExprConstant.cpp
+++ b/lib/CodeGen/CGExprConstant.cpp
@@ -111,7 +111,7 @@ AppendBytes(CharUnits FieldOffsetInChars, llvm::Constant *InitCst) {
 
   // Round up the field offset to the alignment of the field type.
   CharUnits AlignedNextFieldOffsetInChars =
-      NextFieldOffsetInChars.RoundUpToAlignment(FieldAlignment);
+      NextFieldOffsetInChars.alignTo(FieldAlignment);
 
   if (AlignedNextFieldOffsetInChars < FieldOffsetInChars) {
     // We need to append padding.
@@ -121,7 +121,7 @@ AppendBytes(CharUnits FieldOffsetInChars, llvm::Constant *InitCst) {
            "Did not add enough padding!");
 
     AlignedNextFieldOffsetInChars =
-        NextFieldOffsetInChars.RoundUpToAlignment(FieldAlignment);
+        NextFieldOffsetInChars.alignTo(FieldAlignment);
   }
 
   if (AlignedNextFieldOffsetInChars > FieldOffsetInChars) {
@@ -162,8 +162,8 @@ void ConstStructBuilder::AppendBitField(const FieldDecl *Field,
   if (FieldOffset > NextFieldOffsetInBits) {
     // We need to add padding.
     CharUnits PadSize = Context.toCharUnitsFromBits(
-      llvm::RoundUpToAlignment(FieldOffset - NextFieldOffsetInBits, 
-                               Context.getTargetInfo().getCharAlign()));
+        llvm::alignTo(FieldOffset - NextFieldOffsetInBits,
+                      Context.getTargetInfo().getCharAlign()));
 
     AppendPadding(PadSize);
   }
@@ -334,7 +334,7 @@ void ConstStructBuilder::ConvertStructToPacked() {
     CharUnits ElementAlign = CharUnits::fromQuantity(
       CGM.getDataLayout().getABITypeAlignment(C->getType()));
     CharUnits AlignedElementOffsetInChars =
-      ElementOffsetInChars.RoundUpToAlignment(ElementAlign);
+        ElementOffsetInChars.alignTo(ElementAlign);
 
     if (AlignedElementOffsetInChars > ElementOffsetInChars) {
       // We need some padding.
@@ -368,7 +368,14 @@ bool ConstStructBuilder::Build(InitListExpr *ILE) {
 
   unsigned FieldNo = 0;
   unsigned ElementNo = 0;
-  
+
+  // Bail out if we have base classes. We could support these, but they only
+  // arise in C++1z where we will have already constant folded most interesting
+  // cases. FIXME: There are still a few more cases we can handle this way.
+  if (auto *CXXRD = dyn_cast<CXXRecordDecl>(RD))
+    if (CXXRD->getNumBases())
+      return false;
+
   for (RecordDecl::field_iterator Field = RD->field_begin(),
        FieldEnd = RD->field_end(); Field != FieldEnd; ++Field, ++FieldNo) {
     // If this is a union, skip all the fields that aren't being initialized.
@@ -508,13 +515,12 @@ llvm::Constant *ConstStructBuilder::Finalize(QualType Ty) {
   } else {
     // Append tail padding if necessary.
     CharUnits LLVMSizeInChars =
-        NextFieldOffsetInChars.RoundUpToAlignment(LLVMStructAlignment);
+        NextFieldOffsetInChars.alignTo(LLVMStructAlignment);
 
     if (LLVMSizeInChars != LayoutSizeInChars)
       AppendTailPadding(LayoutSizeInChars);
 
-    LLVMSizeInChars =
-        NextFieldOffsetInChars.RoundUpToAlignment(LLVMStructAlignment);
+    LLVMSizeInChars = NextFieldOffsetInChars.alignTo(LLVMStructAlignment);
 
     // Check if we need to convert the struct to a packed struct.
     if (NextFieldOffsetInChars <= LayoutSizeInChars &&
@@ -526,8 +532,7 @@ llvm::Constant *ConstStructBuilder::Finalize(QualType Ty) {
              "Converting to packed did not help!");
     }
 
-    LLVMSizeInChars =
-        NextFieldOffsetInChars.RoundUpToAlignment(LLVMStructAlignment);
+    LLVMSizeInChars = NextFieldOffsetInChars.alignTo(LLVMStructAlignment);
 
     assert(LayoutSizeInChars == LLVMSizeInChars &&
            "Tail padding mismatch!");
@@ -546,8 +551,9 @@ llvm::Constant *ConstStructBuilder::Finalize(QualType Ty) {
 
   llvm::Constant *Result = llvm::ConstantStruct::get(STy, Elements);
 
-  assert(NextFieldOffsetInChars.RoundUpToAlignment(getAlignment(Result)) ==
-         getSizeInChars(Result) && "Size mismatch!");
+  assert(NextFieldOffsetInChars.alignTo(getAlignment(Result)) ==
+             getSizeInChars(Result) &&
+         "Size mismatch!");
 
   return Result;
 }
@@ -758,6 +764,12 @@ public:
     return Visit(DIE->getExpr());
   }
 
+  llvm::Constant *VisitExprWithCleanups(ExprWithCleanups *E) {
+    if (!E->cleanupsHaveSideEffects())
+      return Visit(E->getSubExpr());
+    return nullptr;
+  }
+
   llvm::Constant *VisitMaterializeTemporaryExpr(MaterializeTemporaryExpr *E) {
     return Visit(E->GetTemporaryExpr());
   }
@@ -1125,6 +1137,13 @@ bool ConstStructBuilder::Build(ConstExprEmitter *Emitter,
   unsigned FieldNo = -1;
   unsigned ElementNo = 0;
 
+  // Bail out if we have base classes. We could support these, but they only
+  // arise in C++1z where we will have already constant folded most interesting
+  // cases. FIXME: There are still a few more cases we can handle this way.
+  if (auto *CXXRD = dyn_cast<CXXRecordDecl>(RD))
+    if (CXXRD->getNumBases())
+      return false;
+
   for (FieldDecl *Field : RD->fields()) {
     ++FieldNo;
 
@@ -1301,8 +1320,14 @@ llvm::Constant *CodeGenModule::EmitConstantValue(const APValue &Value,
 
       // Convert to the appropriate type; this could be an lvalue for
       // an integer.
-      if (isa<llvm::PointerType>(DestTy))
+      if (isa<llvm::PointerType>(DestTy)) {
+        // Convert the integer to a pointer-sized integer before converting it
+        // to a pointer.
+        C = llvm::ConstantExpr::getIntegerCast(
+            C, getDataLayout().getIntPtrType(DestTy),
+            /*isSigned=*/false);
         return llvm::ConstantExpr::getIntToPtr(C, DestTy);
+      }
 
       // If the types don't match this should only be a truncate.
       if (C->getType() != DestTy)
diff --git a/lib/CodeGen/CGExprScalar.cpp b/lib/CodeGen/CGExprScalar.cpp
index 268e7967b808..064bc9532a6d 100644
--- a/lib/CodeGen/CGExprScalar.cpp
+++ b/lib/CodeGen/CGExprScalar.cpp
@@ -818,7 +818,7 @@ Value *ScalarExprEmitter::EmitScalarConversion(Value *Src, QualType SrcType,
            "Splatted expr doesn't match with vector element type?");
 
     // Splat the element across to all elements
-    unsigned NumElements = cast<llvm::VectorType>(DstTy)->getNumElements();
+    unsigned NumElements = DstTy->getVectorNumElements();
     return Builder.CreateVectorSplat(NumElements, Src, "splat");
   }
 
@@ -984,8 +984,7 @@ Value *ScalarExprEmitter::VisitExpr(Expr *E) {
 
 Value *ScalarExprEmitter::VisitShuffleVectorExpr(ShuffleVectorExpr *E) {
   // Vector Mask Case
-  if (E->getNumSubExprs() == 2 ||
-      (E->getNumSubExprs() == 3 && E->getExpr(2)->getType()->isVectorType())) {
+  if (E->getNumSubExprs() == 2) {
     Value *LHS = CGF.EmitScalarExpr(E->getExpr(0));
     Value *RHS = CGF.EmitScalarExpr(E->getExpr(1));
     Value *Mask;
@@ -993,22 +992,7 @@ Value *ScalarExprEmitter::VisitShuffleVectorExpr(ShuffleVectorExpr *E) {
     llvm::VectorType *LTy = cast<llvm::VectorType>(LHS->getType());
     unsigned LHSElts = LTy->getNumElements();
 
-    if (E->getNumSubExprs() == 3) {
-      Mask = CGF.EmitScalarExpr(E->getExpr(2));
-
-      // Shuffle LHS & RHS into one input vector.
-      SmallVector<llvm::Constant*, 32> concat;
-      for (unsigned i = 0; i != LHSElts; ++i) {
-        concat.push_back(Builder.getInt32(2*i));
-        concat.push_back(Builder.getInt32(2*i+1));
-      }
-
-      Value* CV = llvm::ConstantVector::get(concat);
-      LHS = Builder.CreateShuffleVector(LHS, RHS, CV, "concat");
-      LHSElts *= 2;
-    } else {
-      Mask = RHS;
-    }
+    Mask = RHS;
 
     llvm::VectorType *MTy = cast<llvm::VectorType>(Mask->getType());
 
@@ -1366,8 +1350,9 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
   QualType DestTy = CE->getType();
   CastKind Kind = CE->getCastKind();
 
-  if (!DestTy->isVoidType())
-    TestAndClearIgnoreResultAssign();
+  // These cases are generally not written to ignore the result of
+  // evaluating their sub-expressions, so we clear this now.
+  bool Ignored = TestAndClearIgnoreResultAssign();
 
   // Since almost all cast kinds apply to scalars, this switch doesn't have
   // a default case, so the compiler will warn on a missing case.  The cases
@@ -1410,7 +1395,10 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
   }
   case CK_AddressSpaceConversion: {
     Value *Src = Visit(const_cast<Expr*>(E));
-    return Builder.CreateAddrSpaceCast(Src, ConvertType(DestTy));
+    // Since target may map different address spaces in AST to the same address
+    // space, an address space conversion may end up as a bitcast.
+    return Builder.CreatePointerBitCastOrAddrSpaceCast(Src,
+                                                       ConvertType(DestTy));
   }
   case CK_AtomicToNonAtomic:
   case CK_NonAtomicToAtomic:
@@ -1494,11 +1482,8 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
     return CGF.EmitARCRetainScalarExpr(E);
   case CK_ARCConsumeObject:
     return CGF.EmitObjCConsumeObject(E->getType(), Visit(E));
-  case CK_ARCReclaimReturnedObject: {
-    llvm::Value *value = Visit(E);
-    value = CGF.EmitARCRetainAutoreleasedReturnValue(value);
-    return CGF.EmitObjCConsumeObject(E->getType(), value);
-  }
+  case CK_ARCReclaimReturnedObject:
+    return CGF.EmitARCReclaimReturnedObject(E, /*allowUnsafe*/ Ignored);
   case CK_ARCExtendBlockObject:
     return CGF.EmitARCExtendBlockObject(E);
 
@@ -1544,7 +1529,7 @@ Value *ScalarExprEmitter::VisitCastExpr(CastExpr *CE) {
     llvm::Type *DstTy = ConvertType(DestTy);
     Value *Elt = Visit(const_cast<Expr*>(E));
     // Splat the element across to all elements
-    unsigned NumElements = cast<llvm::VectorType>(DstTy)->getNumElements();
+    unsigned NumElements = DstTy->getVectorNumElements();
     return Builder.CreateVectorSplat(NumElements, Elt, "splat");
   }
 
@@ -1654,13 +1639,14 @@ ScalarExprEmitter::EmitScalarPrePostIncDec(const UnaryOperator *E, LValue LV,
       llvm::Value *True = CGF.EmitToMemory(Builder.getTrue(), type);
       if (isPre) {
         Builder.CreateStore(True, LV.getAddress(), LV.isVolatileQualified())
-          ->setAtomic(llvm::SequentiallyConsistent);
+          ->setAtomic(llvm::AtomicOrdering::SequentiallyConsistent);
         return Builder.getTrue();
       }
       // For atomic bool increment, we just store true and return it for
       // preincrement, do an atomic swap with true for postincrement
-        return Builder.CreateAtomicRMW(llvm::AtomicRMWInst::Xchg,
-            LV.getPointer(), True, llvm::SequentiallyConsistent);
+      return Builder.CreateAtomicRMW(
+          llvm::AtomicRMWInst::Xchg, LV.getPointer(), True,
+          llvm::AtomicOrdering::SequentiallyConsistent);
     }
     // Special case for atomic increment / decrement on integers, emit
     // atomicrmw instructions.  We skip this if we want to be doing overflow
@@ -1677,7 +1663,7 @@ ScalarExprEmitter::EmitScalarPrePostIncDec(const UnaryOperator *E, LValue LV,
       llvm::Value *amt = CGF.EmitToMemory(
           llvm::ConstantInt::get(ConvertType(type), 1, true), type);
       llvm::Value *old = Builder.CreateAtomicRMW(aop,
-          LV.getPointer(), amt, llvm::SequentiallyConsistent);
+          LV.getPointer(), amt, llvm::AtomicOrdering::SequentiallyConsistent);
       return isPre ? Builder.CreateBinOp(op, old, amt) : old;
     }
     value = EmitLoadOfLValue(LV, E->getExprLoc());
@@ -1794,15 +1780,19 @@ ScalarExprEmitter::EmitScalarPrePostIncDec(const UnaryOperator *E, LValue LV,
       amt = llvm::ConstantFP::get(VMContext,
                                   llvm::APFloat(static_cast<double>(amount)));
     else {
-      // Remaining types are either Half or LongDouble.  Convert from float.
+      // Remaining types are Half, LongDouble or __float128. Convert from float.
       llvm::APFloat F(static_cast<float>(amount));
       bool ignored;
+      const llvm::fltSemantics *FS;
       // Don't use getFloatTypeSemantics because Half isn't
       // necessarily represented using the "half" LLVM type.
-      F.convert(value->getType()->isHalfTy()
-                    ? CGF.getTarget().getHalfFormat()
-                    : CGF.getTarget().getLongDoubleFormat(),
-                llvm::APFloat::rmTowardZero, &ignored);
+      if (value->getType()->isFP128Ty())
+        FS = &CGF.getTarget().getFloat128Format();
+      else if (value->getType()->isHalfTy())
+        FS = &CGF.getTarget().getHalfFormat();
+      else
+        FS = &CGF.getTarget().getLongDoubleFormat();
+      F.convert(*FS, llvm::APFloat::rmTowardZero, &ignored);
       amt = llvm::ConstantFP::get(VMContext, F);
     }
     value = Builder.CreateFAdd(value, amt, isInc ? "inc" : "dec");
@@ -2159,7 +2149,7 @@ LValue ScalarExprEmitter::EmitCompoundAssignLValue(
                                  E->getExprLoc()),
             LHSTy);
         Builder.CreateAtomicRMW(aop, LHSLV.getPointer(), amt,
-            llvm::SequentiallyConsistent);
+            llvm::AtomicOrdering::SequentiallyConsistent);
         return LHSLV;
       }
     }
@@ -2993,15 +2983,17 @@ Value *ScalarExprEmitter::VisitBinAssign(const BinaryOperator *E) {
     std::tie(LHS, RHS) = CGF.EmitARCStoreAutoreleasing(E);
     break;
 
+  case Qualifiers::OCL_ExplicitNone:
+    std::tie(LHS, RHS) = CGF.EmitARCStoreUnsafeUnretained(E, Ignore);
+    break;
+
   case Qualifiers::OCL_Weak:
     RHS = Visit(E->getRHS());
     LHS = EmitCheckedLValue(E->getLHS(), CodeGenFunction::TCK_Store);
     RHS = CGF.EmitARCStoreWeak(LHS.getAddress(), RHS, Ignore);
     break;
 
-  // No reason to do any of these differently.
   case Qualifiers::OCL_None:
-  case Qualifiers::OCL_ExplicitNone:
     // __block variables need to have the rhs evaluated first, plus
     // this should improve codegen just a little.
     RHS = Visit(E->getRHS());
@@ -3366,9 +3358,11 @@ Value *ScalarExprEmitter::VisitVAArgExpr(VAArgExpr *VE) {
 
   llvm::Type *ArgTy = ConvertType(VE->getType());
 
-  // If EmitVAArg fails, we fall back to the LLVM instruction.
-  if (!ArgPtr.isValid())
-    return Builder.CreateVAArg(ArgValue.getPointer(), ArgTy);
+  // If EmitVAArg fails, emit an error.
+  if (!ArgPtr.isValid()) {
+    CGF.ErrorUnsupported(VE, "va_arg expression");
+    return llvm::UndefValue::get(ArgTy);
+  }
 
   // FIXME Volatility.
   llvm::Value *Val = Builder.CreateLoad(ArgPtr);
@@ -3388,50 +3382,48 @@ Value *ScalarExprEmitter::VisitBlockExpr(const BlockExpr *block) {
   return CGF.EmitBlockLiteral(block);
 }
 
+// Convert a vec3 to vec4, or vice versa.
+static Value *ConvertVec3AndVec4(CGBuilderTy &Builder, CodeGenFunction &CGF,
+                                 Value *Src, unsigned NumElementsDst) {
+  llvm::Value *UnV = llvm::UndefValue::get(Src->getType());
+  SmallVector<llvm::Constant*, 4> Args;
+  Args.push_back(Builder.getInt32(0));
+  Args.push_back(Builder.getInt32(1));
+  Args.push_back(Builder.getInt32(2));
+  if (NumElementsDst == 4)
+    Args.push_back(llvm::UndefValue::get(CGF.Int32Ty));
+  llvm::Constant *Mask = llvm::ConstantVector::get(Args);
+  return Builder.CreateShuffleVector(Src, UnV, Mask);
+}
+
 Value *ScalarExprEmitter::VisitAsTypeExpr(AsTypeExpr *E) {
   Value *Src  = CGF.EmitScalarExpr(E->getSrcExpr());
   llvm::Type *DstTy = ConvertType(E->getType());
 
-  // Going from vec4->vec3 or vec3->vec4 is a special case and requires
-  // a shuffle vector instead of a bitcast.
   llvm::Type *SrcTy = Src->getType();
-  if (isa<llvm::VectorType>(DstTy) && isa<llvm::VectorType>(SrcTy)) {
-    unsigned numElementsDst = cast<llvm::VectorType>(DstTy)->getNumElements();
-    unsigned numElementsSrc = cast<llvm::VectorType>(SrcTy)->getNumElements();
-    if ((numElementsDst == 3 && numElementsSrc == 4)
-        || (numElementsDst == 4 && numElementsSrc == 3)) {
-
-
-      // In the case of going from int4->float3, a bitcast is needed before
-      // doing a shuffle.
-      llvm::Type *srcElemTy =
-      cast<llvm::VectorType>(SrcTy)->getElementType();
-      llvm::Type *dstElemTy =
-      cast<llvm::VectorType>(DstTy)->getElementType();
-
-      if ((srcElemTy->isIntegerTy() && dstElemTy->isFloatTy())
-          || (srcElemTy->isFloatTy() && dstElemTy->isIntegerTy())) {
-        // Create a float type of the same size as the source or destination.
-        llvm::VectorType *newSrcTy = llvm::VectorType::get(dstElemTy,
-                                                                 numElementsSrc);
-
-        Src = Builder.CreateBitCast(Src, newSrcTy, "astypeCast");
-      }
-
-      llvm::Value *UnV = llvm::UndefValue::get(Src->getType());
-
-      SmallVector<llvm::Constant*, 3> Args;
-      Args.push_back(Builder.getInt32(0));
-      Args.push_back(Builder.getInt32(1));
-      Args.push_back(Builder.getInt32(2));
-
-      if (numElementsDst == 4)
-        Args.push_back(llvm::UndefValue::get(CGF.Int32Ty));
-
-      llvm::Constant *Mask = llvm::ConstantVector::get(Args);
+  unsigned NumElementsSrc = isa<llvm::VectorType>(SrcTy) ?
+    cast<llvm::VectorType>(SrcTy)->getNumElements() : 0;
+  unsigned NumElementsDst = isa<llvm::VectorType>(DstTy) ?
+    cast<llvm::VectorType>(DstTy)->getNumElements() : 0;
+
+  // Going from vec3 to non-vec3 is a special case and requires a shuffle
+  // vector to get a vec4, then a bitcast if the target type is different.
+  if (NumElementsSrc == 3 && NumElementsDst != 3) {
+    Src = ConvertVec3AndVec4(Builder, CGF, Src, 4);
+    Src = Builder.CreateBitCast(Src, DstTy);
+    Src->setName("astype");
+    return Src;
+  }
 
-      return Builder.CreateShuffleVector(Src, UnV, Mask, "astype");
-    }
+  // Going from non-vec3 to vec3 is a special case and requires a bitcast
+  // to vec4 if the original type is not vec4, then a shuffle vector to
+  // get a vec3.
+  if (NumElementsSrc != 3 && NumElementsDst == 3) {
+    auto Vec4Ty = llvm::VectorType::get(DstTy->getVectorElementType(), 4);
+    Src = Builder.CreateBitCast(Src, Vec4Ty);
+    Src = ConvertVec3AndVec4(Builder, CGF, Src, 3);
+    Src->setName("astype");
+    return Src;
   }
 
   return Builder.CreateBitCast(Src, DstTy, "astype");
diff --git a/lib/CodeGen/CGLoopInfo.cpp b/lib/CodeGen/CGLoopInfo.cpp
index 0afe7dbb9f1d..51474f16a018 100644
--- a/lib/CodeGen/CGLoopInfo.cpp
+++ b/lib/CodeGen/CGLoopInfo.cpp
@@ -19,12 +19,15 @@
 using namespace clang::CodeGen;
 using namespace llvm;
 
-static MDNode *createMetadata(LLVMContext &Ctx, const LoopAttributes &Attrs) {
+static MDNode *createMetadata(LLVMContext &Ctx, const LoopAttributes &Attrs,
+                              llvm::DebugLoc Location) {
 
   if (!Attrs.IsParallel && Attrs.VectorizeWidth == 0 &&
       Attrs.InterleaveCount == 0 && Attrs.UnrollCount == 0 &&
       Attrs.VectorizeEnable == LoopAttributes::Unspecified &&
-      Attrs.UnrollEnable == LoopAttributes::Unspecified)
+      Attrs.UnrollEnable == LoopAttributes::Unspecified &&
+      Attrs.DistributeEnable == LoopAttributes::Unspecified &&
+      !Location)
     return nullptr;
 
   SmallVector<Metadata *, 4> Args;
@@ -32,6 +35,10 @@ static MDNode *createMetadata(LLVMContext &Ctx, const LoopAttributes &Attrs) {
   auto TempNode = MDNode::getTemporary(Ctx, None);
   Args.push_back(TempNode.get());
 
+  // If we have a valid debug location for the loop, add it.
+  if (Location)
+    Args.push_back(Location.getAsMDNode());
+
   // Setting vectorize.width
   if (Attrs.VectorizeWidth > 0) {
     Metadata *Vals[] = {MDString::get(Ctx, "llvm.loop.vectorize.width"),
@@ -78,6 +85,14 @@ static MDNode *createMetadata(LLVMContext &Ctx, const LoopAttributes &Attrs) {
     Args.push_back(MDNode::get(Ctx, Vals));
   }
 
+  if (Attrs.DistributeEnable != LoopAttributes::Unspecified) {
+    Metadata *Vals[] = {MDString::get(Ctx, "llvm.loop.distribute.enable"),
+                        ConstantAsMetadata::get(ConstantInt::get(
+                            Type::getInt1Ty(Ctx), (Attrs.DistributeEnable ==
+                                                   LoopAttributes::Enable)))};
+    Args.push_back(MDNode::get(Ctx, Vals));
+  }
+
   // Set the first operand to itself.
   MDNode *LoopID = MDNode::get(Ctx, Args);
   LoopID->replaceOperandWith(0, LoopID);
@@ -87,7 +102,8 @@ static MDNode *createMetadata(LLVMContext &Ctx, const LoopAttributes &Attrs) {
 LoopAttributes::LoopAttributes(bool IsParallel)
     : IsParallel(IsParallel), VectorizeEnable(LoopAttributes::Unspecified),
       UnrollEnable(LoopAttributes::Unspecified), VectorizeWidth(0),
-      InterleaveCount(0), UnrollCount(0) {}
+      InterleaveCount(0), UnrollCount(0),
+      DistributeEnable(LoopAttributes::Unspecified) {}
 
 void LoopAttributes::clear() {
   IsParallel = false;
@@ -98,37 +114,60 @@ void LoopAttributes::clear() {
   UnrollEnable = LoopAttributes::Unspecified;
 }
 
-LoopInfo::LoopInfo(BasicBlock *Header, const LoopAttributes &Attrs)
+LoopInfo::LoopInfo(BasicBlock *Header, const LoopAttributes &Attrs,
+                   llvm::DebugLoc Location)
     : LoopID(nullptr), Header(Header), Attrs(Attrs) {
-  LoopID = createMetadata(Header->getContext(), Attrs);
+  LoopID = createMetadata(Header->getContext(), Attrs, Location);
 }
 
-void LoopInfoStack::push(BasicBlock *Header) {
-  Active.push_back(LoopInfo(Header, StagedAttrs));
+void LoopInfoStack::push(BasicBlock *Header, llvm::DebugLoc Location) {
+  Active.push_back(LoopInfo(Header, StagedAttrs, Location));
   // Clear the attributes so nested loops do not inherit them.
   StagedAttrs.clear();
 }
 
 void LoopInfoStack::push(BasicBlock *Header, clang::ASTContext &Ctx,
-                         ArrayRef<const clang::Attr *> Attrs) {
+                         ArrayRef<const clang::Attr *> Attrs,
+                         llvm::DebugLoc Location) {
 
   // Identify loop hint attributes from Attrs.
   for (const auto *Attr : Attrs) {
     const LoopHintAttr *LH = dyn_cast<LoopHintAttr>(Attr);
+    const OpenCLUnrollHintAttr *OpenCLHint =
+        dyn_cast<OpenCLUnrollHintAttr>(Attr);
 
     // Skip non loop hint attributes
-    if (!LH)
+    if (!LH && !OpenCLHint) {
       continue;
+    }
 
-    auto *ValueExpr = LH->getValue();
+    LoopHintAttr::OptionType Option = LoopHintAttr::Unroll;
+    LoopHintAttr::LoopHintState State = LoopHintAttr::Disable;
     unsigned ValueInt = 1;
-    if (ValueExpr) {
-      llvm::APSInt ValueAPS = ValueExpr->EvaluateKnownConstInt(Ctx);
-      ValueInt = ValueAPS.getSExtValue();
-    }
+    // Translate opencl_unroll_hint attribute argument to
+    // equivalent LoopHintAttr enums.
+    // OpenCL v2.0 s6.11.5:  
+    // 0 - full unroll (no argument).
+    // 1 - disable unroll.
+    // other positive integer n - unroll by n.
+    if (OpenCLHint) {
+      ValueInt = OpenCLHint->getUnrollHint();
+      if (ValueInt == 0) {
+        State = LoopHintAttr::Full;
+      } else if (ValueInt != 1) {
+        Option = LoopHintAttr::UnrollCount;
+        State = LoopHintAttr::Numeric;
+      }
+    } else if (LH) {
+      auto *ValueExpr = LH->getValue();
+      if (ValueExpr) {
+        llvm::APSInt ValueAPS = ValueExpr->EvaluateKnownConstInt(Ctx);
+        ValueInt = ValueAPS.getSExtValue();
+      }
 
-    LoopHintAttr::OptionType Option = LH->getOption();
-    LoopHintAttr::LoopHintState State = LH->getState();
+      Option = LH->getOption();
+      State = LH->getState();
+    }
     switch (State) {
     case LoopHintAttr::Disable:
       switch (Option) {
@@ -143,6 +182,9 @@ void LoopInfoStack::push(BasicBlock *Header, clang::ASTContext &Ctx,
       case LoopHintAttr::Unroll:
         setUnrollState(LoopAttributes::Disable);
         break;
+      case LoopHintAttr::Distribute:
+        setDistributeState(false);
+        break;
       case LoopHintAttr::UnrollCount:
       case LoopHintAttr::VectorizeWidth:
       case LoopHintAttr::InterleaveCount:
@@ -159,6 +201,9 @@ void LoopInfoStack::push(BasicBlock *Header, clang::ASTContext &Ctx,
       case LoopHintAttr::Unroll:
         setUnrollState(LoopAttributes::Enable);
         break;
+      case LoopHintAttr::Distribute:
+        setDistributeState(true);
+        break;
       case LoopHintAttr::UnrollCount:
       case LoopHintAttr::VectorizeWidth:
       case LoopHintAttr::InterleaveCount:
@@ -178,6 +223,7 @@ void LoopInfoStack::push(BasicBlock *Header, clang::ASTContext &Ctx,
       case LoopHintAttr::UnrollCount:
       case LoopHintAttr::VectorizeWidth:
       case LoopHintAttr::InterleaveCount:
+      case LoopHintAttr::Distribute:
         llvm_unreachable("Options cannot be used to assume mem safety.");
         break;
       }
@@ -192,6 +238,7 @@ void LoopInfoStack::push(BasicBlock *Header, clang::ASTContext &Ctx,
       case LoopHintAttr::UnrollCount:
       case LoopHintAttr::VectorizeWidth:
       case LoopHintAttr::InterleaveCount:
+      case LoopHintAttr::Distribute:
         llvm_unreachable("Options cannot be used with 'full' hint.");
         break;
       }
@@ -210,6 +257,7 @@ void LoopInfoStack::push(BasicBlock *Header, clang::ASTContext &Ctx,
       case LoopHintAttr::Unroll:
       case LoopHintAttr::Vectorize:
       case LoopHintAttr::Interleave:
+      case LoopHintAttr::Distribute:
         llvm_unreachable("Options cannot be assigned a value.");
         break;
       }
@@ -218,7 +266,7 @@ void LoopInfoStack::push(BasicBlock *Header, clang::ASTContext &Ctx,
   }
 
   /// Stage the attributes.
-  push(Header);
+  push(Header, Location);
 }
 
 void LoopInfoStack::pop() {
@@ -237,7 +285,7 @@ void LoopInfoStack::InsertHelper(Instruction *I) const {
   if (TerminatorInst *TI = dyn_cast<TerminatorInst>(I)) {
     for (unsigned i = 0, ie = TI->getNumSuccessors(); i < ie; ++i)
       if (TI->getSuccessor(i) == L.getHeader()) {
-        TI->setMetadata("llvm.loop", L.getLoopID());
+        TI->setMetadata(llvm::LLVMContext::MD_loop, L.getLoopID());
         break;
       }
     return;
diff --git a/lib/CodeGen/CGLoopInfo.h b/lib/CodeGen/CGLoopInfo.h
index ec3390677fa9..a0111edde5de 100644
--- a/lib/CodeGen/CGLoopInfo.h
+++ b/lib/CodeGen/CGLoopInfo.h
@@ -18,6 +18,7 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/IR/DebugLoc.h"
 #include "llvm/IR/Value.h"
 #include "llvm/Support/Compiler.h"
 
@@ -57,13 +58,17 @@ struct LoopAttributes {
 
   /// \brief llvm.unroll.
   unsigned UnrollCount;
+
+  /// \brief Value for llvm.loop.distribute.enable metadata.
+  LVEnableState DistributeEnable;
 };
 
 /// \brief Information used when generating a structured loop.
 class LoopInfo {
 public:
   /// \brief Construct a new LoopInfo for the loop with entry Header.
-  LoopInfo(llvm::BasicBlock *Header, const LoopAttributes &Attrs);
+  LoopInfo(llvm::BasicBlock *Header, const LoopAttributes &Attrs,
+           llvm::DebugLoc Location);
 
   /// \brief Get the loop id metadata for this loop.
   llvm::MDNode *getLoopID() const { return LoopID; }
@@ -95,12 +100,14 @@ public:
 
   /// \brief Begin a new structured loop. The set of staged attributes will be
   /// applied to the loop and then cleared.
-  void push(llvm::BasicBlock *Header);
+  void push(llvm::BasicBlock *Header,
+            llvm::DebugLoc Location = llvm::DebugLoc());
 
   /// \brief Begin a new structured loop. Stage attributes from the Attrs list.
   /// The staged attributes are applied to the loop and then cleared.
   void push(llvm::BasicBlock *Header, clang::ASTContext &Ctx,
-            llvm::ArrayRef<const Attr *> Attrs);
+            llvm::ArrayRef<const Attr *> Attrs,
+            llvm::DebugLoc Location = llvm::DebugLoc());
 
   /// \brief End the current loop.
   void pop();
@@ -126,6 +133,12 @@ public:
         Enable ? LoopAttributes::Enable : LoopAttributes::Disable;
   }
 
+  /// \brief Set the next pushed loop as a distribution candidate.
+  void setDistributeState(bool Enable = true) {
+    StagedAttrs.DistributeEnable =
+        Enable ? LoopAttributes::Enable : LoopAttributes::Disable;
+  }
+
   /// \brief Set the next pushed loop unroll state.
   void setUnrollState(const LoopAttributes::LVEnableState &State) {
     StagedAttrs.UnrollEnable = State;
diff --git a/lib/CodeGen/CGObjC.cpp b/lib/CodeGen/CGObjC.cpp
index 2d5991b71fca..db894ce67470 100644
--- a/lib/CodeGen/CGObjC.cpp
+++ b/lib/CodeGen/CGObjC.cpp
@@ -590,9 +590,7 @@ static void emitStructGetterCall(CodeGenFunction &CGF, ObjCIvarDecl *ivar,
   args.add(RValue::get(CGF.Builder.getInt1(hasStrong)), Context.BoolTy);
 
   llvm::Value *fn = CGF.CGM.getObjCRuntime().GetGetStructFunction();
-  CGF.EmitCall(CGF.getTypes().arrangeFreeFunctionCall(Context.VoidTy, args,
-                                                      FunctionType::ExtInfo(),
-                                                      RequiredArgs::All),
+  CGF.EmitCall(CGF.getTypes().arrangeBuiltinFunctionCall(Context.VoidTy, args),
                fn, ReturnValueSlot(), args);
 }
 
@@ -856,10 +854,8 @@ static void emitCPPObjectAtomicGetterCall(CodeGenFunction &CGF,
   
   llvm::Value *copyCppAtomicObjectFn = 
     CGF.CGM.getObjCRuntime().GetCppAtomicObjectGetFunction();
-  CGF.EmitCall(CGF.getTypes().arrangeFreeFunctionCall(CGF.getContext().VoidTy,
-                                                      args,
-                                                      FunctionType::ExtInfo(),
-                                                      RequiredArgs::All),
+  CGF.EmitCall(
+      CGF.getTypes().arrangeBuiltinFunctionCall(CGF.getContext().VoidTy, args),
                copyCppAtomicObjectFn, ReturnValueSlot(), args);
 }
 
@@ -901,21 +897,29 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl,
 
     // Currently, all atomic accesses have to be through integer
     // types, so there's no point in trying to pick a prettier type.
-    llvm::Type *bitcastType =
-      llvm::Type::getIntNTy(getLLVMContext(),
-                            getContext().toBits(strategy.getIvarSize()));
+    uint64_t ivarSize = getContext().toBits(strategy.getIvarSize());
+    llvm::Type *bitcastType = llvm::Type::getIntNTy(getLLVMContext(), ivarSize);
     bitcastType = bitcastType->getPointerTo(); // addrspace 0 okay
 
     // Perform an atomic load.  This does not impose ordering constraints.
     Address ivarAddr = LV.getAddress();
     ivarAddr = Builder.CreateBitCast(ivarAddr, bitcastType);
     llvm::LoadInst *load = Builder.CreateLoad(ivarAddr, "load");
-    load->setAtomic(llvm::Unordered);
+    load->setAtomic(llvm::AtomicOrdering::Unordered);
 
     // Store that value into the return address.  Doing this with a
     // bitcast is likely to produce some pretty ugly IR, but it's not
     // the *most* terrible thing in the world.
-    Builder.CreateStore(load, Builder.CreateBitCast(ReturnValue, bitcastType));
+    llvm::Type *retTy = ConvertType(getterMethod->getReturnType());
+    uint64_t retTySize = CGM.getDataLayout().getTypeSizeInBits(retTy);
+    llvm::Value *ivarVal = load;
+    if (ivarSize > retTySize) {
+      llvm::Type *newTy = llvm::Type::getIntNTy(getLLVMContext(), retTySize);
+      ivarVal = Builder.CreateTrunc(load, newTy);
+      bitcastType = newTy->getPointerTo();
+    }
+    Builder.CreateStore(ivarVal,
+                        Builder.CreateBitCast(ReturnValue, bitcastType));
 
     // Make sure we don't do an autorelease.
     AutoreleaseResult = false;
@@ -950,8 +954,7 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl,
     // runtime already should have computed it to build the function.
     llvm::Instruction *CallInstruction;
     RValue RV = EmitCall(
-        getTypes().arrangeFreeFunctionCall(
-            propType, args, FunctionType::ExtInfo(), RequiredArgs::All),
+        getTypes().arrangeBuiltinFunctionCall(propType, args),
         getPropertyFn, ReturnValueSlot(), args, CGCalleeInfo(),
         &CallInstruction);
     if (llvm::CallInst *call = dyn_cast<llvm::CallInst>(CallInstruction))
@@ -1015,7 +1018,6 @@ CodeGenFunction::generateObjCGetterBody(const ObjCImplementationDecl *classImpl,
           AutoreleaseResult = false;
         }
 
-        value = Builder.CreateBitCast(value, ConvertType(propType));
         value = Builder.CreateBitCast(
             value, ConvertType(GetterMethodDecl->getReturnType()));
       }
@@ -1067,10 +1069,8 @@ static void emitStructSetterCall(CodeGenFunction &CGF, ObjCMethodDecl *OMD,
   args.add(RValue::get(CGF.Builder.getFalse()), CGF.getContext().BoolTy);
 
   llvm::Value *copyStructFn = CGF.CGM.getObjCRuntime().GetSetStructFunction();
-  CGF.EmitCall(CGF.getTypes().arrangeFreeFunctionCall(CGF.getContext().VoidTy,
-                                                      args,
-                                                      FunctionType::ExtInfo(),
-                                                      RequiredArgs::All),
+  CGF.EmitCall(
+      CGF.getTypes().arrangeBuiltinFunctionCall(CGF.getContext().VoidTy, args),
                copyStructFn, ReturnValueSlot(), args);
 }
 
@@ -1105,10 +1105,8 @@ static void emitCPPObjectAtomicSetterCall(CodeGenFunction &CGF,
   
   llvm::Value *copyCppAtomicObjectFn = 
     CGF.CGM.getObjCRuntime().GetCppAtomicObjectSetFunction();
-  CGF.EmitCall(CGF.getTypes().arrangeFreeFunctionCall(CGF.getContext().VoidTy,
-                                                      args,
-                                                      FunctionType::ExtInfo(),
-                                                      RequiredArgs::All),
+  CGF.EmitCall(
+      CGF.getTypes().arrangeBuiltinFunctionCall(CGF.getContext().VoidTy, args),
                copyCppAtomicObjectFn, ReturnValueSlot(), args);
 }
 
@@ -1192,7 +1190,7 @@ CodeGenFunction::generateObjCSetterBody(const ObjCImplementationDecl *classImpl,
 
     // Perform an atomic store.  There are no memory ordering requirements.
     llvm::StoreInst *store = Builder.CreateStore(load, ivarAddr);
-    store->setAtomic(llvm::Unordered);
+    store->setAtomic(llvm::AtomicOrdering::Unordered);
     return;
   }
 
@@ -1238,9 +1236,7 @@ CodeGenFunction::generateObjCSetterBody(const ObjCImplementationDecl *classImpl,
     if (setOptimizedPropertyFn) {
       args.add(RValue::get(arg), getContext().getObjCIdType());
       args.add(RValue::get(ivarOffset), getContext().getPointerDiffType());
-      EmitCall(getTypes().arrangeFreeFunctionCall(getContext().VoidTy, args,
-                                                  FunctionType::ExtInfo(),
-                                                  RequiredArgs::All),
+      EmitCall(getTypes().arrangeBuiltinFunctionCall(getContext().VoidTy, args),
                setOptimizedPropertyFn, ReturnValueSlot(), args);
     } else {
       args.add(RValue::get(ivarOffset), getContext().getPointerDiffType());
@@ -1251,9 +1247,7 @@ CodeGenFunction::generateObjCSetterBody(const ObjCImplementationDecl *classImpl,
                getContext().BoolTy);
       // FIXME: We shouldn't need to get the function info here, the runtime
       // already should have computed it to build the function.
-      EmitCall(getTypes().arrangeFreeFunctionCall(getContext().VoidTy, args,
-                                                  FunctionType::ExtInfo(),
-                                                  RequiredArgs::All),
+      EmitCall(getTypes().arrangeBuiltinFunctionCall(getContext().VoidTy, args),
                setPropertyFn, ReturnValueSlot(), args);
     }
     
@@ -1498,6 +1492,8 @@ void CodeGenFunction::EmitObjCForCollectionStmt(const ObjCForCollectionStmt &S){
                                       ArrayType::Normal, 0);
   Address ItemsPtr = CreateMemTemp(ItemsTy, "items.ptr");
 
+  RunCleanupsScope ForScope(*this);
+
   // Emit the collection pointer.  In ARC, we do a retain.
   llvm::Value *Collection;
   if (getLangOpts().ObjCAutoRefCount) {
@@ -1610,9 +1606,8 @@ void CodeGenFunction::EmitObjCForCollectionStmt(const ObjCForCollectionStmt &S){
   Args2.add(RValue::get(V), getContext().getObjCIdType());
   // FIXME: We shouldn't need to get the function info here, the runtime already
   // should have computed it to build the function.
-  EmitCall(CGM.getTypes().arrangeFreeFunctionCall(getContext().VoidTy, Args2,
-                                                  FunctionType::ExtInfo(),
-                                                  RequiredArgs::All),
+  EmitCall(
+          CGM.getTypes().arrangeBuiltinFunctionCall(getContext().VoidTy, Args2),
            EnumerationMutationFn, ReturnValueSlot(), Args2);
 
   // Otherwise, or if the mutation function returns, just continue.
@@ -1739,10 +1734,7 @@ void CodeGenFunction::EmitObjCForCollectionStmt(const ObjCForCollectionStmt &S){
   if (DI)
     DI->EmitLexicalBlockEnd(Builder, S.getSourceRange().getEnd());
 
-  // Leave the cleanup we entered in ARC.
-  if (getLangOpts().ObjCAutoRefCount)
-    PopCleanupBlock();
-
+  ForScope.ForceCleanup();
   EmitBlock(LoopEnd.getBlock());
 }
 
@@ -1980,20 +1972,14 @@ llvm::Value *CodeGenFunction::EmitARCRetainBlock(llvm::Value *value,
   return result;
 }
 
-/// Retain the given object which is the result of a function call.
-///   call i8* \@objc_retainAutoreleasedReturnValue(i8* %value)
-///
-/// Yes, this function name is one character away from a different
-/// call with completely different semantics.
-llvm::Value *
-CodeGenFunction::EmitARCRetainAutoreleasedReturnValue(llvm::Value *value) {
+static void emitAutoreleasedReturnValueMarker(CodeGenFunction &CGF) {
   // Fetch the void(void) inline asm which marks that we're going to
-  // retain the autoreleased return value.
+  // do something with the autoreleased return value.
   llvm::InlineAsm *&marker
-    = CGM.getObjCEntrypoints().retainAutoreleasedReturnValueMarker;
+    = CGF.CGM.getObjCEntrypoints().retainAutoreleasedReturnValueMarker;
   if (!marker) {
     StringRef assembly
-      = CGM.getTargetCodeGenInfo()
+      = CGF.CGM.getTargetCodeGenInfo()
            .getARCRetainAutoreleasedReturnValueMarker();
 
     // If we have an empty assembly string, there's nothing to do.
@@ -2001,9 +1987,9 @@ CodeGenFunction::EmitARCRetainAutoreleasedReturnValue(llvm::Value *value) {
 
     // Otherwise, at -O0, build an inline asm that we're going to call
     // in a moment.
-    } else if (CGM.getCodeGenOpts().OptimizationLevel == 0) {
+    } else if (CGF.CGM.getCodeGenOpts().OptimizationLevel == 0) {
       llvm::FunctionType *type =
-        llvm::FunctionType::get(VoidTy, /*variadic*/false);
+        llvm::FunctionType::get(CGF.VoidTy, /*variadic*/false);
       
       marker = llvm::InlineAsm::get(type, assembly, "", /*sideeffects*/ true);
 
@@ -2012,25 +1998,50 @@ CodeGenFunction::EmitARCRetainAutoreleasedReturnValue(llvm::Value *value) {
     // optimizer to pick up.
     } else {
       llvm::NamedMDNode *metadata =
-        CGM.getModule().getOrInsertNamedMetadata(
+        CGF.CGM.getModule().getOrInsertNamedMetadata(
                             "clang.arc.retainAutoreleasedReturnValueMarker");
       assert(metadata->getNumOperands() <= 1);
       if (metadata->getNumOperands() == 0) {
-        metadata->addOperand(llvm::MDNode::get(
-            getLLVMContext(), llvm::MDString::get(getLLVMContext(), assembly)));
+        auto &ctx = CGF.getLLVMContext();
+        metadata->addOperand(llvm::MDNode::get(ctx,
+                                     llvm::MDString::get(ctx, assembly)));
       }
     }
   }
 
   // Call the marker asm if we made one, which we do only at -O0.
   if (marker)
-    Builder.CreateCall(marker);
+    CGF.Builder.CreateCall(marker);
+}
 
+/// Retain the given object which is the result of a function call.
+///   call i8* \@objc_retainAutoreleasedReturnValue(i8* %value)
+///
+/// Yes, this function name is one character away from a different
+/// call with completely different semantics.
+llvm::Value *
+CodeGenFunction::EmitARCRetainAutoreleasedReturnValue(llvm::Value *value) {
+  emitAutoreleasedReturnValueMarker(*this);
   return emitARCValueOperation(*this, value,
-                     CGM.getObjCEntrypoints().objc_retainAutoreleasedReturnValue,
+              CGM.getObjCEntrypoints().objc_retainAutoreleasedReturnValue,
                                "objc_retainAutoreleasedReturnValue");
 }
 
+/// Claim a possibly-autoreleased return value at +0.  This is only
+/// valid to do in contexts which do not rely on the retain to keep
+/// the object valid for for all of its uses; for example, when
+/// the value is ignored, or when it is being assigned to an
+/// __unsafe_unretained variable.
+///
+///   call i8* \@objc_unsafeClaimAutoreleasedReturnValue(i8* %value)
+llvm::Value *
+CodeGenFunction::EmitARCUnsafeClaimAutoreleasedReturnValue(llvm::Value *value) {
+  emitAutoreleasedReturnValueMarker(*this);
+  return emitARCValueOperation(*this, value,
+              CGM.getObjCEntrypoints().objc_unsafeClaimAutoreleasedReturnValue,
+                               "objc_unsafeClaimAutoreleasedReturnValue");
+}
+
 /// Release the given object.
 ///   call void \@objc_release(i8* %value)
 void CodeGenFunction::EmitARCRelease(llvm::Value *value,
@@ -2446,25 +2457,22 @@ static TryEmitResult tryEmitARCRetainLoadOfScalar(CodeGenFunction &CGF,
   return tryEmitARCRetainLoadOfScalar(CGF, CGF.EmitLValue(e), type);
 }
 
-static llvm::Value *emitARCRetainAfterCall(CodeGenFunction &CGF,
-                                           llvm::Value *value);
+typedef llvm::function_ref<llvm::Value *(CodeGenFunction &CGF,
+                                         llvm::Value *value)>
+  ValueTransform;
 
-/// Given that the given expression is some sort of call (which does
-/// not return retained), emit a retain following it.
-static llvm::Value *emitARCRetainCall(CodeGenFunction &CGF, const Expr *e) {
-  llvm::Value *value = CGF.EmitScalarExpr(e);
-  return emitARCRetainAfterCall(CGF, value);
-}
-
-static llvm::Value *emitARCRetainAfterCall(CodeGenFunction &CGF,
-                                           llvm::Value *value) {
+/// Insert code immediately after a call.
+static llvm::Value *emitARCOperationAfterCall(CodeGenFunction &CGF,
+                                              llvm::Value *value,
+                                              ValueTransform doAfterCall,
+                                              ValueTransform doFallback) {
   if (llvm::CallInst *call = dyn_cast<llvm::CallInst>(value)) {
     CGBuilderTy::InsertPoint ip = CGF.Builder.saveIP();
 
     // Place the retain immediately following the call.
     CGF.Builder.SetInsertPoint(call->getParent(),
                                ++llvm::BasicBlock::iterator(call));
-    value = CGF.EmitARCRetainAutoreleasedReturnValue(value);
+    value = doAfterCall(CGF, value);
 
     CGF.Builder.restoreIP(ip);
     return value;
@@ -2474,7 +2482,7 @@ static llvm::Value *emitARCRetainAfterCall(CodeGenFunction &CGF,
     // Place the retain at the beginning of the normal destination block.
     llvm::BasicBlock *BB = invoke->getNormalDest();
     CGF.Builder.SetInsertPoint(BB, BB->begin());
-    value = CGF.EmitARCRetainAutoreleasedReturnValue(value);
+    value = doAfterCall(CGF, value);
 
     CGF.Builder.restoreIP(ip);
     return value;
@@ -2483,7 +2491,7 @@ static llvm::Value *emitARCRetainAfterCall(CodeGenFunction &CGF,
   // the operand.
   } else if (llvm::BitCastInst *bitcast = dyn_cast<llvm::BitCastInst>(value)) {
     llvm::Value *operand = bitcast->getOperand(0);
-    operand = emitARCRetainAfterCall(CGF, operand);
+    operand = emitARCOperationAfterCall(CGF, operand, doAfterCall, doFallback);
     bitcast->setOperand(0, operand);
     return bitcast;
 
@@ -2491,7 +2499,46 @@ static llvm::Value *emitARCRetainAfterCall(CodeGenFunction &CGF,
   } else {
     // Retain using the non-block variant: we never need to do a copy
     // of a block that's been returned to us.
-    return CGF.EmitARCRetainNonBlock(value);
+    return doFallback(CGF, value);
+  }
+}
+
+/// Given that the given expression is some sort of call (which does
+/// not return retained), emit a retain following it.
+static llvm::Value *emitARCRetainCallResult(CodeGenFunction &CGF,
+                                            const Expr *e) {
+  llvm::Value *value = CGF.EmitScalarExpr(e);
+  return emitARCOperationAfterCall(CGF, value,
+           [](CodeGenFunction &CGF, llvm::Value *value) {
+             return CGF.EmitARCRetainAutoreleasedReturnValue(value);
+           },
+           [](CodeGenFunction &CGF, llvm::Value *value) {
+             return CGF.EmitARCRetainNonBlock(value);
+           });
+}
+
+/// Given that the given expression is some sort of call (which does
+/// not return retained), perform an unsafeClaim following it.
+static llvm::Value *emitARCUnsafeClaimCallResult(CodeGenFunction &CGF,
+                                                 const Expr *e) {
+  llvm::Value *value = CGF.EmitScalarExpr(e);
+  return emitARCOperationAfterCall(CGF, value,
+           [](CodeGenFunction &CGF, llvm::Value *value) {
+             return CGF.EmitARCUnsafeClaimAutoreleasedReturnValue(value);
+           },
+           [](CodeGenFunction &CGF, llvm::Value *value) {
+             return value;
+           });
+}
+
+llvm::Value *CodeGenFunction::EmitARCReclaimReturnedObject(const Expr *E,
+                                                      bool allowUnsafeClaim) {
+  if (allowUnsafeClaim &&
+      CGM.getLangOpts().ObjCRuntime.hasARCUnsafeClaimAutoreleasedReturnValue()) {
+    return emitARCUnsafeClaimCallResult(*this, E);
+  } else {
+    llvm::Value *value = emitARCRetainCallResult(*this, E);
+    return EmitObjCConsumeObject(E->getType(), value);
   }
 }
 
@@ -2531,17 +2578,52 @@ static bool shouldEmitSeparateBlockRetain(const Expr *e) {
   return true;
 }
 
-/// Try to emit a PseudoObjectExpr at +1.
+namespace {
+/// A CRTP base class for emitting expressions of retainable object
+/// pointer type in ARC.
+template <typename Impl, typename Result> class ARCExprEmitter {
+protected:
+  CodeGenFunction &CGF;
+  Impl &asImpl() { return *static_cast<Impl*>(this); }
+
+  ARCExprEmitter(CodeGenFunction &CGF) : CGF(CGF) {}
+
+public:
+  Result visit(const Expr *e);
+  Result visitCastExpr(const CastExpr *e);
+  Result visitPseudoObjectExpr(const PseudoObjectExpr *e);
+  Result visitBinaryOperator(const BinaryOperator *e);
+  Result visitBinAssign(const BinaryOperator *e);
+  Result visitBinAssignUnsafeUnretained(const BinaryOperator *e);
+  Result visitBinAssignAutoreleasing(const BinaryOperator *e);
+  Result visitBinAssignWeak(const BinaryOperator *e);
+  Result visitBinAssignStrong(const BinaryOperator *e);
+
+  // Minimal implementation:
+  //   Result visitLValueToRValue(const Expr *e)
+  //   Result visitConsumeObject(const Expr *e)
+  //   Result visitExtendBlockObject(const Expr *e)
+  //   Result visitReclaimReturnedObject(const Expr *e)
+  //   Result visitCall(const Expr *e)
+  //   Result visitExpr(const Expr *e)
+  //
+  //   Result emitBitCast(Result result, llvm::Type *resultType)
+  //   llvm::Value *getValueOfResult(Result result)
+};
+}
+
+/// Try to emit a PseudoObjectExpr under special ARC rules.
 ///
 /// This massively duplicates emitPseudoObjectRValue.
-static TryEmitResult tryEmitARCRetainPseudoObject(CodeGenFunction &CGF,
-                                                  const PseudoObjectExpr *E) {
+template <typename Impl, typename Result>
+Result
+ARCExprEmitter<Impl,Result>::visitPseudoObjectExpr(const PseudoObjectExpr *E) {
   SmallVector<CodeGenFunction::OpaqueValueMappingData, 4> opaques;
 
   // Find the result expression.
   const Expr *resultExpr = E->getResultExpr();
   assert(resultExpr);
-  TryEmitResult result;
+  Result result;
 
   for (PseudoObjectExpr::const_semantics_iterator
          i = E->semantics_begin(), e = E->semantics_end(); i != e; ++i) {
@@ -2557,8 +2639,9 @@ static TryEmitResult tryEmitARCRetainPseudoObject(CodeGenFunction &CGF,
       // expression, try to evaluate the source as +1.
       if (ov == resultExpr) {
         assert(!OVMA::shouldBindAsLValue(ov));
-        result = tryEmitARCRetainScalarExpr(CGF, ov->getSourceExpr());
-        opaqueData = OVMA::bind(CGF, ov, RValue::get(result.getPointer()));
+        result = asImpl().visit(ov->getSourceExpr());
+        opaqueData = OVMA::bind(CGF, ov,
+                            RValue::get(asImpl().getValueOfResult(result)));
 
       // Otherwise, just bind it.
       } else {
@@ -2569,7 +2652,7 @@ static TryEmitResult tryEmitARCRetainPseudoObject(CodeGenFunction &CGF,
     // Otherwise, if the expression is the result, evaluate it
     // and remember the result.
     } else if (semantic == resultExpr) {
-      result = tryEmitARCRetainScalarExpr(CGF, semantic);
+      result = asImpl().visit(semantic);
 
     // Otherwise, evaluate the expression in an ignored context.
     } else {
@@ -2584,146 +2667,240 @@ static TryEmitResult tryEmitARCRetainPseudoObject(CodeGenFunction &CGF,
   return result;
 }
 
-static TryEmitResult
-tryEmitARCRetainScalarExpr(CodeGenFunction &CGF, const Expr *e) {
+template <typename Impl, typename Result>
+Result ARCExprEmitter<Impl,Result>::visitCastExpr(const CastExpr *e) {
+  switch (e->getCastKind()) {
+
+  // No-op casts don't change the type, so we just ignore them.
+  case CK_NoOp:
+    return asImpl().visit(e->getSubExpr());
+
+  // These casts can change the type.
+  case CK_CPointerToObjCPointerCast:
+  case CK_BlockPointerToObjCPointerCast:
+  case CK_AnyPointerToBlockPointerCast:
+  case CK_BitCast: {
+    llvm::Type *resultType = CGF.ConvertType(e->getType());
+    assert(e->getSubExpr()->getType()->hasPointerRepresentation());
+    Result result = asImpl().visit(e->getSubExpr());
+    return asImpl().emitBitCast(result, resultType);
+  }
+
+  // Handle some casts specially.
+  case CK_LValueToRValue:
+    return asImpl().visitLValueToRValue(e->getSubExpr());
+  case CK_ARCConsumeObject:
+    return asImpl().visitConsumeObject(e->getSubExpr());
+  case CK_ARCExtendBlockObject:
+    return asImpl().visitExtendBlockObject(e->getSubExpr());
+  case CK_ARCReclaimReturnedObject:
+    return asImpl().visitReclaimReturnedObject(e->getSubExpr());
+
+  // Otherwise, use the default logic.
+  default:
+    return asImpl().visitExpr(e);
+  }
+}
+
+template <typename Impl, typename Result>
+Result
+ARCExprEmitter<Impl,Result>::visitBinaryOperator(const BinaryOperator *e) {
+  switch (e->getOpcode()) {
+  case BO_Comma:
+    CGF.EmitIgnoredExpr(e->getLHS());
+    CGF.EnsureInsertPoint();
+    return asImpl().visit(e->getRHS());
+
+  case BO_Assign:
+    return asImpl().visitBinAssign(e);
+
+  default:
+    return asImpl().visitExpr(e);
+  }
+}
+
+template <typename Impl, typename Result>
+Result ARCExprEmitter<Impl,Result>::visitBinAssign(const BinaryOperator *e) {
+  switch (e->getLHS()->getType().getObjCLifetime()) {
+  case Qualifiers::OCL_ExplicitNone:
+    return asImpl().visitBinAssignUnsafeUnretained(e);
+
+  case Qualifiers::OCL_Weak:
+    return asImpl().visitBinAssignWeak(e);
+
+  case Qualifiers::OCL_Autoreleasing:
+    return asImpl().visitBinAssignAutoreleasing(e);
+
+  case Qualifiers::OCL_Strong:
+    return asImpl().visitBinAssignStrong(e);
+
+  case Qualifiers::OCL_None:
+    return asImpl().visitExpr(e);
+  }
+  llvm_unreachable("bad ObjC ownership qualifier");
+}
+
+/// The default rule for __unsafe_unretained emits the RHS recursively,
+/// stores into the unsafe variable, and propagates the result outward.
+template <typename Impl, typename Result>
+Result ARCExprEmitter<Impl,Result>::
+                    visitBinAssignUnsafeUnretained(const BinaryOperator *e) {
+  // Recursively emit the RHS.
+  // For __block safety, do this before emitting the LHS.
+  Result result = asImpl().visit(e->getRHS());
+
+  // Perform the store.
+  LValue lvalue =
+    CGF.EmitCheckedLValue(e->getLHS(), CodeGenFunction::TCK_Store);
+  CGF.EmitStoreThroughLValue(RValue::get(asImpl().getValueOfResult(result)),
+                             lvalue);
+
+  return result;
+}
+
+template <typename Impl, typename Result>
+Result
+ARCExprEmitter<Impl,Result>::visitBinAssignAutoreleasing(const BinaryOperator *e) {
+  return asImpl().visitExpr(e);
+}
+
+template <typename Impl, typename Result>
+Result
+ARCExprEmitter<Impl,Result>::visitBinAssignWeak(const BinaryOperator *e) {
+  return asImpl().visitExpr(e);
+}
+
+template <typename Impl, typename Result>
+Result
+ARCExprEmitter<Impl,Result>::visitBinAssignStrong(const BinaryOperator *e) {
+  return asImpl().visitExpr(e);
+}
+
+/// The general expression-emission logic.
+template <typename Impl, typename Result>
+Result ARCExprEmitter<Impl,Result>::visit(const Expr *e) {
   // We should *never* see a nested full-expression here, because if
   // we fail to emit at +1, our caller must not retain after we close
-  // out the full-expression.
+  // out the full-expression.  This isn't as important in the unsafe
+  // emitter.
   assert(!isa<ExprWithCleanups>(e));
 
-  // The desired result type, if it differs from the type of the
-  // ultimate opaque expression.
-  llvm::Type *resultType = nullptr;
-
-  while (true) {
-    e = e->IgnoreParens();
-
-    // There's a break at the end of this if-chain;  anything
-    // that wants to keep looping has to explicitly continue.
-    if (const CastExpr *ce = dyn_cast<CastExpr>(e)) {
-      switch (ce->getCastKind()) {
-      // No-op casts don't change the type, so we just ignore them.
-      case CK_NoOp:
-        e = ce->getSubExpr();
-        continue;
-
-      case CK_LValueToRValue: {
-        TryEmitResult loadResult
-          = tryEmitARCRetainLoadOfScalar(CGF, ce->getSubExpr());
-        if (resultType) {
-          llvm::Value *value = loadResult.getPointer();
-          value = CGF.Builder.CreateBitCast(value, resultType);
-          loadResult.setPointer(value);
-        }
-        return loadResult;
-      }
+  // Look through parens, __extension__, generic selection, etc.
+  e = e->IgnoreParens();
 
-      // These casts can change the type, so remember that and
-      // soldier on.  We only need to remember the outermost such
-      // cast, though.
-      case CK_CPointerToObjCPointerCast:
-      case CK_BlockPointerToObjCPointerCast:
-      case CK_AnyPointerToBlockPointerCast:
-      case CK_BitCast:
-        if (!resultType)
-          resultType = CGF.ConvertType(ce->getType());
-        e = ce->getSubExpr();
-        assert(e->getType()->hasPointerRepresentation());
-        continue;
-
-      // For consumptions, just emit the subexpression and thus elide
-      // the retain/release pair.
-      case CK_ARCConsumeObject: {
-        llvm::Value *result = CGF.EmitScalarExpr(ce->getSubExpr());
-        if (resultType) result = CGF.Builder.CreateBitCast(result, resultType);
-        return TryEmitResult(result, true);
-      }
+  // Handle certain kinds of casts.
+  if (const CastExpr *ce = dyn_cast<CastExpr>(e)) {
+    return asImpl().visitCastExpr(ce);
 
-      // Block extends are net +0.  Naively, we could just recurse on
-      // the subexpression, but actually we need to ensure that the
-      // value is copied as a block, so there's a little filter here.
-      case CK_ARCExtendBlockObject: {
-        llvm::Value *result; // will be a +0 value
+  // Handle the comma operator.
+  } else if (auto op = dyn_cast<BinaryOperator>(e)) {
+    return asImpl().visitBinaryOperator(op);
 
-        // If we can't safely assume the sub-expression will produce a
-        // block-copied value, emit the sub-expression at +0.
-        if (shouldEmitSeparateBlockRetain(ce->getSubExpr())) {
-          result = CGF.EmitScalarExpr(ce->getSubExpr());
+  // TODO: handle conditional operators here
 
-        // Otherwise, try to emit the sub-expression at +1 recursively.
-        } else {
-          TryEmitResult subresult
-            = tryEmitARCRetainScalarExpr(CGF, ce->getSubExpr());
-          result = subresult.getPointer();
-
-          // If that produced a retained value, just use that,
-          // possibly casting down.
-          if (subresult.getInt()) {
-            if (resultType)
-              result = CGF.Builder.CreateBitCast(result, resultType);
-            return TryEmitResult(result, true);
-          }
+  // For calls and message sends, use the retained-call logic.
+  // Delegate inits are a special case in that they're the only
+  // returns-retained expression that *isn't* surrounded by
+  // a consume.
+  } else if (isa<CallExpr>(e) ||
+             (isa<ObjCMessageExpr>(e) &&
+              !cast<ObjCMessageExpr>(e)->isDelegateInitCall())) {
+    return asImpl().visitCall(e);
 
-          // Otherwise it's +0.
-        }
+  // Look through pseudo-object expressions.
+  } else if (const PseudoObjectExpr *pseudo = dyn_cast<PseudoObjectExpr>(e)) {
+    return asImpl().visitPseudoObjectExpr(pseudo);
+  }
 
-        // Retain the object as a block, then cast down.
-        result = CGF.EmitARCRetainBlock(result, /*mandatory*/ true);
-        if (resultType) result = CGF.Builder.CreateBitCast(result, resultType);
-        return TryEmitResult(result, true);
-      }
+  return asImpl().visitExpr(e);
+}
 
-      // For reclaims, emit the subexpression as a retained call and
-      // skip the consumption.
-      case CK_ARCReclaimReturnedObject: {
-        llvm::Value *result = emitARCRetainCall(CGF, ce->getSubExpr());
-        if (resultType) result = CGF.Builder.CreateBitCast(result, resultType);
-        return TryEmitResult(result, true);
-      }
+namespace {
 
-      default:
-        break;
-      }
+/// An emitter for +1 results.
+struct ARCRetainExprEmitter :
+  public ARCExprEmitter<ARCRetainExprEmitter, TryEmitResult> {
 
-    // Skip __extension__.
-    } else if (const UnaryOperator *op = dyn_cast<UnaryOperator>(e)) {
-      if (op->getOpcode() == UO_Extension) {
-        e = op->getSubExpr();
-        continue;
-      }
+  ARCRetainExprEmitter(CodeGenFunction &CGF) : ARCExprEmitter(CGF) {}
+
+  llvm::Value *getValueOfResult(TryEmitResult result) {
+    return result.getPointer();
+  }
 
-    // For calls and message sends, use the retained-call logic.
-    // Delegate inits are a special case in that they're the only
-    // returns-retained expression that *isn't* surrounded by
-    // a consume.
-    } else if (isa<CallExpr>(e) ||
-               (isa<ObjCMessageExpr>(e) &&
-                !cast<ObjCMessageExpr>(e)->isDelegateInitCall())) {
-      llvm::Value *result = emitARCRetainCall(CGF, e);
-      if (resultType) result = CGF.Builder.CreateBitCast(result, resultType);
-      return TryEmitResult(result, true);
-
-    // Look through pseudo-object expressions.
-    } else if (const PseudoObjectExpr *pseudo = dyn_cast<PseudoObjectExpr>(e)) {
-      TryEmitResult result
-        = tryEmitARCRetainPseudoObject(CGF, pseudo);
-      if (resultType) {
-        llvm::Value *value = result.getPointer();
-        value = CGF.Builder.CreateBitCast(value, resultType);
-        result.setPointer(value);
+  TryEmitResult emitBitCast(TryEmitResult result, llvm::Type *resultType) {
+    llvm::Value *value = result.getPointer();
+    value = CGF.Builder.CreateBitCast(value, resultType);
+    result.setPointer(value);
+    return result;
+  }
+
+  TryEmitResult visitLValueToRValue(const Expr *e) {
+    return tryEmitARCRetainLoadOfScalar(CGF, e);
+  }
+
+  /// For consumptions, just emit the subexpression and thus elide
+  /// the retain/release pair.
+  TryEmitResult visitConsumeObject(const Expr *e) {
+    llvm::Value *result = CGF.EmitScalarExpr(e);
+    return TryEmitResult(result, true);
+  }
+
+  /// Block extends are net +0.  Naively, we could just recurse on
+  /// the subexpression, but actually we need to ensure that the
+  /// value is copied as a block, so there's a little filter here.
+  TryEmitResult visitExtendBlockObject(const Expr *e) {
+    llvm::Value *result; // will be a +0 value
+
+    // If we can't safely assume the sub-expression will produce a
+    // block-copied value, emit the sub-expression at +0.
+    if (shouldEmitSeparateBlockRetain(e)) {
+      result = CGF.EmitScalarExpr(e);
+
+    // Otherwise, try to emit the sub-expression at +1 recursively.
+    } else {
+      TryEmitResult subresult = asImpl().visit(e);
+
+      // If that produced a retained value, just use that.
+      if (subresult.getInt()) {
+        return subresult;
       }
-      return result;
+
+      // Otherwise it's +0.
+      result = subresult.getPointer();
     }
 
-    // Conservatively halt the search at any other expression kind.
-    break;
+    // Retain the object as a block.
+    result = CGF.EmitARCRetainBlock(result, /*mandatory*/ true);
+    return TryEmitResult(result, true);
   }
 
-  // We didn't find an obvious production, so emit what we've got and
-  // tell the caller that we didn't manage to retain.
-  llvm::Value *result = CGF.EmitScalarExpr(e);
-  if (resultType) result = CGF.Builder.CreateBitCast(result, resultType);
-  return TryEmitResult(result, false);
+  /// For reclaims, emit the subexpression as a retained call and
+  /// skip the consumption.
+  TryEmitResult visitReclaimReturnedObject(const Expr *e) {
+    llvm::Value *result = emitARCRetainCallResult(CGF, e);
+    return TryEmitResult(result, true);
+  }
+
+  /// When we have an undecorated call, retroactively do a claim.
+  TryEmitResult visitCall(const Expr *e) {
+    llvm::Value *result = emitARCRetainCallResult(CGF, e);
+    return TryEmitResult(result, true);
+  }
+
+  // TODO: maybe special-case visitBinAssignWeak?
+
+  TryEmitResult visitExpr(const Expr *e) {
+    // We didn't find an obvious production, so emit what we've got and
+    // tell the caller that we didn't manage to retain.
+    llvm::Value *result = CGF.EmitScalarExpr(e);
+    return TryEmitResult(result, false);
+  }
+};
+}
+
+static TryEmitResult
+tryEmitARCRetainScalarExpr(CodeGenFunction &CGF, const Expr *e) {
+  return ARCRetainExprEmitter(CGF).visit(e);
 }
 
 static llvm::Value *emitARCRetainLoadOfScalar(CodeGenFunction &CGF,
@@ -2807,6 +2984,96 @@ llvm::Value *CodeGenFunction::EmitObjCThrowOperand(const Expr *expr) {
   return EmitScalarExpr(expr);
 }
 
+namespace {
+
+/// An emitter for assigning into an __unsafe_unretained context.
+struct ARCUnsafeUnretainedExprEmitter :
+  public ARCExprEmitter<ARCUnsafeUnretainedExprEmitter, llvm::Value*> {
+
+  ARCUnsafeUnretainedExprEmitter(CodeGenFunction &CGF) : ARCExprEmitter(CGF) {}
+
+  llvm::Value *getValueOfResult(llvm::Value *value) {
+    return value;
+  }
+
+  llvm::Value *emitBitCast(llvm::Value *value, llvm::Type *resultType) {
+    return CGF.Builder.CreateBitCast(value, resultType);
+  }
+
+  llvm::Value *visitLValueToRValue(const Expr *e) {
+    return CGF.EmitScalarExpr(e);
+  }
+
+  /// For consumptions, just emit the subexpression and perform the
+  /// consumption like normal.
+  llvm::Value *visitConsumeObject(const Expr *e) {
+    llvm::Value *value = CGF.EmitScalarExpr(e);
+    return CGF.EmitObjCConsumeObject(e->getType(), value);
+  }
+
+  /// No special logic for block extensions.  (This probably can't
+  /// actually happen in this emitter, though.)
+  llvm::Value *visitExtendBlockObject(const Expr *e) {
+    return CGF.EmitARCExtendBlockObject(e);
+  }
+
+  /// For reclaims, perform an unsafeClaim if that's enabled.
+  llvm::Value *visitReclaimReturnedObject(const Expr *e) {
+    return CGF.EmitARCReclaimReturnedObject(e, /*unsafe*/ true);
+  }
+
+  /// When we have an undecorated call, just emit it without adding
+  /// the unsafeClaim.
+  llvm::Value *visitCall(const Expr *e) {
+    return CGF.EmitScalarExpr(e);
+  }
+
+  /// Just do normal scalar emission in the default case.
+  llvm::Value *visitExpr(const Expr *e) {
+    return CGF.EmitScalarExpr(e);
+  }
+};
+}
+
+static llvm::Value *emitARCUnsafeUnretainedScalarExpr(CodeGenFunction &CGF,
+                                                      const Expr *e) {
+  return ARCUnsafeUnretainedExprEmitter(CGF).visit(e);
+}
+
+/// EmitARCUnsafeUnretainedScalarExpr - Semantically equivalent to
+/// immediately releasing the resut of EmitARCRetainScalarExpr, but
+/// avoiding any spurious retains, including by performing reclaims
+/// with objc_unsafeClaimAutoreleasedReturnValue.
+llvm::Value *CodeGenFunction::EmitARCUnsafeUnretainedScalarExpr(const Expr *e) {
+  // Look through full-expressions.
+  if (const ExprWithCleanups *cleanups = dyn_cast<ExprWithCleanups>(e)) {
+    enterFullExpression(cleanups);
+    RunCleanupsScope scope(*this);
+    return emitARCUnsafeUnretainedScalarExpr(*this, cleanups->getSubExpr());
+  }
+
+  return emitARCUnsafeUnretainedScalarExpr(*this, e);
+}
+
+std::pair<LValue,llvm::Value*>
+CodeGenFunction::EmitARCStoreUnsafeUnretained(const BinaryOperator *e,
+                                              bool ignored) {
+  // Evaluate the RHS first.  If we're ignoring the result, assume
+  // that we can emit at an unsafe +0.
+  llvm::Value *value;
+  if (ignored) {
+    value = EmitARCUnsafeUnretainedScalarExpr(e->getRHS());
+  } else {
+    value = EmitScalarExpr(e->getRHS());
+  }
+
+  // Emit the LHS and perform the store.
+  LValue lvalue = EmitLValue(e->getLHS());
+  EmitStoreOfScalar(value, lvalue);
+
+  return std::pair<LValue,llvm::Value*>(std::move(lvalue), value);
+}
+
 std::pair<LValue,llvm::Value*>
 CodeGenFunction::EmitARCStoreStrong(const BinaryOperator *e,
                                     bool ignored) {
@@ -2935,8 +3202,8 @@ CodeGenFunction::GenerateObjCAtomicSetterCopyHelperFunction(
   ImplicitParamDecl srcDecl(getContext(), FD, SourceLocation(), nullptr, SrcTy);
   args.push_back(&srcDecl);
 
-  const CGFunctionInfo &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      C.VoidTy, args, FunctionType::ExtInfo(), RequiredArgs::All);
+  const CGFunctionInfo &FI =
+    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, args);
 
   llvm::FunctionType *LTy = CGM.getTypes().GetFunctionType(FI);
   
@@ -3016,8 +3283,8 @@ CodeGenFunction::GenerateObjCAtomicGetterCopyHelperFunction(
   ImplicitParamDecl srcDecl(getContext(), FD, SourceLocation(), nullptr, SrcTy);
   args.push_back(&srcDecl);
 
-  const CGFunctionInfo &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      C.VoidTy, args, FunctionType::ExtInfo(), RequiredArgs::All);
+  const CGFunctionInfo &FI =
+    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, args);
 
   llvm::FunctionType *LTy = CGM.getTypes().GetFunctionType(FI);
   
diff --git a/lib/CodeGen/CGObjCGNU.cpp b/lib/CodeGen/CGObjCGNU.cpp
index f0af3e924c09..caafef84c333 100644
--- a/lib/CodeGen/CGObjCGNU.cpp
+++ b/lib/CodeGen/CGObjCGNU.cpp
@@ -35,11 +35,9 @@
 #include "llvm/Support/Compiler.h"
 #include <cstdarg>
 
-
 using namespace clang;
 using namespace CodeGen;
 
-
 namespace {
 /// Class that lazily initialises the runtime function.  Avoids inserting the
 /// types and the function declaration into a module if they're not used, and
@@ -161,6 +159,7 @@ protected:
   /// runtime provides some LLVM passes that can use this to do things like
   /// automatic IMP caching and speculative inlining.
   unsigned msgSendMDKind;
+
   /// Helper function that generates a constant string and returns a pointer to
   /// the start of the string.  The result of this function can be used anywhere
   /// where the C code specifies const char*.  
@@ -170,6 +169,7 @@ protected:
     return llvm::ConstantExpr::getGetElementPtr(Array.getElementType(),
                                                 Array.getPointer(), Zeros);
   }
+
   /// Emits a linkonce_odr string, whose name is the prefix followed by the
   /// string value.  This allows the linker to combine the strings between
   /// different modules.  Used for EH typeinfo names, selector strings, and a
@@ -186,6 +186,7 @@ protected:
     return llvm::ConstantExpr::getGetElementPtr(ConstStr->getValueType(),
                                                 ConstStr, Zeros);
   }
+
   /// Generates a global structure, initialized by the elements in the vector.
   /// The element types must match the types of the structure elements in the
   /// first argument.
@@ -201,6 +202,7 @@ protected:
     GV->setAlignment(Align.getQuantity());
     return GV;
   }
+
   /// Generates a global array.  The vector must contain the same number of
   /// elements that the array type declares, of the type specified as the array
   /// element type.
@@ -216,6 +218,7 @@ protected:
     GV->setAlignment(Align.getQuantity());
     return GV;
   }
+
   /// Generates a global array, inferring the array type from the specified
   /// element type and the size of the initialiser.  
   llvm::GlobalVariable *MakeGlobalArray(llvm::Type *Ty,
@@ -227,6 +230,7 @@ protected:
     llvm::ArrayType *ArrayTy = llvm::ArrayType::get(Ty, V.size());
     return MakeGlobal(ArrayTy, V, Align, Name, linkage);
   }
+
   /// Returns a property name and encoding string.
   llvm::Constant *MakePropertyEncodingString(const ObjCPropertyDecl *PD,
                                              const Decl *Container) {
@@ -245,6 +249,7 @@ protected:
     }
     return MakeConstantString(PD->getNameAsString());
   }
+
   /// Push the property attributes into two structure fields. 
   void PushPropertyAttributes(std::vector<llvm::Constant*> &Fields,
       ObjCPropertyDecl *property, bool isSynthesized=true, bool
@@ -273,6 +278,7 @@ protected:
     Fields.push_back(llvm::ConstantInt::get(Int8Ty, 0));
     Fields.push_back(llvm::ConstantInt::get(Int8Ty, 0));
   }
+
   /// Ensures that the value has the required type, by inserting a bitcast if
   /// required.  This function lets us avoid inserting bitcasts that are
   /// redundant.
@@ -284,12 +290,14 @@ protected:
     if (V.getType() == Ty) return V;
     return B.CreateBitCast(V, Ty);
   }
+
   // Some zeros used for GEPs in lots of places.
   llvm::Constant *Zeros[2];
   /// Null pointer value.  Mainly used as a terminator in various arrays.
   llvm::Constant *NULLPtr;
   /// LLVM context.
   llvm::LLVMContext &VMContext;
+
 private:
   /// Placeholder for the class.  Lots of things refer to the class before we've
   /// actually emitted it.  We use this alias as a placeholder, and then replace
@@ -360,7 +368,6 @@ protected:
   LazyRuntimeFunction SyncExitFn;
 
 private:
-
   /// Function called if fast enumeration detects that the collection is
   /// modified during the update.
   LazyRuntimeFunction EnumerationMutationFn;
@@ -385,7 +392,7 @@ private:
   /// Objective-C 1 property structures when targeting the GCC runtime or it
   /// will abort.
   const int ProtocolVersion;
-private:
+
   /// Generates an instance variable list structure.  This is a structure
   /// containing a size and an array of structures containing instance variable
   /// metadata.  This is used purely for introspection in the fragile ABI.  In
@@ -393,6 +400,7 @@ private:
   llvm::Constant *GenerateIvarList(ArrayRef<llvm::Constant *> IvarNames,
                                    ArrayRef<llvm::Constant *> IvarTypes,
                                    ArrayRef<llvm::Constant *> IvarOffsets);
+
   /// Generates a method list structure.  This is a structure containing a size
   /// and an array of structures containing method metadata.
   ///
@@ -403,23 +411,28 @@ private:
       ArrayRef<Selector> MethodSels,
       ArrayRef<llvm::Constant *> MethodTypes,
       bool isClassMethodList);
+
   /// Emits an empty protocol.  This is used for \@protocol() where no protocol
   /// is found.  The runtime will (hopefully) fix up the pointer to refer to the
   /// real protocol.
   llvm::Constant *GenerateEmptyProtocol(const std::string &ProtocolName);
+
   /// Generates a list of property metadata structures.  This follows the same
   /// pattern as method and instance variable metadata lists.
   llvm::Constant *GeneratePropertyList(const ObjCImplementationDecl *OID,
         SmallVectorImpl<Selector> &InstanceMethodSels,
         SmallVectorImpl<llvm::Constant*> &InstanceMethodTypes);
+
   /// Generates a list of referenced protocols.  Classes, categories, and
   /// protocols all use this structure.
   llvm::Constant *GenerateProtocolList(ArrayRef<std::string> Protocols);
+
   /// To ensure that all protocols are seen by the runtime, we add a category on
   /// a class defined in the runtime, declaring no methods, but adopting the
   /// protocols.  This is a horribly ugly hack, but it allows us to collect all
   /// of the protocols without changing the ABI.
   void GenerateProtocolHolderCategory();
+
   /// Generates a class structure.
   llvm::Constant *GenerateClassStructure(
       llvm::Constant *MetaClass,
@@ -436,25 +449,31 @@ private:
       llvm::Constant *StrongIvarBitmap,
       llvm::Constant *WeakIvarBitmap,
       bool isMeta=false);
+
   /// Generates a method list.  This is used by protocols to define the required
   /// and optional methods.
   llvm::Constant *GenerateProtocolMethodList(
       ArrayRef<llvm::Constant *> MethodNames,
       ArrayRef<llvm::Constant *> MethodTypes);
+
   /// Returns a selector with the specified type encoding.  An empty string is
   /// used to return an untyped selector (with the types field set to NULL).
   llvm::Value *GetSelector(CodeGenFunction &CGF, Selector Sel,
                            const std::string &TypeEncoding);
+
   /// Returns the variable used to store the offset of an instance variable.
   llvm::GlobalVariable *ObjCIvarOffsetVariable(const ObjCInterfaceDecl *ID,
       const ObjCIvarDecl *Ivar);
   /// Emits a reference to a class.  This allows the linker to object if there
   /// is no class of the matching name.
+
 protected:
   void EmitClassRef(const std::string &className);
+
   /// Emits a pointer to the named class
   virtual llvm::Value *GetClassNamed(CodeGenFunction &CGF,
                                      const std::string &Name, bool isWeak);
+
   /// Looks up the method for sending a message to the specified object.  This
   /// mechanism differs between the GCC and GNU runtimes, so this method must be
   /// overridden in subclasses.
@@ -463,6 +482,7 @@ protected:
                                  llvm::Value *cmd,
                                  llvm::MDNode *node,
                                  MessageSendInfo &MSI) = 0;
+
   /// Looks up the method for sending a message to a superclass.  This
   /// mechanism differs between the GCC and GNU runtimes, so this method must
   /// be overridden in subclasses.
@@ -470,6 +490,7 @@ protected:
                                       Address ObjCSuper,
                                       llvm::Value *cmd,
                                       MessageSendInfo &MSI) = 0;
+
   /// Libobjc2 uses a bitfield representation where small(ish) bitfields are
   /// stored in a 64-bit value with the low bit set to 1 and the remaining 63
   /// bits set to their values, LSB first, while larger ones are stored in a
@@ -482,6 +503,7 @@ protected:
   /// a bitfield with the 64th bit set will be (int64_t)&{ 2, [0, 1<<31] },
   /// while a bitfield / with the 63rd bit set will be 1<<64.
   llvm::Constant *MakeBitField(ArrayRef<bool> bits);
+
 public:
   CGObjCGNU(CodeGenModule &cgm, unsigned runtimeABIVersion,
       unsigned protocolClassVersion);
@@ -569,11 +591,12 @@ public:
     return NULLPtr;
   }
 
-  llvm::GlobalVariable *GetClassGlobal(const std::string &Name,
+  llvm::GlobalVariable *GetClassGlobal(StringRef Name,
                                        bool Weak = false) override {
     return nullptr;
   }
 };
+
 /// Class representing the legacy GCC Objective-C ABI.  This is the default when
 /// -fobjc-nonfragile-abi is not specified.
 ///
@@ -590,6 +613,7 @@ class CGObjCGCC : public CGObjCGNU {
   /// structure describing the receiver and the class, and a selector as
   /// arguments.  Returns the IMP for the corresponding method.
   LazyRuntimeFunction MsgLookupSuperFn;
+
 protected:
   llvm::Value *LookupIMP(CodeGenFunction &CGF, llvm::Value *&Receiver,
                          llvm::Value *cmd, llvm::MDNode *node,
@@ -602,23 +626,26 @@ protected:
     imp->setMetadata(msgSendMDKind, node);
     return imp.getInstruction();
   }
+
   llvm::Value *LookupIMPSuper(CodeGenFunction &CGF, Address ObjCSuper,
                               llvm::Value *cmd, MessageSendInfo &MSI) override {
-      CGBuilderTy &Builder = CGF.Builder;
-      llvm::Value *lookupArgs[] = {EnforceType(Builder, ObjCSuper,
-          PtrToObjCSuperTy).getPointer(), cmd};
-      return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFn, lookupArgs);
-    }
-  public:
-    CGObjCGCC(CodeGenModule &Mod) : CGObjCGNU(Mod, 8, 2) {
-      // IMP objc_msg_lookup(id, SEL);
-      MsgLookupFn.init(&CGM, "objc_msg_lookup", IMPTy, IdTy, SelectorTy,
-                       nullptr);
-      // IMP objc_msg_lookup_super(struct objc_super*, SEL);
-      MsgLookupSuperFn.init(&CGM, "objc_msg_lookup_super", IMPTy,
-              PtrToObjCSuperTy, SelectorTy, nullptr);
-    }
+    CGBuilderTy &Builder = CGF.Builder;
+    llvm::Value *lookupArgs[] = {EnforceType(Builder, ObjCSuper,
+        PtrToObjCSuperTy).getPointer(), cmd};
+    return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFn, lookupArgs);
+  }
+
+public:
+  CGObjCGCC(CodeGenModule &Mod) : CGObjCGNU(Mod, 8, 2) {
+    // IMP objc_msg_lookup(id, SEL);
+    MsgLookupFn.init(&CGM, "objc_msg_lookup", IMPTy, IdTy, SelectorTy,
+                     nullptr);
+    // IMP objc_msg_lookup_super(struct objc_super*, SEL);
+    MsgLookupSuperFn.init(&CGM, "objc_msg_lookup_super", IMPTy,
+                          PtrToObjCSuperTy, SelectorTy, nullptr);
+  }
 };
+
 /// Class used when targeting the new GNUstep runtime ABI.
 class CGObjCGNUstep : public CGObjCGNU {
     /// The slot lookup function.  Returns a pointer to a cacheable structure
@@ -646,8 +673,10 @@ class CGObjCGNUstep : public CGObjCGNU {
     /// Type of an slot structure pointer.  This is returned by the various
     /// lookup functions.
     llvm::Type *SlotTy;
+
   public:
     llvm::Constant *GetEHType(QualType T) override;
+
   protected:
     llvm::Value *LookupIMP(CodeGenFunction &CGF, llvm::Value *&Receiver,
                            llvm::Value *cmd, llvm::MDNode *node,
@@ -689,6 +718,7 @@ class CGObjCGNUstep : public CGObjCGNU {
       Receiver = Builder.CreateLoad(ReceiverPtr, true);
       return imp;
     }
+
     llvm::Value *LookupIMPSuper(CodeGenFunction &CGF, Address ObjCSuper,
                                 llvm::Value *cmd,
                                 MessageSendInfo &MSI) override {
@@ -702,6 +732,7 @@ class CGObjCGNUstep : public CGObjCGNU {
       return Builder.CreateAlignedLoad(Builder.CreateStructGEP(nullptr, slot, 4),
                                        CGF.getPointerAlign());
     }
+
   public:
     CGObjCGNUstep(CodeGenModule &Mod) : CGObjCGNU(Mod, 9, 3) {
       const ObjCRuntime &R = CGM.getLangOpts().ObjCRuntime;
@@ -753,6 +784,7 @@ class CGObjCGNUstep : public CGObjCGNU {
       CxxAtomicObjectGetFn.init(&CGM, "objc_getCppObjectAtomic", VoidTy, PtrTy,
           PtrTy, PtrTy, nullptr);
     }
+
     llvm::Constant *GetCppAtomicObjectGetFunction() override {
       // The optimised functions were added in version 1.7 of the GNUstep
       // runtime.
@@ -760,6 +792,7 @@ class CGObjCGNUstep : public CGObjCGNU {
           VersionTuple(1, 7));
       return CxxAtomicObjectGetFn;
     }
+
     llvm::Constant *GetCppAtomicObjectSetFunction() override {
       // The optimised functions were added in version 1.7 of the GNUstep
       // runtime.
@@ -767,6 +800,7 @@ class CGObjCGNUstep : public CGObjCGNU {
           VersionTuple(1, 7));
       return CxxAtomicObjectSetFn;
     }
+
     llvm::Constant *GetOptimizedPropertySetFunction(bool atomic,
                                                     bool copy) override {
       // The optimised property functions omit the GC check, and so are not
@@ -821,32 +855,29 @@ protected:
 
   llvm::Value *LookupIMPSuper(CodeGenFunction &CGF, Address ObjCSuper,
                               llvm::Value *cmd, MessageSendInfo &MSI) override {
-      CGBuilderTy &Builder = CGF.Builder;
-      llvm::Value *lookupArgs[] = {EnforceType(Builder, ObjCSuper.getPointer(),
-          PtrToObjCSuperTy), cmd};
+    CGBuilderTy &Builder = CGF.Builder;
+    llvm::Value *lookupArgs[] = {
+        EnforceType(Builder, ObjCSuper.getPointer(), PtrToObjCSuperTy), cmd,
+    };
 
-      if (CGM.ReturnTypeUsesSRet(MSI.CallInfo))
-        return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFnSRet, lookupArgs);
-      else
-        return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFn, lookupArgs);
-    }
+    if (CGM.ReturnTypeUsesSRet(MSI.CallInfo))
+      return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFnSRet, lookupArgs);
+    else
+      return CGF.EmitNounwindRuntimeCall(MsgLookupSuperFn, lookupArgs);
+  }
 
-  llvm::Value *GetClassNamed(CodeGenFunction &CGF,
-                             const std::string &Name, bool isWeak) override {
+  llvm::Value *GetClassNamed(CodeGenFunction &CGF, const std::string &Name,
+                             bool isWeak) override {
     if (isWeak)
       return CGObjCGNU::GetClassNamed(CGF, Name, isWeak);
 
     EmitClassRef(Name);
-
     std::string SymbolName = "_OBJC_CLASS_" + Name;
-
     llvm::GlobalVariable *ClassSymbol = TheModule.getGlobalVariable(SymbolName);
-
     if (!ClassSymbol)
       ClassSymbol = new llvm::GlobalVariable(TheModule, LongTy, false,
                                              llvm::GlobalValue::ExternalLinkage,
                                              nullptr, SymbolName);
-
     return ClassSymbol;
   }
 
@@ -865,7 +896,6 @@ public:
 };
 } // end anonymous namespace
 
-
 /// Emits a reference to a dummy variable which is emitted with each class.
 /// This ensures that a linker error will be generated when trying to link
 /// together modules where a referenced class is not defined.
@@ -1021,8 +1051,7 @@ CGObjCGNU::CGObjCGNU(CodeGenModule &cgm, unsigned runtimeABIVersion,
 }
 
 llvm::Value *CGObjCGNU::GetClassNamed(CodeGenFunction &CGF,
-                                      const std::string &Name,
-                                      bool isWeak) {
+                                      const std::string &Name, bool isWeak) {
   llvm::Constant *ClassName = MakeConstantString(Name);
   // With the incompatible ABI, this will need to be replaced with a direct
   // reference to the class symbol.  For the compatible nonfragile ABI we are
@@ -1044,15 +1073,48 @@ llvm::Value *CGObjCGNU::GetClassNamed(CodeGenFunction &CGF,
 // techniques can modify the name -> class mapping.
 llvm::Value *CGObjCGNU::GetClass(CodeGenFunction &CGF,
                                  const ObjCInterfaceDecl *OID) {
-  return GetClassNamed(CGF, OID->getNameAsString(), OID->isWeakImported());
+  auto *Value =
+      GetClassNamed(CGF, OID->getNameAsString(), OID->isWeakImported());
+  if (CGM.getTriple().isOSBinFormatCOFF()) {
+    if (auto *ClassSymbol = dyn_cast<llvm::GlobalVariable>(Value)) {
+      auto DLLStorage = llvm::GlobalValue::DefaultStorageClass;
+      if (OID->hasAttr<DLLExportAttr>())
+        DLLStorage = llvm::GlobalValue::DLLExportStorageClass;
+      else if (OID->hasAttr<DLLImportAttr>())
+        DLLStorage = llvm::GlobalValue::DLLImportStorageClass;
+      ClassSymbol->setDLLStorageClass(DLLStorage);
+    }
+  }
+  return Value;
 }
+
 llvm::Value *CGObjCGNU::EmitNSAutoreleasePoolClassRef(CodeGenFunction &CGF) {
-  return GetClassNamed(CGF, "NSAutoreleasePool", false);
+  auto *Value  = GetClassNamed(CGF, "NSAutoreleasePool", false);
+  if (CGM.getTriple().isOSBinFormatCOFF()) {
+    if (auto *ClassSymbol = dyn_cast<llvm::GlobalVariable>(Value)) {
+      IdentifierInfo &II = CGF.CGM.getContext().Idents.get("NSAutoreleasePool");
+      TranslationUnitDecl *TUDecl = CGM.getContext().getTranslationUnitDecl();
+      DeclContext *DC = TranslationUnitDecl::castToDeclContext(TUDecl);
+
+      const VarDecl *VD = nullptr;
+      for (const auto &Result : DC->lookup(&II))
+        if ((VD = dyn_cast<VarDecl>(Result)))
+          break;
+
+      auto DLLStorage = llvm::GlobalValue::DefaultStorageClass;
+      if (!VD || VD->hasAttr<DLLImportAttr>())
+        DLLStorage = llvm::GlobalValue::DLLImportStorageClass;
+      else if (VD->hasAttr<DLLExportAttr>())
+        DLLStorage = llvm::GlobalValue::DLLExportStorageClass;
+
+      ClassSymbol->setDLLStorageClass(DLLStorage);
+    }
+  }
+  return Value;
 }
 
 llvm::Value *CGObjCGNU::GetSelector(CodeGenFunction &CGF, Selector Sel,
                                     const std::string &TypeEncoding) {
-
   SmallVectorImpl<TypedSelector> &Types = SelectorTable[Sel];
   llvm::GlobalAlias *SelValue = nullptr;
 
@@ -1247,8 +1309,6 @@ CGObjCGNU::GenerateMessageSendSuper(CodeGenFunction &CGF,
   }
 
   llvm::Value *cmd = GetSelector(CGF, Sel);
-
-
   CallArgList ActualArgs;
 
   ActualArgs.add(RValue::get(EnforceType(Builder, Receiver, IdTy)), ASTIdTy);
@@ -1497,21 +1557,17 @@ GenerateMethodList(StringRef ClassName,
     IMPTy, //Method pointer
     nullptr);
   std::vector<llvm::Constant*> Methods;
-  std::vector<llvm::Constant*> Elements;
   for (unsigned int i = 0, e = MethodTypes.size(); i < e; ++i) {
-    Elements.clear();
     llvm::Constant *Method =
       TheModule.getFunction(SymbolNameForMethod(ClassName, CategoryName,
                                                 MethodSels[i],
                                                 isClassMethodList));
     assert(Method && "Can't generate metadata for method that doesn't exist");
     llvm::Constant *C = MakeConstantString(MethodSels[i].getAsString());
-    Elements.push_back(C);
-    Elements.push_back(MethodTypes[i]);
     Method = llvm::ConstantExpr::getBitCast(Method,
         IMPTy);
-    Elements.push_back(Method);
-    Methods.push_back(llvm::ConstantStruct::get(ObjCMethodTy, Elements));
+    Methods.push_back(
+        llvm::ConstantStruct::get(ObjCMethodTy, {C, MethodTypes[i], Method}));
   }
 
   // Array of method structures
@@ -1554,23 +1610,18 @@ GenerateIvarList(ArrayRef<llvm::Constant *> IvarNames,
     IntTy,
     nullptr);
   std::vector<llvm::Constant*> Ivars;
-  std::vector<llvm::Constant*> Elements;
   for (unsigned int i = 0, e = IvarNames.size() ; i < e ; i++) {
-    Elements.clear();
-    Elements.push_back(IvarNames[i]);
-    Elements.push_back(IvarTypes[i]);
-    Elements.push_back(IvarOffsets[i]);
-    Ivars.push_back(llvm::ConstantStruct::get(ObjCIvarTy, Elements));
+    Ivars.push_back(llvm::ConstantStruct::get(
+        ObjCIvarTy, {IvarNames[i], IvarTypes[i], IvarOffsets[i]}));
   }
 
   // Array of method structures
   llvm::ArrayType *ObjCIvarArrayTy = llvm::ArrayType::get(ObjCIvarTy,
       IvarNames.size());
 
-
-  Elements.clear();
-  Elements.push_back(llvm::ConstantInt::get(IntTy, (int)IvarNames.size()));
-  Elements.push_back(llvm::ConstantArray::get(ObjCIvarArrayTy, Ivars));
+  llvm::Constant *Elements[] = {
+      llvm::ConstantInt::get(IntTy, (int)IvarNames.size()),
+      llvm::ConstantArray::get(ObjCIvarArrayTy, Ivars)};
   // Structure containing array and array count
   llvm::StructType *ObjCIvarListTy = llvm::StructType::get(IntTy,
     ObjCIvarArrayTy,
@@ -1682,12 +1733,9 @@ GenerateProtocolMethodList(ArrayRef<llvm::Constant *> MethodNames,
     PtrToInt8Ty,
     nullptr);
   std::vector<llvm::Constant*> Methods;
-  std::vector<llvm::Constant*> Elements;
   for (unsigned int i = 0, e = MethodTypes.size() ; i < e ; i++) {
-    Elements.clear();
-    Elements.push_back(MethodNames[i]);
-    Elements.push_back(MethodTypes[i]);
-    Methods.push_back(llvm::ConstantStruct::get(ObjCMethodDescTy, Elements));
+    Methods.push_back(llvm::ConstantStruct::get(
+        ObjCMethodDescTy, {MethodNames[i], MethodTypes[i]}));
   }
   llvm::ArrayType *ObjCMethodArrayTy = llvm::ArrayType::get(ObjCMethodDescTy,
       MethodNames.size());
@@ -1762,17 +1810,13 @@ llvm::Constant *CGObjCGNU::GenerateEmptyProtocol(
       MethodList->getType(),
       MethodList->getType(),
       nullptr);
-  std::vector<llvm::Constant*> Elements;
   // The isa pointer must be set to a magic number so the runtime knows it's
   // the correct layout.
-  Elements.push_back(llvm::ConstantExpr::getIntToPtr(
-        llvm::ConstantInt::get(Int32Ty, ProtocolVersion), IdTy));
-  Elements.push_back(MakeConstantString(ProtocolName, ".objc_protocol_name"));
-  Elements.push_back(ProtocolList);
-  Elements.push_back(MethodList);
-  Elements.push_back(MethodList);
-  Elements.push_back(MethodList);
-  Elements.push_back(MethodList);
+  llvm::Constant *Elements[] = {
+      llvm::ConstantExpr::getIntToPtr(
+          llvm::ConstantInt::get(Int32Ty, ProtocolVersion), IdTy),
+      MakeConstantString(ProtocolName, ".objc_protocol_name"), ProtocolList,
+      MethodList, MethodList, MethodList, MethodList};
   return MakeGlobal(ProtocolTy, Elements, CGM.getPointerAlign(),
                     ".objc_protocol");
 }
@@ -1849,7 +1893,7 @@ void CGObjCGNU::GenerateProtocol(const ObjCProtocolDecl *PD) {
 
   // Add all of the property methods need adding to the method list and to the
   // property metadata list.
-  for (auto *property : PD->properties()) {
+  for (auto *property : PD->instance_properties()) {
     std::vector<llvm::Constant*> Fields;
 
     Fields.push_back(MakePropertyEncodingString(property, nullptr));
@@ -1920,19 +1964,14 @@ void CGObjCGNU::GenerateProtocol(const ObjCProtocolDecl *PD) {
       PropertyList->getType(),
       OptionalPropertyList->getType(),
       nullptr);
-  std::vector<llvm::Constant*> Elements;
   // The isa pointer must be set to a magic number so the runtime knows it's
   // the correct layout.
-  Elements.push_back(llvm::ConstantExpr::getIntToPtr(
-        llvm::ConstantInt::get(Int32Ty, ProtocolVersion), IdTy));
-  Elements.push_back(MakeConstantString(ProtocolName, ".objc_protocol_name"));
-  Elements.push_back(ProtocolList);
-  Elements.push_back(InstanceMethodList);
-  Elements.push_back(ClassMethodList);
-  Elements.push_back(OptionalInstanceMethodList);
-  Elements.push_back(OptionalClassMethodList);
-  Elements.push_back(PropertyList);
-  Elements.push_back(OptionalPropertyList);
+  llvm::Constant *Elements[] = {
+      llvm::ConstantExpr::getIntToPtr(
+          llvm::ConstantInt::get(Int32Ty, ProtocolVersion), IdTy),
+      MakeConstantString(ProtocolName, ".objc_protocol_name"), ProtocolList,
+      InstanceMethodList, ClassMethodList, OptionalInstanceMethodList,
+      OptionalClassMethodList, PropertyList, OptionalPropertyList};
   ExistingProtocols[ProtocolName] =
     llvm::ConstantExpr::getBitCast(MakeGlobal(ProtocolTy, Elements,
           CGM.getPointerAlign(), ".objc_protocol"), IdTy);
@@ -2058,20 +2097,20 @@ void CGObjCGNU::GenerateCategory(const ObjCCategoryImplDecl *OCD) {
        E = Protos.end(); I != E; ++I)
     Protocols.push_back((*I)->getNameAsString());
 
-  std::vector<llvm::Constant*> Elements;
-  Elements.push_back(MakeConstantString(CategoryName));
-  Elements.push_back(MakeConstantString(ClassName));
-  // Instance method list
-  Elements.push_back(llvm::ConstantExpr::getBitCast(GenerateMethodList(
-          ClassName, CategoryName, InstanceMethodSels, InstanceMethodTypes,
-          false), PtrTy));
-  // Class method list
-  Elements.push_back(llvm::ConstantExpr::getBitCast(GenerateMethodList(
-          ClassName, CategoryName, ClassMethodSels, ClassMethodTypes, true),
-        PtrTy));
-  // Protocol list
-  Elements.push_back(llvm::ConstantExpr::getBitCast(
-        GenerateProtocolList(Protocols), PtrTy));
+  llvm::Constant *Elements[] = {
+      MakeConstantString(CategoryName), MakeConstantString(ClassName),
+      // Instance method list
+      llvm::ConstantExpr::getBitCast(
+          GenerateMethodList(ClassName, CategoryName, InstanceMethodSels,
+                             InstanceMethodTypes, false),
+          PtrTy),
+      // Class method list
+      llvm::ConstantExpr::getBitCast(GenerateMethodList(ClassName, CategoryName,
+                                                        ClassMethodSels,
+                                                        ClassMethodTypes, true),
+                                     PtrTy),
+      // Protocol list
+      llvm::ConstantExpr::getBitCast(GenerateProtocolList(Protocols), PtrTy)};
   Categories.push_back(llvm::ConstantExpr::getBitCast(
         MakeGlobal(llvm::StructType::get(PtrToInt8Ty, PtrToInt8Ty,
             PtrTy, PtrTy, PtrTy, nullptr), Elements, CGM.getPointerAlign()),
@@ -2167,18 +2206,19 @@ void CGObjCGNU::GenerateClass(const ObjCImplementationDecl *OID) {
 
   // Get the class name
   ObjCInterfaceDecl *ClassDecl =
-    const_cast<ObjCInterfaceDecl *>(OID->getClassInterface());
+      const_cast<ObjCInterfaceDecl *>(OID->getClassInterface());
   std::string ClassName = ClassDecl->getNameAsString();
+
   // Emit the symbol that is used to generate linker errors if this class is
   // referenced in other modules but not declared.
   std::string classSymbolName = "__objc_class_name_" + ClassName;
-  if (llvm::GlobalVariable *symbol =
-      TheModule.getGlobalVariable(classSymbolName)) {
+  if (auto *symbol = TheModule.getGlobalVariable(classSymbolName)) {
     symbol->setInitializer(llvm::ConstantInt::get(LongTy, 0));
   } else {
     new llvm::GlobalVariable(TheModule, LongTy, false,
-    llvm::GlobalValue::ExternalLinkage, llvm::ConstantInt::get(LongTy, 0),
-    classSymbolName);
+                             llvm::GlobalValue::ExternalLinkage,
+                             llvm::ConstantInt::get(LongTy, 0),
+                             classSymbolName);
   }
 
   // Get the size of instances.
@@ -2256,7 +2296,6 @@ void CGObjCGNU::GenerateClass(const ObjCImplementationDecl *OID) {
     MakeGlobalArray(PtrToIntTy, IvarOffsetValues, CGM.getPointerAlign(),
                     ".ivar.offsets");
 
-
   // Collect information about instance methods
   SmallVector<Selector, 16> InstanceMethodSels;
   SmallVector<llvm::Constant*, 16> InstanceMethodTypes;
@@ -2270,7 +2309,6 @@ void CGObjCGNU::GenerateClass(const ObjCImplementationDecl *OID) {
   llvm::Constant *Properties = GeneratePropertyList(OID, InstanceMethodSels,
           InstanceMethodTypes);
 
-
   // Collect information about class methods
   SmallVector<Selector, 16> ClassMethodSels;
   SmallVector<llvm::Constant*, 16> ClassMethodTypes;
@@ -2343,19 +2381,35 @@ void CGObjCGNU::GenerateClass(const ObjCImplementationDecl *OID) {
       ++ivarIndex;
   }
   llvm::Constant *ZeroPtr = llvm::ConstantInt::get(IntPtrTy, 0);
+
   //Generate metaclass for class methods
-  llvm::Constant *MetaClassStruct = GenerateClassStructure(NULLPtr,
-      NULLPtr, 0x12L, ClassName.c_str(), nullptr, Zeros[0], GenerateIvarList(
-        empty, empty, empty), ClassMethodList, NULLPtr,
-      NULLPtr, NULLPtr, ZeroPtr, ZeroPtr, true);
+  llvm::Constant *MetaClassStruct = GenerateClassStructure(
+      NULLPtr, NULLPtr, 0x12L, ClassName.c_str(), nullptr, Zeros[0],
+      GenerateIvarList(empty, empty, empty), ClassMethodList, NULLPtr, NULLPtr,
+      NULLPtr, ZeroPtr, ZeroPtr, true);
+  if (CGM.getTriple().isOSBinFormatCOFF()) {
+    auto Storage = llvm::GlobalValue::DefaultStorageClass;
+    if (OID->getClassInterface()->hasAttr<DLLImportAttr>())
+      Storage = llvm::GlobalValue::DLLImportStorageClass;
+    else if (OID->getClassInterface()->hasAttr<DLLExportAttr>())
+      Storage = llvm::GlobalValue::DLLExportStorageClass;
+    cast<llvm::GlobalValue>(MetaClassStruct)->setDLLStorageClass(Storage);
+  }
 
   // Generate the class structure
-  llvm::Constant *ClassStruct =
-    GenerateClassStructure(MetaClassStruct, SuperClass, 0x11L,
-                           ClassName.c_str(), nullptr,
-      llvm::ConstantInt::get(LongTy, instanceSize), IvarList,
-      MethodList, GenerateProtocolList(Protocols), IvarOffsetArray,
-      Properties, StrongIvarBitmap, WeakIvarBitmap);
+  llvm::Constant *ClassStruct = GenerateClassStructure(
+      MetaClassStruct, SuperClass, 0x11L, ClassName.c_str(), nullptr,
+      llvm::ConstantInt::get(LongTy, instanceSize), IvarList, MethodList,
+      GenerateProtocolList(Protocols), IvarOffsetArray, Properties,
+      StrongIvarBitmap, WeakIvarBitmap);
+  if (CGM.getTriple().isOSBinFormatCOFF()) {
+    auto Storage = llvm::GlobalValue::DefaultStorageClass;
+    if (OID->getClassInterface()->hasAttr<DLLImportAttr>())
+      Storage = llvm::GlobalValue::DLLImportStorageClass;
+    else if (OID->getClassInterface()->hasAttr<DLLExportAttr>())
+      Storage = llvm::GlobalValue::DLLExportStorageClass;
+    cast<llvm::GlobalValue>(ClassStruct)->setDLLStorageClass(Storage);
+  }
 
   // Resolve the class aliases, if they exist.
   if (ClassPtrAlias) {
@@ -2376,7 +2430,6 @@ void CGObjCGNU::GenerateClass(const ObjCImplementationDecl *OID) {
   Classes.push_back(ClassStruct);
 }
 
-
 llvm::Function *CGObjCGNU::ModuleInitFunction() {
   // Only emit an ObjC load function if no Objective-C stuff has been called
   if (Classes.empty() && Categories.empty() && ConstantStrings.empty() &&
@@ -2651,12 +2704,15 @@ llvm::Constant *CGObjCGNU::GetOptimizedPropertySetFunction(bool atomic,
 llvm::Constant *CGObjCGNU::GetGetStructFunction() {
   return GetStructPropertyFn;
 }
+
 llvm::Constant *CGObjCGNU::GetSetStructFunction() {
   return SetStructPropertyFn;
 }
+
 llvm::Constant *CGObjCGNU::GetCppAtomicObjectGetFunction() {
   return nullptr;
 }
+
 llvm::Constant *CGObjCGNU::GetCppAtomicObjectSetFunction() {
   return nullptr;
 }
@@ -2685,7 +2741,6 @@ void CGObjCGNU::EmitTryStmt(CodeGenFunction &CGF,
   // In Objective-C++ mode, we actually emit something equivalent to the C++
   // exception handler. 
   EmitTryCatchStmt(CGF, S, EnterCatchFn, ExitCatchFn, ExceptionReThrowFn);
-  return ;
 }
 
 void CGObjCGNU::EmitThrowStmt(CodeGenFunction &CGF,
@@ -2800,7 +2855,7 @@ llvm::GlobalVariable *CGObjCGNU::ObjCIvarOffsetVariable(
     // to replace it with the real version for a library.  In non-PIC code you
     // must compile with the fragile ABI if you want to use ivars from a
     // GCC-compiled class.
-    if (CGM.getLangOpts().PICLevel || CGM.getLangOpts().PIELevel) {
+    if (CGM.getLangOpts().PICLevel) {
       llvm::GlobalVariable *IvarOffsetGV = new llvm::GlobalVariable(TheModule,
             Int32Ty, false,
             llvm::GlobalValue::PrivateLinkage, OffsetGuess, Name+".guess");
@@ -2848,7 +2903,12 @@ llvm::Value *CGObjCGNU::EmitIvarOffset(CodeGenFunction &CGF,
                          const ObjCIvarDecl *Ivar) {
   if (CGM.getLangOpts().ObjCRuntime.isNonFragile()) {
     Interface = FindIvarInterface(CGM.getContext(), Interface, Ivar);
-    if (RuntimeVersion < 10)
+
+    // The MSVC linker cannot have a single global defined as LinkOnceAnyLinkage
+    // and ExternalLinkage, so create a reference to the ivar global and rely on
+    // the definition being created as part of GenerateClass.
+    if (RuntimeVersion < 10 ||
+        CGF.CGM.getTarget().getTriple().isKnownWindowsMSVCEnvironment())
       return CGF.Builder.CreateZExtOrBitCast(
           CGF.Builder.CreateDefaultAlignedLoad(CGF.Builder.CreateAlignedLoad(
                   ObjCIvarOffsetVariable(Interface, Ivar),
diff --git a/lib/CodeGen/CGObjCMac.cpp b/lib/CodeGen/CGObjCMac.cpp
index e30b2875f209..5ab9fc4f9710 100644
--- a/lib/CodeGen/CGObjCMac.cpp
+++ b/lib/CodeGen/CGObjCMac.cpp
@@ -236,17 +236,14 @@ public:
     CodeGen::CodeGenTypes &Types = CGM.getTypes();
     ASTContext &Ctx = CGM.getContext();
     // id objc_getProperty (id, SEL, ptrdiff_t, bool)
-    SmallVector<CanQualType,4> Params;
     CanQualType IdType = Ctx.getCanonicalParamType(Ctx.getObjCIdType());
     CanQualType SelType = Ctx.getCanonicalParamType(Ctx.getObjCSelType());
-    Params.push_back(IdType);
-    Params.push_back(SelType);
-    Params.push_back(Ctx.getPointerDiffType()->getCanonicalTypeUnqualified());
-    Params.push_back(Ctx.BoolTy);
+    CanQualType Params[] = {
+        IdType, SelType,
+        Ctx.getPointerDiffType()->getCanonicalTypeUnqualified(), Ctx.BoolTy};
     llvm::FunctionType *FTy =
-        Types.GetFunctionType(Types.arrangeLLVMFunctionInfo(
-            IdType, false, false, Params, FunctionType::ExtInfo(),
-            RequiredArgs::All));
+        Types.GetFunctionType(
+          Types.arrangeBuiltinFunctionDeclaration(IdType, Params));
     return CGM.CreateRuntimeFunction(FTy, "objc_getProperty");
   }
 
@@ -254,19 +251,18 @@ public:
     CodeGen::CodeGenTypes &Types = CGM.getTypes();
     ASTContext &Ctx = CGM.getContext();
     // void objc_setProperty (id, SEL, ptrdiff_t, id, bool, bool)
-    SmallVector<CanQualType,6> Params;
     CanQualType IdType = Ctx.getCanonicalParamType(Ctx.getObjCIdType());
     CanQualType SelType = Ctx.getCanonicalParamType(Ctx.getObjCSelType());
-    Params.push_back(IdType);
-    Params.push_back(SelType);
-    Params.push_back(Ctx.getPointerDiffType()->getCanonicalTypeUnqualified());
-    Params.push_back(IdType);
-    Params.push_back(Ctx.BoolTy);
-    Params.push_back(Ctx.BoolTy);
+    CanQualType Params[] = {
+        IdType,
+        SelType,
+        Ctx.getPointerDiffType()->getCanonicalTypeUnqualified(),
+        IdType,
+        Ctx.BoolTy,
+        Ctx.BoolTy};
     llvm::FunctionType *FTy =
-        Types.GetFunctionType(Types.arrangeLLVMFunctionInfo(
-            Ctx.VoidTy, false, false, Params, FunctionType::ExtInfo(),
-            RequiredArgs::All));
+        Types.GetFunctionType(
+          Types.arrangeBuiltinFunctionDeclaration(Ctx.VoidTy, Params));
     return CGM.CreateRuntimeFunction(FTy, "objc_setProperty");
   }
 
@@ -290,9 +286,8 @@ public:
     Params.push_back(IdType);
     Params.push_back(Ctx.getPointerDiffType()->getCanonicalTypeUnqualified());
     llvm::FunctionType *FTy =
-        Types.GetFunctionType(Types.arrangeLLVMFunctionInfo(
-            Ctx.VoidTy, false, false, Params, FunctionType::ExtInfo(),
-            RequiredArgs::All));
+        Types.GetFunctionType(
+          Types.arrangeBuiltinFunctionDeclaration(Ctx.VoidTy, Params));
     const char *name;
     if (atomic && copy)
       name = "objc_setProperty_atomic_copy";
@@ -317,9 +312,8 @@ public:
     Params.push_back(Ctx.BoolTy);
     Params.push_back(Ctx.BoolTy);
     llvm::FunctionType *FTy =
-        Types.GetFunctionType(Types.arrangeLLVMFunctionInfo(
-            Ctx.VoidTy, false, false, Params, FunctionType::ExtInfo(),
-            RequiredArgs::All));
+        Types.GetFunctionType(
+          Types.arrangeBuiltinFunctionDeclaration(Ctx.VoidTy, Params));
     return CGM.CreateRuntimeFunction(FTy, "objc_copyStruct");
   }
   
@@ -336,10 +330,8 @@ public:
     Params.push_back(Ctx.VoidPtrTy);
     Params.push_back(Ctx.VoidPtrTy);
     llvm::FunctionType *FTy =
-      Types.GetFunctionType(Types.arrangeLLVMFunctionInfo(Ctx.VoidTy, false, false,
-                                                          Params,
-                                                          FunctionType::ExtInfo(),
-                                                          RequiredArgs::All));
+        Types.GetFunctionType(
+          Types.arrangeBuiltinFunctionDeclaration(Ctx.VoidTy, Params));
     return CGM.CreateRuntimeFunction(FTy, "objc_copyCppObjectAtomic");
   }
   
@@ -350,12 +342,25 @@ public:
     SmallVector<CanQualType,1> Params;
     Params.push_back(Ctx.getCanonicalParamType(Ctx.getObjCIdType()));
     llvm::FunctionType *FTy =
-        Types.GetFunctionType(Types.arrangeLLVMFunctionInfo(
-            Ctx.VoidTy, false, false, Params, FunctionType::ExtInfo(),
-            RequiredArgs::All));
+        Types.GetFunctionType(
+          Types.arrangeBuiltinFunctionDeclaration(Ctx.VoidTy, Params));
     return CGM.CreateRuntimeFunction(FTy, "objc_enumerationMutation");
   }
 
+  llvm::Constant *getLookUpClassFn() {
+    CodeGen::CodeGenTypes &Types = CGM.getTypes();
+    ASTContext &Ctx = CGM.getContext();
+    // Class objc_lookUpClass (const char *)
+    SmallVector<CanQualType,1> Params;
+    Params.push_back(
+      Ctx.getCanonicalType(Ctx.getPointerType(Ctx.CharTy.withConst())));
+    llvm::FunctionType *FTy =
+        Types.GetFunctionType(Types.arrangeBuiltinFunctionDeclaration(
+                                Ctx.getCanonicalType(Ctx.getObjCClassType()),
+                                Params));
+    return CGM.CreateRuntimeFunction(FTy, "objc_lookUpClass");
+  }
+
   /// GcReadWeakFn -- LLVM objc_read_weak (id *src) function.
   llvm::Constant *getGcReadWeakFn() {
     // id objc_read_weak (id *)
@@ -576,7 +581,6 @@ public:
     return CGM.CreateRuntimeFunction(
       llvm::FunctionType::get(CGM.Int32Ty, params, false),
       "objc_exception_match");
-
   }
 
   /// SetJmpFn - LLVM _setjmp function.
@@ -600,7 +604,6 @@ public:
 /// modern abi
 class ObjCNonFragileABITypesHelper : public ObjCCommonTypesHelper {
 public:
-
   // MethodListnfABITy - LLVM for struct _method_list_t
   llvm::StructType *MethodListnfABITy;
 
@@ -967,7 +970,8 @@ protected:
   llvm::Constant *EmitPropertyList(Twine Name,
                                    const Decl *Container,
                                    const ObjCContainerDecl *OCD,
-                                   const ObjCCommonTypesHelper &ObjCTypes);
+                                   const ObjCCommonTypesHelper &ObjCTypes,
+                                   bool IsClassProperty);
 
   /// EmitProtocolMethodTypes - Generate the array of extended method type 
   /// strings. The return value has type Int8PtrPtrTy.
@@ -981,13 +985,20 @@ protected:
     SmallVectorImpl<llvm::Constant*> &Properties,
     const Decl *Container,
     const ObjCProtocolDecl *Proto,
-    const ObjCCommonTypesHelper &ObjCTypes);
+    const ObjCCommonTypesHelper &ObjCTypes,
+    bool IsClassProperty);
 
   /// GetProtocolRef - Return a reference to the internal protocol
   /// description, creating an empty one if it has not been
   /// defined. The return value has type ProtocolPtrTy.
   llvm::Constant *GetProtocolRef(const ObjCProtocolDecl *PD);
 
+  /// Return a reference to the given Class using runtime calls rather than
+  /// by a symbol reference.
+  llvm::Value *EmitClassRefViaRuntime(CodeGenFunction &CGF,
+                                      const ObjCInterfaceDecl *ID,
+                                      ObjCCommonTypesHelper &ObjCTypes);
+
 public:
   /// CreateMetadataVar - Create a global variable with internal
   /// linkage for use by the Objective-C runtime.
@@ -1079,7 +1090,8 @@ private:
   /// has type ClassExtensionPtrTy.
   llvm::Constant *EmitClassExtension(const ObjCImplementationDecl *ID,
                                      CharUnits instanceSize,
-                                     bool hasMRCWeakIvars);
+                                     bool hasMRCWeakIvars,
+                                     bool isClassProperty);
 
   /// EmitClassRef - Return a Value*, of type ObjCTypes.ClassPtrTy,
   /// for the given class.
@@ -1119,9 +1131,8 @@ private:
 
   /// EmitMethodList - Emit the method list for the given
   /// implementation. The return value has type MethodListPtrTy.
-  llvm::Constant *EmitMethodList(Twine Name,
-                                 const char *Section,
-                                 ArrayRef<llvm::Constant*> Methods);
+  llvm::Constant *EmitMethodList(Twine Name, StringRef Section,
+                                 ArrayRef<llvm::Constant *> Methods);
 
   /// EmitMethodDescList - Emit a method description list for a list of
   /// method declarations.
@@ -1134,9 +1145,8 @@ private:
   ///  - begin, end: The method list to output.
   ///
   /// The return value has type MethodDescriptionListPtrTy.
-  llvm::Constant *EmitMethodDescList(Twine Name,
-                                     const char *Section,
-                                     ArrayRef<llvm::Constant*> Methods);
+  llvm::Constant *EmitMethodDescList(Twine Name, StringRef Section,
+                                     ArrayRef<llvm::Constant *> Methods);
 
   /// GetOrEmitProtocol - Get the protocol object for the given
   /// declaration, emitting it if necessary. The return value has type
@@ -1255,7 +1265,7 @@ public:
 
   /// GetClassGlobal - Return the global variable for the Objective-C
   /// class of the given name.
-  llvm::GlobalVariable *GetClassGlobal(const std::string &Name,
+  llvm::GlobalVariable *GetClassGlobal(StringRef Name,
                                        bool Weak = false) override {
     llvm_unreachable("CGObjCMac::GetClassGlobal");
   }
@@ -1293,9 +1303,8 @@ private:
 
   /// AddModuleClassList - Add the given list of class pointers to the
   /// module with the provided symbol and section names.
-  void AddModuleClassList(ArrayRef<llvm::GlobalValue*> Container,
-                          const char *SymbolName,
-                          const char *SectionName);
+  void AddModuleClassList(ArrayRef<llvm::GlobalValue *> Container,
+                          StringRef SymbolName, StringRef SectionName);
 
   llvm::GlobalVariable * BuildClassRoTInitializer(unsigned flags,
                                               unsigned InstanceStart,
@@ -1314,9 +1323,8 @@ private:
 
   /// EmitMethodList - Emit the method list for the given
   /// implementation. The return value has type MethodListnfABITy.
-  llvm::Constant *EmitMethodList(Twine Name,
-                                 const char *Section,
-                                 ArrayRef<llvm::Constant*> Methods);
+  llvm::Constant *EmitMethodList(Twine Name, StringRef Section,
+                                 ArrayRef<llvm::Constant *> Methods);
   /// EmitIvarList - Emit the ivar list for the given
   /// implementation. If ForClass is true the list of class ivars
   /// (i.e. metaclass ivars) is emitted, otherwise the list of
@@ -1357,7 +1365,7 @@ private:
   
   /// GetClassGlobal - Return the global variable for the Objective-C
   /// class of the given name.
-  llvm::GlobalVariable *GetClassGlobal(const std::string &Name,
+  llvm::GlobalVariable *GetClassGlobal(StringRef Name,
                                        bool Weak = false) override;
 
   /// EmitClassRef - Return a Value*, of type ObjCTypes.ClassPtrTy,
@@ -1398,13 +1406,9 @@ private:
   llvm::Constant *GetInterfaceEHType(const ObjCInterfaceDecl *ID,
                                   bool ForDefinition);
 
-  const char *getMetaclassSymbolPrefix() const {
-    return "OBJC_METACLASS_$_";
-  }
+  StringRef getMetaclassSymbolPrefix() const { return "OBJC_METACLASS_$_"; }
 
-  const char *getClassSymbolPrefix() const {
-    return "OBJC_CLASS_$_";
-  }
+  StringRef getClassSymbolPrefix() const { return "OBJC_CLASS_$_"; }
 
   void GetClassSizeInfo(const ObjCImplementationDecl *OID,
                         uint32_t &InstanceStart,
@@ -1506,12 +1510,15 @@ public:
   llvm::Constant *GetSetStructFunction() override {
     return ObjCTypes.getCopyStructFn();
   }
+
   llvm::Constant *GetGetStructFunction() override {
     return ObjCTypes.getCopyStructFn();
   }
+
   llvm::Constant *GetCppAtomicObjectSetFunction() override {
     return ObjCTypes.getCppAtomicObjectFunction();
   }
+
   llvm::Constant *GetCppAtomicObjectGetFunction() override {
     return ObjCTypes.getCppAtomicObjectFunction();
   }
@@ -1934,7 +1941,7 @@ CGObjCCommonMac::EmitMessageSend(CodeGen::CodeGenFunction &CGF,
   // Emit a null-check if there's a consumed argument other than the receiver.
   bool RequiresNullCheck = false;
   if (ReceiverCanBeNull && CGM.getLangOpts().ObjCAutoRefCount && Method) {
-    for (const auto *ParamDecl : Method->params()) {
+    for (const auto *ParamDecl : Method->parameters()) {
       if (ParamDecl->hasAttr<NSConsumedAttr>()) {
         if (!nullReturn.NullBB)
           nullReturn.init(CGF, Arg0);
@@ -2027,6 +2034,7 @@ namespace {
     bool IsDisordered = false;
 
     llvm::SmallVector<IvarInfo, 8> IvarsInfo;
+
   public:
     IvarLayoutBuilder(CodeGenModule &CGM, CharUnits instanceBegin,
                       CharUnits instanceEnd, bool forStrongLayout)
@@ -2062,7 +2070,7 @@ namespace {
       printf("\n");
     }
   };
-}
+} // end anonymous namespace
 
 llvm::Constant *CGObjCCommonMac::BuildGCBlockLayout(CodeGenModule &CGM,
                                                 const CGBlockInfo &blockInfo) {
@@ -2141,7 +2149,6 @@ void IvarLayoutBuilder::visitBlock(const CGBlockInfo &blockInfo) {
   }
 }
 
-
 /// getBlockCaptureLifetime - This routine returns life time of the captured
 /// block variable for the purpose of block layout meta-data generation. FQT is
 /// the type of the variable captured in the block.
@@ -2629,7 +2636,6 @@ llvm::Constant *CGObjCCommonMac::BuildRCBlockLayout(CodeGenModule &CGM,
   return getBitmapBlockLayout(false);
 }
 
-
 llvm::Constant *CGObjCCommonMac::BuildByrefLayout(CodeGen::CodeGenModule &CGM,
                                                   QualType T) {
   assert(CGM.getLangOpts().getGC() == LangOptions::NonGC);
@@ -2677,6 +2683,25 @@ llvm::Constant *CGObjCCommonMac::GetProtocolRef(const ObjCProtocolDecl *PD) {
   return GetOrEmitProtocolRef(PD);
 }
 
+llvm::Value *CGObjCCommonMac::EmitClassRefViaRuntime(
+               CodeGenFunction &CGF,
+               const ObjCInterfaceDecl *ID,
+               ObjCCommonTypesHelper &ObjCTypes) {
+  llvm::Constant *lookUpClassFn = ObjCTypes.getLookUpClassFn();
+
+  llvm::Value *className =
+      CGF.CGM.GetAddrOfConstantCString(ID->getObjCRuntimeNameAsString())
+        .getPointer();
+  ASTContext &ctx = CGF.CGM.getContext();
+  className =
+      CGF.Builder.CreateBitCast(className,
+                                CGF.ConvertType(
+                                  ctx.getPointerType(ctx.CharTy.withConst())));
+  llvm::CallInst *call = CGF.Builder.CreateCall(lookUpClassFn, className);
+  call->setDoesNotThrow();
+  return call;
+}
+
 /*
 // Objective-C 1.0 extensions
 struct _objc_protocol {
@@ -2798,6 +2823,7 @@ llvm::Constant *CGObjCMac::GetOrEmitProtocolRef(const ObjCProtocolDecl *PD) {
   struct objc_method_description_list *optional_class_methods;
   struct objc_property_list *instance_properties;
   const char ** extendedMethodTypes;
+  struct objc_property_list *class_properties;
   };
 */
 llvm::Constant *
@@ -2816,13 +2842,16 @@ CGObjCMac::EmitProtocolExtension(const ObjCProtocolDecl *PD,
                          "__OBJC,__cat_cls_meth,regular,no_dead_strip",
                          OptClassMethods),
       EmitPropertyList("OBJC_$_PROP_PROTO_LIST_" + PD->getName(), nullptr, PD,
-                       ObjCTypes),
+                       ObjCTypes, false),
       EmitProtocolMethodTypes("OBJC_PROTOCOL_METHOD_TYPES_" + PD->getName(),
-                              MethodTypesExt, ObjCTypes)};
+                              MethodTypesExt, ObjCTypes),
+      EmitPropertyList("OBJC_$_CLASS_PROP_PROTO_LIST_" + PD->getName(), nullptr,
+                       PD, ObjCTypes, true)};
 
   // Return null if no extension bits are used.
   if (Values[1]->isNullValue() && Values[2]->isNullValue() &&
-      Values[3]->isNullValue() && Values[4]->isNullValue())
+      Values[3]->isNullValue() && Values[4]->isNullValue() &&
+      Values[5]->isNullValue())
     return llvm::Constant::getNullValue(ObjCTypes.ProtocolExtensionPtrTy);
 
   llvm::Constant *Init =
@@ -2878,10 +2907,15 @@ PushProtocolProperties(llvm::SmallPtrSet<const IdentifierInfo*,16> &PropertySet,
                        SmallVectorImpl<llvm::Constant *> &Properties,
                        const Decl *Container,
                        const ObjCProtocolDecl *Proto,
-                       const ObjCCommonTypesHelper &ObjCTypes) {
+                       const ObjCCommonTypesHelper &ObjCTypes,
+                       bool IsClassProperty) {
   for (const auto *P : Proto->protocols()) 
-    PushProtocolProperties(PropertySet, Properties, Container, P, ObjCTypes);
+    PushProtocolProperties(PropertySet, Properties, Container, P, ObjCTypes,
+                           IsClassProperty);
+
   for (const auto *PD : Proto->properties()) {
+    if (IsClassProperty != PD->isClassProperty())
+      continue;
     if (!PropertySet.insert(PD->getIdentifier()).second)
       continue;
     llvm::Constant *Prop[] = {
@@ -2907,7 +2941,17 @@ PushProtocolProperties(llvm::SmallPtrSet<const IdentifierInfo*,16> &PropertySet,
 llvm::Constant *CGObjCCommonMac::EmitPropertyList(Twine Name,
                                        const Decl *Container,
                                        const ObjCContainerDecl *OCD,
-                                       const ObjCCommonTypesHelper &ObjCTypes) {
+                                       const ObjCCommonTypesHelper &ObjCTypes,
+                                       bool IsClassProperty) {
+  if (IsClassProperty) {
+    // Make this entry NULL for OS X with deployment target < 10.11, for iOS
+    // with deployment target < 9.0.
+    const llvm::Triple &Triple = CGM.getTarget().getTriple();
+    if ((Triple.isMacOSX() && Triple.isMacOSXVersionLT(10, 11)) ||
+        (Triple.isiOS() && Triple.isOSVersionLT(9)))
+      return llvm::Constant::getNullValue(ObjCTypes.PropertyListPtrTy);
+  }
+
   SmallVector<llvm::Constant *, 16> Properties;
   llvm::SmallPtrSet<const IdentifierInfo*, 16> PropertySet;
 
@@ -2919,10 +2963,15 @@ llvm::Constant *CGObjCCommonMac::EmitPropertyList(Twine Name,
   if (const ObjCInterfaceDecl *OID = dyn_cast<ObjCInterfaceDecl>(OCD))
     for (const ObjCCategoryDecl *ClassExt : OID->known_extensions())
       for (auto *PD : ClassExt->properties()) {
+        if (IsClassProperty != PD->isClassProperty())
+          continue;
         PropertySet.insert(PD->getIdentifier());
         AddProperty(PD);
       }
+
   for (const auto *PD : OCD->properties()) {
+    if (IsClassProperty != PD->isClassProperty())
+      continue;
     // Don't emit duplicate metadata for properties that were already in a
     // class extension.
     if (!PropertySet.insert(PD->getIdentifier()).second)
@@ -2932,11 +2981,13 @@ llvm::Constant *CGObjCCommonMac::EmitPropertyList(Twine Name,
 
   if (const ObjCInterfaceDecl *OID = dyn_cast<ObjCInterfaceDecl>(OCD)) {
     for (const auto *P : OID->all_referenced_protocols())
-      PushProtocolProperties(PropertySet, Properties, Container, P, ObjCTypes);
+      PushProtocolProperties(PropertySet, Properties, Container, P, ObjCTypes,
+                             IsClassProperty);
   }
   else if (const ObjCCategoryDecl *CD = dyn_cast<ObjCCategoryDecl>(OCD)) {
     for (const auto *P : CD->protocols())
-      PushProtocolProperties(PropertySet, Properties, Container, P, ObjCTypes);
+      PushProtocolProperties(PropertySet, Properties, Container, P, ObjCTypes,
+                             IsClassProperty);
   }
 
   // Return null for empty list.
@@ -3001,8 +3052,8 @@ CGObjCMac::GetMethodDescriptionConstant(const ObjCMethodDecl *MD) {
 }
 
 llvm::Constant *
-CGObjCMac::EmitMethodDescList(Twine Name, const char *Section,
-                              ArrayRef<llvm::Constant*> Methods) {
+CGObjCMac::EmitMethodDescList(Twine Name, StringRef Section,
+                              ArrayRef<llvm::Constant *> Methods) {
   // Return null for empty list.
   if (Methods.empty())
     return llvm::Constant::getNullValue(ObjCTypes.MethodDescriptionListPtrTy);
@@ -3029,6 +3080,7 @@ CGObjCMac::EmitMethodDescList(Twine Name, const char *Section,
   struct _objc_protocol_list *protocols;
   uint32_t size; // <rdar://4585769>
   struct _objc_property_list *instance_properties;
+  struct _objc_property_list *class_properties;
   };
 */
 void CGObjCMac::GenerateCategory(const ObjCCategoryImplDecl *OCD) {
@@ -3055,7 +3107,7 @@ void CGObjCMac::GenerateCategory(const ObjCCategoryImplDecl *OCD) {
     // Class methods should always be defined.
     ClassMethods.push_back(GetMethodConstant(I));
 
-  llvm::Constant *Values[7];
+  llvm::Constant *Values[8];
   Values[0] = GetClassName(OCD->getName());
   Values[1] = GetClassName(Interface->getObjCRuntimeNameAsString());
   LazySymbols.insert(Interface->getIdentifier());
@@ -3077,9 +3129,12 @@ void CGObjCMac::GenerateCategory(const ObjCCategoryImplDecl *OCD) {
   // If there is no category @interface then there can be no properties.
   if (Category) {
     Values[6] = EmitPropertyList("\01l_OBJC_$_PROP_LIST_" + ExtName.str(),
-                                 OCD, Category, ObjCTypes);
+                                 OCD, Category, ObjCTypes, false);
+    Values[7] = EmitPropertyList("\01l_OBJC_$_CLASS_PROP_LIST_" + ExtName.str(),
+                                 OCD, Category, ObjCTypes, true);
   } else {
     Values[6] = llvm::Constant::getNullValue(ObjCTypes.PropertyListPtrTy);
+    Values[7] = llvm::Constant::getNullValue(ObjCTypes.PropertyListPtrTy);
   }
 
   llvm::Constant *Init = llvm::ConstantStruct::get(ObjCTypes.CategoryTy,
@@ -3274,7 +3329,8 @@ void CGObjCMac::GenerateClass(const ObjCImplementationDecl *ID) {
   Values[ 8] = llvm::Constant::getNullValue(ObjCTypes.CachePtrTy);
   Values[ 9] = Protocols;
   Values[10] = BuildStrongIvarLayout(ID, CharUnits::Zero(), Size);
-  Values[11] = EmitClassExtension(ID, Size, hasMRCWeak);
+  Values[11] = EmitClassExtension(ID, Size, hasMRCWeak,
+                                  false/*isClassProperty*/);
   llvm::Constant *Init = llvm::ConstantStruct::get(ObjCTypes.ClassTy,
                                                    Values);
   std::string Name("OBJC_CLASS_");
@@ -3338,8 +3394,9 @@ llvm::Constant *CGObjCMac::EmitMetaClass(const ObjCImplementationDecl *ID,
   Values[ 9] = Protocols;
   // ivar_layout for metaclass is always NULL.
   Values[10] = llvm::Constant::getNullValue(ObjCTypes.Int8PtrTy);
-  // The class extension is always unused for metaclasses.
-  Values[11] = llvm::Constant::getNullValue(ObjCTypes.ClassExtensionPtrTy);
+  // The class extension is used to store class properties for metaclasses.
+  Values[11] = EmitClassExtension(ID, CharUnits::Zero(), false/*hasMRCWeak*/,
+                                  true/*isClassProperty*/);
   llvm::Constant *Init = llvm::ConstantStruct::get(ObjCTypes.ClassTy,
                                                    Values);
 
@@ -3413,19 +3470,28 @@ llvm::Value *CGObjCMac::EmitSuperClassRef(const ObjCInterfaceDecl *ID) {
 */
 llvm::Constant *
 CGObjCMac::EmitClassExtension(const ObjCImplementationDecl *ID,
-                              CharUnits InstanceSize, bool hasMRCWeakIvars) {
+                              CharUnits InstanceSize, bool hasMRCWeakIvars,
+                              bool isClassProperty) {
   uint64_t Size =
     CGM.getDataLayout().getTypeAllocSize(ObjCTypes.ClassExtensionTy);
 
   llvm::Constant *Values[3];
   Values[0] = llvm::ConstantInt::get(ObjCTypes.IntTy, Size);
-  Values[1] = BuildWeakIvarLayout(ID, CharUnits::Zero(), InstanceSize,
-                                  hasMRCWeakIvars);
-  Values[2] = EmitPropertyList("\01l_OBJC_$_PROP_LIST_" + ID->getName(),
-                               ID, ID->getClassInterface(), ObjCTypes);
+  if (isClassProperty) {
+    llvm::Type *PtrTy = CGM.Int8PtrTy;
+    Values[1] = llvm::Constant::getNullValue(PtrTy);
+  } else
+    Values[1] = BuildWeakIvarLayout(ID, CharUnits::Zero(), InstanceSize,
+                                    hasMRCWeakIvars);
+  if (isClassProperty)
+    Values[2] = EmitPropertyList("\01l_OBJC_$_CLASS_PROP_LIST_" + ID->getName(),
+                                 ID, ID->getClassInterface(), ObjCTypes, true);
+  else
+    Values[2] = EmitPropertyList("\01l_OBJC_$_PROP_LIST_" + ID->getName(),
+                                 ID, ID->getClassInterface(), ObjCTypes, false);
 
   // Return null if no extension bits are used.
-  if (Values[1]->isNullValue() && Values[2]->isNullValue())
+  if ((!Values[1] || Values[1]->isNullValue()) && Values[2]->isNullValue())
     return llvm::Constant::getNullValue(ObjCTypes.ClassExtensionPtrTy);
 
   llvm::Constant *Init =
@@ -3530,9 +3596,8 @@ llvm::Constant *CGObjCMac::GetMethodConstant(const ObjCMethodDecl *MD) {
   return llvm::ConstantStruct::get(ObjCTypes.MethodTy, Method);
 }
 
-llvm::Constant *CGObjCMac::EmitMethodList(Twine Name,
-                                          const char *Section,
-                                          ArrayRef<llvm::Constant*> Methods) {
+llvm::Constant *CGObjCMac::EmitMethodList(Twine Name, StringRef Section,
+                                          ArrayRef<llvm::Constant *> Methods) {
   // Return null for empty list.
   if (Methods.empty())
     return llvm::Constant::getNullValue(ObjCTypes.MethodListPtrTy);
@@ -3607,6 +3672,7 @@ llvm::Constant *CGObjCMac::GetOptimizedPropertySetFunction(bool atomic,
 llvm::Constant *CGObjCMac::GetGetStructFunction() {
   return ObjCTypes.getCopyStructFn();
 }
+
 llvm::Constant *CGObjCMac::GetSetStructFunction() {
   return ObjCTypes.getCopyStructFn();
 }
@@ -3614,6 +3680,7 @@ llvm::Constant *CGObjCMac::GetSetStructFunction() {
 llvm::Constant *CGObjCMac::GetCppAtomicObjectGetFunction() {
   return ObjCTypes.getCppAtomicObjectFunction();
 }
+
 llvm::Constant *CGObjCMac::GetCppAtomicObjectSetFunction() {
   return ObjCTypes.getCppAtomicObjectFunction();
 }
@@ -3711,7 +3778,7 @@ namespace {
     void emitWriteHazard();
     void emitHazardsInNewBlocks();
   };
-}
+} // end anonymous namespace
 
 /// Create the fragile-ABI read and write hazards based on the current
 /// state of the function, which is presumed to be immediately prior
@@ -4332,7 +4399,6 @@ void CGObjCMac::EmitObjCWeakAssign(CodeGen::CodeGenFunction &CGF,
   llvm::Value *args[] = { src, dst.getPointer() };
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignWeakFn(),
                               args, "weakassign");
-  return;
 }
 
 /// EmitObjCGlobalAssign - Code gen for assigning to a __strong object.
@@ -4358,7 +4424,6 @@ void CGObjCMac::EmitObjCGlobalAssign(CodeGen::CodeGenFunction &CGF,
   else
     CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignThreadLocalFn(),
                                 args, "threadlocalassign");
-  return;
 }
 
 /// EmitObjCIvarAssign - Code gen for assigning to a __strong object.
@@ -4380,7 +4445,6 @@ void CGObjCMac::EmitObjCIvarAssign(CodeGen::CodeGenFunction &CGF,
   dst = CGF.Builder.CreateBitCast(dst, ObjCTypes.PtrObjectPtrTy);
   llvm::Value *args[] = { src, dst.getPointer(), ivarOffset };
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignIvarFn(), args);
-  return;
 }
 
 /// EmitObjCStrongCastAssign - Code gen for assigning to a __strong cast object.
@@ -4401,7 +4465,6 @@ void CGObjCMac::EmitObjCStrongCastAssign(CodeGen::CodeGenFunction &CGF,
   llvm::Value *args[] = { src, dst.getPointer() };
   CGF.EmitNounwindRuntimeCall(ObjCTypes.getGcAssignStrongCastFn(),
                               args, "strongassign");
-  return;
 }
 
 void CGObjCMac::EmitGCMemmoveCollectable(CodeGen::CodeGenFunction &CGF,
@@ -4455,7 +4518,8 @@ enum ImageInfoFlags {
   // A flag indicating that the module has no instances of a @synthesize of a
   // superclass variable. <rdar://problem/6803242>
   eImageInfo_CorrectedSynthesize = (1 << 4), // This flag is no longer set by clang.
-  eImageInfo_ImageIsSimulated    = (1 << 5)
+  eImageInfo_ImageIsSimulated    = (1 << 5),
+  eImageInfo_ClassProperties     = (1 << 6)
 };
 
 void CGObjCCommonMac::EmitImageInfo() {
@@ -4507,6 +4571,10 @@ void CGObjCCommonMac::EmitImageInfo() {
        Triple.getArch() == llvm::Triple::x86_64))
     Mod.addModuleFlag(llvm::Module::Error, "Objective-C Is Simulated",
                       eImageInfo_ImageIsSimulated);
+
+  // Indicate whether we are generating class properties.
+  Mod.addModuleFlag(llvm::Module::Error, "Objective-C Class Properties",
+                    eImageInfo_ClassProperties);
 }
 
 // struct objc_module {
@@ -4602,6 +4670,11 @@ llvm::Value *CGObjCMac::EmitClassRefFromId(CodeGenFunction &CGF,
 
 llvm::Value *CGObjCMac::EmitClassRef(CodeGenFunction &CGF,
                                      const ObjCInterfaceDecl *ID) {
+  // If the class has the objc_runtime_visible attribute, we need to
+  // use the Objective-C runtime to get the class.
+  if (ID->hasAttr<ObjCRuntimeVisibleAttr>())
+    return EmitClassRefViaRuntime(CGF, ID, ObjCTypes);
+
   return EmitClassRefFromId(CGF, ID->getIdentifier());
 }
 
@@ -4954,7 +5027,7 @@ CGObjCCommonMac::BuildIvarLayout(const ObjCImplementationDecl *OMD,
       baseOffset = CharUnits::Zero();
     }
 
-    baseOffset = baseOffset.RoundUpToAlignment(CGM.getPointerAlign());
+    baseOffset = baseOffset.alignTo(CGM.getPointerAlign());
   }
   else {
     CGM.getContext().DeepCollectObjCIvars(OI, true, ivars);
@@ -5131,9 +5204,8 @@ void CGObjCMac::FinishModule() {
 }
 
 CGObjCNonFragileABIMac::CGObjCNonFragileABIMac(CodeGen::CodeGenModule &cgm)
-  : CGObjCCommonMac(cgm),
-    ObjCTypes(cgm) {
-  ObjCEmptyCacheVar = ObjCEmptyVtableVar = nullptr;
+    : CGObjCCommonMac(cgm), ObjCTypes(cgm), ObjCEmptyCacheVar(nullptr),
+      ObjCEmptyVtableVar(nullptr) {
   ObjCABI = 2;
 }
 
@@ -5223,7 +5295,6 @@ ObjCCommonTypesHelper::ObjCCommonTypesHelper(CodeGen::CodeGenModule &cgm)
   // struct _objc_cache *
   CacheTy = llvm::StructType::create(VMContext, "struct._objc_cache");
   CachePtrTy = llvm::PointerType::getUnqual(CacheTy);
-    
 }
 
 ObjCTypesHelper::ObjCTypesHelper(CodeGen::CodeGenModule &cgm)
@@ -5256,12 +5327,13 @@ ObjCTypesHelper::ObjCTypesHelper(CodeGen::CodeGenModule &cgm)
   //   struct _objc_method_description_list *optional_class_methods;
   //   struct _objc_property_list *instance_properties;
   //   const char ** extendedMethodTypes;
+  //   struct _objc_property_list *class_properties;
   // }
   ProtocolExtensionTy =
     llvm::StructType::create("struct._objc_protocol_extension",
                              IntTy, MethodDescriptionListPtrTy,
                              MethodDescriptionListPtrTy, PropertyListPtrTy,
-                             Int8PtrPtrTy, nullptr);
+                             Int8PtrPtrTy, PropertyListPtrTy, nullptr);
 
   // struct _objc_protocol_extension *
   ProtocolExtensionPtrTy = llvm::PointerType::getUnqual(ProtocolExtensionTy);
@@ -5359,14 +5431,17 @@ ObjCTypesHelper::ObjCTypesHelper(CodeGen::CodeGenModule &cgm)
   //   char *class_name;
   //   struct _objc_method_list *instance_method;
   //   struct _objc_method_list *class_method;
+  //   struct _objc_protocol_list *protocols;
   //   uint32_t size;  // sizeof(struct _objc_category)
   //   struct _objc_property_list *instance_properties;// category's @property
+  //   struct _objc_property_list *class_properties;
   // }
   CategoryTy =
     llvm::StructType::create("struct._objc_category",
                              Int8PtrTy, Int8PtrTy, MethodListPtrTy,
                              MethodListPtrTy, ProtocolListPtrTy,
-                             IntTy, PropertyListPtrTy, nullptr);
+                             IntTy, PropertyListPtrTy, PropertyListPtrTy,
+                             nullptr);
 
   // Global metadata structures
 
@@ -5405,7 +5480,6 @@ ObjCTypesHelper::ObjCTypesHelper(CodeGen::CodeGenModule &cgm)
     llvm::StructType::create("struct._objc_exception_data",
                              llvm::ArrayType::get(CGM.Int32Ty,SetJmpBufferSize),
                              StackPtrTy, nullptr);
-
 }
 
 ObjCNonFragileABITypesHelper::ObjCNonFragileABITypesHelper(CodeGen::CodeGenModule &cgm)
@@ -5434,6 +5508,7 @@ ObjCNonFragileABITypesHelper::ObjCNonFragileABITypesHelper(CodeGen::CodeGenModul
   //   const uint32_t flags;  // = 0
   //   const char ** extendedMethodTypes;
   //   const char *demangledName;
+  //   const struct _prop_list_t * class_properties;
   // }
 
   // Holder for struct _protocol_list_t *
@@ -5446,7 +5521,7 @@ ObjCNonFragileABITypesHelper::ObjCNonFragileABITypesHelper(CodeGen::CodeGenModul
                              MethodListnfABIPtrTy, MethodListnfABIPtrTy,
                              MethodListnfABIPtrTy, MethodListnfABIPtrTy,
                              PropertyListPtrTy, IntTy, IntTy, Int8PtrPtrTy,
-                             Int8PtrTy,
+                             Int8PtrTy, PropertyListPtrTy,
                              nullptr);
 
   // struct _protocol_t*
@@ -5539,6 +5614,8 @@ ObjCNonFragileABITypesHelper::ObjCNonFragileABITypesHelper(CodeGen::CodeGenModul
   //   const struct _method_list_t * const class_methods;
   //   const struct _protocol_list_t * const protocols;
   //   const struct _prop_list_t * const properties;
+  //   const struct _prop_list_t * const class_properties;
+  //   const uint32_t size;
   // }
   CategorynfABITy = llvm::StructType::create("struct._category_t",
                                              Int8PtrTy, ClassnfABIPtrTy,
@@ -5546,6 +5623,8 @@ ObjCNonFragileABITypesHelper::ObjCNonFragileABITypesHelper(CodeGen::CodeGenModul
                                              MethodListnfABIPtrTy,
                                              ProtocolListnfABIPtrTy,
                                              PropertyListPtrTy,
+                                             PropertyListPtrTy,
+                                             IntTy,
                                              nullptr);
 
   // New types for nonfragile abi messaging.
@@ -5609,10 +5688,9 @@ llvm::Function *CGObjCNonFragileABIMac::ModuleInitFunction() {
   return nullptr;
 }
 
-void CGObjCNonFragileABIMac::
-AddModuleClassList(ArrayRef<llvm::GlobalValue*> Container,
-                   const char *SymbolName,
-                   const char *SectionName) {
+void CGObjCNonFragileABIMac::AddModuleClassList(
+    ArrayRef<llvm::GlobalValue *> Container, StringRef SymbolName,
+    StringRef SectionName) {
   unsigned NumClasses = Container.size();
 
   if (!NumClasses)
@@ -5814,13 +5892,16 @@ llvm::GlobalVariable * CGObjCNonFragileABIMac::BuildClassRoTInitializer(
   if (flags & NonFragileABI_Class_Meta) {
     Values[ 7] = llvm::Constant::getNullValue(ObjCTypes.IvarListnfABIPtrTy);
     Values[ 8] = GetIvarLayoutName(nullptr, ObjCTypes);
-    Values[ 9] = llvm::Constant::getNullValue(ObjCTypes.PropertyListPtrTy);
+    Values[ 9] = EmitPropertyList(
+        "\01l_OBJC_$_CLASS_PROP_LIST_" + ID->getObjCRuntimeNameAsString(),
+        ID, ID->getClassInterface(), ObjCTypes, true);
   } else {
     Values[ 7] = EmitIvarList(ID);
     Values[ 8] = BuildWeakIvarLayout(ID, beginInstance, endInstance,
                                      hasMRCWeak);
-    Values[ 9] = EmitPropertyList("\01l_OBJC_$_PROP_LIST_" + ID->getObjCRuntimeNameAsString(),
-                                  ID, ID->getClassInterface(), ObjCTypes);
+    Values[ 9] = EmitPropertyList(
+        "\01l_OBJC_$_PROP_LIST_" + ID->getObjCRuntimeNameAsString(),
+        ID, ID->getClassInterface(), ObjCTypes, false);
   }
   llvm::Constant *Init = llvm::ConstantStruct::get(ObjCTypes.ClassRonfABITy,
                                                    Values);
@@ -5870,8 +5951,9 @@ llvm::GlobalVariable *CGObjCNonFragileABIMac::BuildClassMetaData(
   GV->setSection("__DATA, __objc_data");
   GV->setAlignment(
     CGM.getDataLayout().getABITypeAlignment(ObjCTypes.ClassnfABITy));
-  if (HiddenVisibility)
-    GV->setVisibility(llvm::GlobalValue::HiddenVisibility);
+  if (!CGM.getTriple().isOSBinFormatCOFF())
+    if (HiddenVisibility)
+      GV->setVisibility(llvm::GlobalValue::HiddenVisibility);
   return GV;
 }
 
@@ -5896,49 +5978,60 @@ void CGObjCNonFragileABIMac::GetClassSizeInfo(const ObjCImplementationDecl *OID,
     InstanceStart = RL.getFieldOffset(0) / CGM.getContext().getCharWidth();
 }
 
+static llvm::GlobalValue::DLLStorageClassTypes getStorage(CodeGenModule &CGM,
+                                                          StringRef Name) {
+  IdentifierInfo &II = CGM.getContext().Idents.get(Name);
+  TranslationUnitDecl *TUDecl = CGM.getContext().getTranslationUnitDecl();
+  DeclContext *DC = TranslationUnitDecl::castToDeclContext(TUDecl);
+
+  const VarDecl *VD = nullptr;
+  for (const auto &Result : DC->lookup(&II))
+    if ((VD = dyn_cast<VarDecl>(Result)))
+      break;
+
+  if (!VD)
+    return llvm::GlobalValue::DLLImportStorageClass;
+  if (VD->hasAttr<DLLExportAttr>())
+    return llvm::GlobalValue::DLLExportStorageClass;
+  if (VD->hasAttr<DLLImportAttr>())
+    return llvm::GlobalValue::DLLImportStorageClass;
+  return llvm::GlobalValue::DefaultStorageClass;
+}
+
 void CGObjCNonFragileABIMac::GenerateClass(const ObjCImplementationDecl *ID) {
-  std::string ClassName = ID->getObjCRuntimeNameAsString();
   if (!ObjCEmptyCacheVar) {
-    ObjCEmptyCacheVar = new llvm::GlobalVariable(
-      CGM.getModule(),
-      ObjCTypes.CacheTy,
-      false,
-      llvm::GlobalValue::ExternalLinkage,
-      nullptr,
-      "_objc_empty_cache");
-
-    // Make this entry NULL for any iOS device target, any iOS simulator target,
-    // OS X with deployment target 10.9 or later.
+    ObjCEmptyCacheVar =
+        new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.CacheTy, false,
+                                 llvm::GlobalValue::ExternalLinkage, nullptr,
+                                 "_objc_empty_cache");
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      ObjCEmptyCacheVar->setDLLStorageClass(getStorage(CGM, "_objc_empty_cache"));
+
+    // Only OS X with deployment version <10.9 use the empty vtable symbol
     const llvm::Triple &Triple = CGM.getTarget().getTriple();
-    if (Triple.isiOS() || Triple.isWatchOS() ||
-        (Triple.isMacOSX() && !Triple.isMacOSXVersionLT(10, 9)))
-      // This entry will be null.
-      ObjCEmptyVtableVar = nullptr;
-    else
-      ObjCEmptyVtableVar = new llvm::GlobalVariable(
-                                                    CGM.getModule(),
-                                                    ObjCTypes.ImpnfABITy,
-                                                    false,
-                                                    llvm::GlobalValue::ExternalLinkage,
-                                                    nullptr,
-                                                    "_objc_empty_vtable");
-  }
-  assert(ID->getClassInterface() &&
-         "CGObjCNonFragileABIMac::GenerateClass - class is 0");
+    if (Triple.isMacOSX() && Triple.isMacOSXVersionLT(10, 9))
+      ObjCEmptyVtableVar =
+          new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.ImpnfABITy, false,
+                                   llvm::GlobalValue::ExternalLinkage, nullptr,
+                                   "_objc_empty_vtable");
+  }
+
   // FIXME: Is this correct (that meta class size is never computed)?
   uint32_t InstanceStart =
     CGM.getDataLayout().getTypeAllocSize(ObjCTypes.ClassnfABITy);
   uint32_t InstanceSize = InstanceStart;
   uint32_t flags = NonFragileABI_Class_Meta;
-  llvm::SmallString<64> ObjCMetaClassName(getMetaclassSymbolPrefix());
-  llvm::SmallString<64> ObjCClassName(getClassSymbolPrefix());
-  llvm::SmallString<64> TClassName;
 
   llvm::GlobalVariable *SuperClassGV, *IsAGV;
 
+  StringRef ClassName = ID->getObjCRuntimeNameAsString();
+  const auto *CI = ID->getClassInterface();
+  assert(CI && "CGObjCNonFragileABIMac::GenerateClass - class is 0");
+
   // Build the flags for the metaclass.
-  bool classIsHidden =
-    ID->getClassInterface()->getVisibility() == HiddenVisibility;
+  bool classIsHidden = (CGM.getTriple().isOSBinFormatCOFF())
+                           ? !CI->hasAttr<DLLExportAttr>()
+                           : CI->getVisibility() == HiddenVisibility;
   if (classIsHidden)
     flags |= NonFragileABI_Class_Hidden;
 
@@ -5947,45 +6040,59 @@ void CGObjCNonFragileABIMac::GenerateClass(const ObjCImplementationDecl *ID) {
   if (ID->hasNonZeroConstructors() || ID->hasDestructors()) {
     flags |= NonFragileABI_Class_HasCXXStructors;
     if (!ID->hasNonZeroConstructors())
-      flags |= NonFragileABI_Class_HasCXXDestructorOnly;  
+      flags |= NonFragileABI_Class_HasCXXDestructorOnly;
   }
 
-  if (!ID->getClassInterface()->getSuperClass()) {
+  if (!CI->getSuperClass()) {
     // class is root
     flags |= NonFragileABI_Class_Root;
-    TClassName = ObjCClassName;
-    TClassName += ClassName;
-    SuperClassGV = GetClassGlobal(TClassName.str(),
-                                  ID->getClassInterface()->isWeakImported());
-    TClassName = ObjCMetaClassName;
-    TClassName += ClassName;
-    IsAGV = GetClassGlobal(TClassName.str(),
-                           ID->getClassInterface()->isWeakImported());
+
+    SuperClassGV = GetClassGlobal((getClassSymbolPrefix() + ClassName).str(),
+                                  CI->isWeakImported());
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      if (CI->hasAttr<DLLImportAttr>())
+        SuperClassGV->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
+
+    IsAGV = GetClassGlobal((getMetaclassSymbolPrefix() + ClassName).str(),
+                           CI->isWeakImported());
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      if (CI->hasAttr<DLLImportAttr>())
+        IsAGV->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
   } else {
     // Has a root. Current class is not a root.
     const ObjCInterfaceDecl *Root = ID->getClassInterface();
     while (const ObjCInterfaceDecl *Super = Root->getSuperClass())
       Root = Super;
-    TClassName = ObjCMetaClassName ;
-    TClassName += Root->getObjCRuntimeNameAsString();
-    IsAGV = GetClassGlobal(TClassName.str(),
+
+    const auto *Super = CI->getSuperClass();
+    StringRef RootClassName = Root->getObjCRuntimeNameAsString();
+    StringRef SuperClassName = Super->getObjCRuntimeNameAsString();
+
+    IsAGV = GetClassGlobal((getMetaclassSymbolPrefix() + RootClassName).str(),
                            Root->isWeakImported());
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      if (Root->hasAttr<DLLImportAttr>())
+        IsAGV->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
 
     // work on super class metadata symbol.
-    TClassName = ObjCMetaClassName;
-    TClassName += ID->getClassInterface()->getSuperClass()->getObjCRuntimeNameAsString();
-    SuperClassGV = GetClassGlobal(
-                                  TClassName.str(),
-                                  ID->getClassInterface()->getSuperClass()->isWeakImported());
-  }
-  llvm::GlobalVariable *CLASS_RO_GV = BuildClassRoTInitializer(flags,
-                                                               InstanceStart,
-                                                               InstanceSize,ID);
-  TClassName = ObjCMetaClassName;
-  TClassName += ClassName;
-  llvm::GlobalVariable *MetaTClass = BuildClassMetaData(
-      TClassName.str(), IsAGV, SuperClassGV, CLASS_RO_GV, classIsHidden,
-      ID->getClassInterface()->isWeakImported());
+    SuperClassGV =
+        GetClassGlobal((getMetaclassSymbolPrefix() + SuperClassName).str(),
+                       Super->isWeakImported());
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      if (Super->hasAttr<DLLImportAttr>())
+        SuperClassGV->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
+  }
+
+  llvm::GlobalVariable *CLASS_RO_GV =
+      BuildClassRoTInitializer(flags, InstanceStart, InstanceSize, ID);
+
+  llvm::GlobalVariable *MetaTClass =
+      BuildClassMetaData((getMetaclassSymbolPrefix() + ClassName).str(), IsAGV,
+                         SuperClassGV, CLASS_RO_GV, classIsHidden,
+                         CI->isWeakImported());
+  if (CGM.getTriple().isOSBinFormatCOFF())
+    if (CI->hasAttr<DLLExportAttr>())
+      MetaTClass->setDLLStorageClass(llvm::GlobalValue::DLLExportStorageClass);
   DefinedMetaClasses.push_back(MetaTClass);
 
   // Metadata for the class
@@ -6006,34 +6113,38 @@ void CGObjCNonFragileABIMac::GenerateClass(const ObjCImplementationDecl *ID) {
       flags |= NonFragileABI_Class_HasCXXDestructorOnly;
   }
 
-  if (hasObjCExceptionAttribute(CGM.getContext(), ID->getClassInterface()))
+  if (hasObjCExceptionAttribute(CGM.getContext(), CI))
     flags |= NonFragileABI_Class_Exception;
 
-  if (!ID->getClassInterface()->getSuperClass()) {
+  if (!CI->getSuperClass()) {
     flags |= NonFragileABI_Class_Root;
     SuperClassGV = nullptr;
   } else {
     // Has a root. Current class is not a root.
-    TClassName = ObjCClassName;
-    TClassName += ID->getClassInterface()->getSuperClass()->getObjCRuntimeNameAsString();
-    SuperClassGV = GetClassGlobal(
-                                  TClassName.str(),
-                                  ID->getClassInterface()->getSuperClass()->isWeakImported());
+    const auto *Super = CI->getSuperClass();
+    StringRef SuperClassName = Super->getObjCRuntimeNameAsString();
+
+    SuperClassGV =
+        GetClassGlobal((getClassSymbolPrefix() + SuperClassName).str(),
+                       Super->isWeakImported());
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      if (Super->hasAttr<DLLImportAttr>())
+        SuperClassGV->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
   }
+
   GetClassSizeInfo(ID, InstanceStart, InstanceSize);
-  CLASS_RO_GV = BuildClassRoTInitializer(flags,
-                                         InstanceStart,
-                                         InstanceSize,
-                                         ID);
+  CLASS_RO_GV =
+      BuildClassRoTInitializer(flags, InstanceStart, InstanceSize, ID);
 
-  TClassName = ObjCClassName;
-  TClassName += ClassName;
   llvm::GlobalVariable *ClassMD =
-    BuildClassMetaData(TClassName.str(), MetaTClass, SuperClassGV, CLASS_RO_GV,
-                       classIsHidden,
-                       ID->getClassInterface()->isWeakImported());
+    BuildClassMetaData((getClassSymbolPrefix() + ClassName).str(), MetaTClass,
+                       SuperClassGV, CLASS_RO_GV, classIsHidden,
+                       CI->isWeakImported());
+  if (CGM.getTriple().isOSBinFormatCOFF())
+    if (CI->hasAttr<DLLExportAttr>())
+      ClassMD->setDLLStorageClass(llvm::GlobalValue::DLLExportStorageClass);
   DefinedClasses.push_back(ClassMD);
-  ImplementedClasses.push_back(ID->getClassInterface());
+  ImplementedClasses.push_back(CI);
 
   // Determine if this class is also "non-lazy".
   if (ImplementationIsNonLazy(ID))
@@ -6041,7 +6152,7 @@ void CGObjCNonFragileABIMac::GenerateClass(const ObjCImplementationDecl *ID) {
 
   // Force the definition of the EHType if necessary.
   if (flags & NonFragileABI_Class_Exception)
-    GetInterfaceEHType(ID->getClassInterface(), true);
+    GetInterfaceEHType(CI, true);
   // Make sure method definition entries are all clear for next implementation.
   MethodDefinitions.clear();
 }
@@ -6093,6 +6204,8 @@ llvm::Value *CGObjCNonFragileABIMac::GenerateProtocolRef(CodeGenFunction &CGF,
 ///   const struct _method_list_t * const class_methods;
 ///   const struct _protocol_list_t * const protocols;
 ///   const struct _prop_list_t * const properties;
+///   const struct _prop_list_t * const class_properties;
+///   const uint32_t size;
 /// }
 ///
 void CGObjCNonFragileABIMac::GenerateCategory(const ObjCCategoryImplDecl *OCD) {
@@ -6107,7 +6220,7 @@ void CGObjCNonFragileABIMac::GenerateCategory(const ObjCCategoryImplDecl *OCD) {
   llvm::SmallString<64> ExtClassName(getClassSymbolPrefix());
   ExtClassName += Interface->getObjCRuntimeNameAsString();
 
-  llvm::Constant *Values[6];
+  llvm::Constant *Values[8];
   Values[0] = GetClassName(OCD->getIdentifier()->getName());
   // meta-class entry symbol
   llvm::GlobalVariable *ClassGV =
@@ -6156,12 +6269,18 @@ void CGObjCNonFragileABIMac::GenerateCategory(const ObjCCategoryImplDecl *OCD) {
                                    Category->protocol_begin(),
                                    Category->protocol_end());
     Values[5] = EmitPropertyList("\01l_OBJC_$_PROP_LIST_" + ExtName.str(),
-                                 OCD, Category, ObjCTypes);
+                                 OCD, Category, ObjCTypes, false);
+    Values[6] = EmitPropertyList("\01l_OBJC_$_CLASS_PROP_LIST_" + ExtName.str(),
+                                 OCD, Category, ObjCTypes, true);
   } else {
     Values[4] = llvm::Constant::getNullValue(ObjCTypes.ProtocolListnfABIPtrTy);
     Values[5] = llvm::Constant::getNullValue(ObjCTypes.PropertyListPtrTy);
+    Values[6] = llvm::Constant::getNullValue(ObjCTypes.PropertyListPtrTy);
   }
 
+  unsigned Size = CGM.getDataLayout().getTypeAllocSize(ObjCTypes.CategorynfABITy);
+  Values[7] = llvm::ConstantInt::get(ObjCTypes.IntTy, Size);
+
   llvm::Constant *Init =
     llvm::ConstantStruct::get(ObjCTypes.CategorynfABITy,
                               Values);
@@ -6210,9 +6329,8 @@ llvm::Constant *CGObjCNonFragileABIMac::GetMethodConstant(
 /// }
 ///
 llvm::Constant *
-CGObjCNonFragileABIMac::EmitMethodList(Twine Name,
-                                       const char *Section,
-                                       ArrayRef<llvm::Constant*> Methods) {
+CGObjCNonFragileABIMac::EmitMethodList(Twine Name, StringRef Section,
+                                       ArrayRef<llvm::Constant *> Methods) {
   // Return null for empty list.
   if (Methods.empty())
     return llvm::Constant::getNullValue(ObjCTypes.MethodListnfABIPtrTy);
@@ -6242,18 +6360,28 @@ CGObjCNonFragileABIMac::EmitMethodList(Twine Name,
 llvm::GlobalVariable *
 CGObjCNonFragileABIMac::ObjCIvarOffsetVariable(const ObjCInterfaceDecl *ID,
                                                const ObjCIvarDecl *Ivar) {
-    
   const ObjCInterfaceDecl *Container = Ivar->getContainingInterface();
   llvm::SmallString<64> Name("OBJC_IVAR_$_");
   Name += Container->getObjCRuntimeNameAsString();
   Name += ".";
   Name += Ivar->getName();
-  llvm::GlobalVariable *IvarOffsetGV =
-    CGM.getModule().getGlobalVariable(Name);
-  if (!IvarOffsetGV)
-    IvarOffsetGV = new llvm::GlobalVariable(
-      CGM.getModule(), ObjCTypes.IvarOffsetVarTy, false,
-      llvm::GlobalValue::ExternalLinkage, nullptr, Name.str());
+  llvm::GlobalVariable *IvarOffsetGV = CGM.getModule().getGlobalVariable(Name);
+  if (!IvarOffsetGV) {
+    IvarOffsetGV =
+        new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.IvarOffsetVarTy,
+                                 false, llvm::GlobalValue::ExternalLinkage,
+                                 nullptr, Name.str());
+    if (CGM.getTriple().isOSBinFormatCOFF()) {
+      bool IsPrivateOrPackage =
+          Ivar->getAccessControl() == ObjCIvarDecl::Private ||
+          Ivar->getAccessControl() == ObjCIvarDecl::Package;
+
+      if (ID->hasAttr<DLLExportAttr>() && !IsPrivateOrPackage)
+        IvarOffsetGV->setDLLStorageClass(llvm::GlobalValue::DLLExportStorageClass);
+      else if (ID->hasAttr<DLLImportAttr>())
+        IvarOffsetGV->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
+    }
+  }
   return IvarOffsetGV;
 }
 
@@ -6267,14 +6395,17 @@ CGObjCNonFragileABIMac::EmitIvarOffsetVar(const ObjCInterfaceDecl *ID,
   IvarOffsetGV->setAlignment(
       CGM.getDataLayout().getABITypeAlignment(ObjCTypes.IvarOffsetVarTy));
 
-  // FIXME: This matches gcc, but shouldn't the visibility be set on the use as
-  // well (i.e., in ObjCIvarOffsetVariable).
-  if (Ivar->getAccessControl() == ObjCIvarDecl::Private ||
-      Ivar->getAccessControl() == ObjCIvarDecl::Package ||
-      ID->getVisibility() == HiddenVisibility)
-    IvarOffsetGV->setVisibility(llvm::GlobalValue::HiddenVisibility);
-  else
-    IvarOffsetGV->setVisibility(llvm::GlobalValue::DefaultVisibility);
+  if (!CGM.getTriple().isOSBinFormatCOFF()) {
+    // FIXME: This matches gcc, but shouldn't the visibility be set on the use
+    // as well (i.e., in ObjCIvarOffsetVariable).
+    if (Ivar->getAccessControl() == ObjCIvarDecl::Private ||
+        Ivar->getAccessControl() == ObjCIvarDecl::Package ||
+        ID->getVisibility() == HiddenVisibility)
+      IvarOffsetGV->setVisibility(llvm::GlobalValue::HiddenVisibility);
+    else
+      IvarOffsetGV->setVisibility(llvm::GlobalValue::DefaultVisibility);
+  }
+
   IvarOffsetGV->setSection("__DATA, __objc_ivar");
   return IvarOffsetGV;
 }
@@ -6361,7 +6492,7 @@ llvm::Constant *CGObjCNonFragileABIMac::GetOrEmitProtocolRef(
   const ObjCProtocolDecl *PD) {
   llvm::GlobalVariable *&Entry = Protocols[PD->getIdentifier()];
 
-  if (!Entry) {
+  if (!Entry)
     // We use the initializer as a marker of whether this is a forward
     // reference or not. At module finalization we add the empty
     // contents for protocols which were referenced but never defined.
@@ -6370,8 +6501,6 @@ llvm::Constant *CGObjCNonFragileABIMac::GetOrEmitProtocolRef(
                                  false, llvm::GlobalValue::ExternalLinkage,
                                  nullptr,
                                  "\01l_OBJC_PROTOCOL_$_" + PD->getObjCRuntimeNameAsString());
-    Entry->setSection("__DATA,__datacoal_nt,coalesced");
-  }
 
   return Entry;
 }
@@ -6391,6 +6520,7 @@ llvm::Constant *CGObjCNonFragileABIMac::GetOrEmitProtocolRef(
 ///   const uint32_t flags;  // = 0
 ///   const char ** extendedMethodTypes;
 ///   const char *demangledName;
+///   const struct _prop_list_t * class_properties;
 /// }
 /// @endcode
 ///
@@ -6442,7 +6572,7 @@ llvm::Constant *CGObjCNonFragileABIMac::GetOrEmitProtocol(
   MethodTypesExt.insert(MethodTypesExt.end(),
                         OptMethodTypesExt.begin(), OptMethodTypesExt.end());
 
-  llvm::Constant *Values[12];
+  llvm::Constant *Values[13];
   // isa is NULL
   Values[0] = llvm::Constant::getNullValue(ObjCTypes.ObjectPtrTy);
   Values[1] = GetClassName(PD->getObjCRuntimeNameAsString());
@@ -6466,8 +6596,9 @@ llvm::Constant *CGObjCNonFragileABIMac::GetOrEmitProtocol(
                              + PD->getObjCRuntimeNameAsString(),
                              "__DATA, __objc_const",
                              OptClassMethods);
-  Values[7] = EmitPropertyList("\01l_OBJC_$_PROP_LIST_" + PD->getObjCRuntimeNameAsString(),
-                               nullptr, PD, ObjCTypes);
+  Values[7] = EmitPropertyList(
+      "\01l_OBJC_$_PROP_LIST_" + PD->getObjCRuntimeNameAsString(),
+      nullptr, PD, ObjCTypes, false);
   uint32_t Size =
     CGM.getDataLayout().getTypeAllocSize(ObjCTypes.ProtocolnfABITy);
   Values[8] = llvm::ConstantInt::get(ObjCTypes.IntTy, Size);
@@ -6477,6 +6608,10 @@ llvm::Constant *CGObjCNonFragileABIMac::GetOrEmitProtocol(
                                        MethodTypesExt, ObjCTypes);
   // const char *demangledName;
   Values[11] = llvm::Constant::getNullValue(ObjCTypes.Int8PtrTy);
+
+  Values[12] = EmitPropertyList(
+      "\01l_OBJC_$_CLASS_PROP_LIST_" + PD->getObjCRuntimeNameAsString(),
+      nullptr, PD, ObjCTypes, true);
     
   llvm::Constant *Init = llvm::ConstantStruct::get(ObjCTypes.ProtocolnfABITy,
                                                    Values);
@@ -6492,7 +6627,6 @@ llvm::Constant *CGObjCNonFragileABIMac::GetOrEmitProtocol(
                                "\01l_OBJC_PROTOCOL_$_" + PD->getObjCRuntimeNameAsString());
     Entry->setAlignment(
       CGM.getDataLayout().getABITypeAlignment(ObjCTypes.ProtocolnfABITy));
-    Entry->setSection("__DATA,__datacoal_nt,coalesced");
 
     Protocols[PD->getIdentifier()] = Entry;
   }
@@ -6640,7 +6774,7 @@ static void appendSelectorForMessageRefTable(std::string &buffer,
   }
 }
 
-/// Emit a "v-table" message send.  We emit a weak hidden-visibility
+/// Emit a "vtable" message send.  We emit a weak hidden-visibility
 /// struct, initially containing the selector pointer and a pointer to
 /// a "fixup" variant of the appropriate objc_msgSend.  To call, we
 /// load and call the function pointer, passing the address of the
@@ -6734,7 +6868,7 @@ CGObjCNonFragileABIMac::EmitVTableMessageSend(CodeGenFunction &CGF,
   
   bool requiresnullCheck = false;
   if (CGM.getLangOpts().ObjCAutoRefCount && method)
-    for (const auto *ParamDecl : method->params()) {
+    for (const auto *ParamDecl : method->parameters()) {
       if (ParamDecl->hasAttr<NSConsumedAttr>()) {
         if (!nullReturn.NullBB)
           nullReturn.init(CGF, arg0);
@@ -6783,7 +6917,7 @@ CGObjCNonFragileABIMac::GenerateMessageSend(CodeGen::CodeGenFunction &CGF,
 }
 
 llvm::GlobalVariable *
-CGObjCNonFragileABIMac::GetClassGlobal(const std::string &Name, bool Weak) {
+CGObjCNonFragileABIMac::GetClassGlobal(StringRef Name, bool Weak) {
   llvm::GlobalValue::LinkageTypes L =
       Weak ? llvm::GlobalValue::ExternalWeakLinkage
            : llvm::GlobalValue::ExternalLinkage;
@@ -6806,9 +6940,8 @@ llvm::Value *CGObjCNonFragileABIMac::EmitClassRefFromId(CodeGenFunction &CGF,
   llvm::GlobalVariable *&Entry = ClassReferences[II];
   
   if (!Entry) {
-    std::string ClassName(
-      getClassSymbolPrefix() +
-      (ID ? ID->getObjCRuntimeNameAsString() : II->getName()).str());
+    StringRef Name = ID ? ID->getObjCRuntimeNameAsString() : II->getName();
+    std::string ClassName = (getClassSymbolPrefix() + Name).str();
     llvm::GlobalVariable *ClassGV = GetClassGlobal(ClassName, Weak);
     Entry = new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.ClassnfABIPtrTy,
                                      false, llvm::GlobalValue::PrivateLinkage,
@@ -6822,6 +6955,11 @@ llvm::Value *CGObjCNonFragileABIMac::EmitClassRefFromId(CodeGenFunction &CGF,
 
 llvm::Value *CGObjCNonFragileABIMac::EmitClassRef(CodeGenFunction &CGF,
                                                   const ObjCInterfaceDecl *ID) {
+  // If the class has the objc_runtime_visible attribute, we need to
+  // use the Objective-C runtime to get the class.
+  if (ID->hasAttr<ObjCRuntimeVisibleAttr>())
+    return EmitClassRefViaRuntime(CGF, ID, ObjCTypes);
+
   return EmitClassRefFromId(CGF, ID->getIdentifier(), ID->isWeakImported(), ID);
 }
 
@@ -7100,27 +7238,28 @@ CGObjCNonFragileABIMac::EmitSynchronizedStmt(CodeGen::CodeGenFunction &CGF,
 llvm::Constant *
 CGObjCNonFragileABIMac::GetEHType(QualType T) {
   // There's a particular fixed type info for 'id'.
-  if (T->isObjCIdType() ||
-      T->isObjCQualifiedIdType()) {
-    llvm::Constant *IDEHType =
-      CGM.getModule().getGlobalVariable("OBJC_EHTYPE_id");
-    if (!IDEHType)
+  if (T->isObjCIdType() || T->isObjCQualifiedIdType()) {
+    auto *IDEHType = CGM.getModule().getGlobalVariable("OBJC_EHTYPE_id");
+    if (!IDEHType) {
       IDEHType =
-        new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.EHTypeTy,
-                                 false,
-                                 llvm::GlobalValue::ExternalLinkage,
-                                 nullptr, "OBJC_EHTYPE_id");
+          new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.EHTypeTy, false,
+                                   llvm::GlobalValue::ExternalLinkage, nullptr,
+                                   "OBJC_EHTYPE_id");
+      if (CGM.getTriple().isOSBinFormatCOFF())
+        IDEHType->setDLLStorageClass(getStorage(CGM, "OBJC_EHTYPE_id"));
+    }
     return IDEHType;
   }
 
   // All other types should be Objective-C interface pointer types.
-  const ObjCObjectPointerType *PT =
-    T->getAs<ObjCObjectPointerType>();
+  const ObjCObjectPointerType *PT = T->getAs<ObjCObjectPointerType>();
   assert(PT && "Invalid @catch type.");
+
   const ObjCInterfaceType *IT = PT->getInterfaceType();
   assert(IT && "Invalid @catch type.");
+
   return GetInterfaceEHType(IT->getDecl(), false);
-}                                                  
+}
 
 void CGObjCNonFragileABIMac::EmitTryStmt(CodeGen::CodeGenFunction &CGF,
                                          const ObjCAtTryStmt &S) {
@@ -7153,6 +7292,7 @@ llvm::Constant *
 CGObjCNonFragileABIMac::GetInterfaceEHType(const ObjCInterfaceDecl *ID,
                                            bool ForDefinition) {
   llvm::GlobalVariable * &Entry = EHTypeReferences[ID->getIdentifier()];
+  StringRef ClassName = ID->getObjCRuntimeNameAsString();
 
   // If we don't need a definition, return the entry if found or check
   // if we use an external reference.
@@ -7162,38 +7302,43 @@ CGObjCNonFragileABIMac::GetInterfaceEHType(const ObjCInterfaceDecl *ID,
 
     // If this type (or a super class) has the __objc_exception__
     // attribute, emit an external reference.
-    if (hasObjCExceptionAttribute(CGM.getContext(), ID))
-      return Entry =
-          new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.EHTypeTy, false,
-                                   llvm::GlobalValue::ExternalLinkage,
-                                   nullptr,
-                                   ("OBJC_EHTYPE_$_" +
-                                    ID->getObjCRuntimeNameAsString()));
+    if (hasObjCExceptionAttribute(CGM.getContext(), ID)) {
+      std::string EHTypeName = ("OBJC_EHTYPE_$_" + ClassName).str();
+      Entry = new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.EHTypeTy,
+                                       false, llvm::GlobalValue::ExternalLinkage,
+                                       nullptr, EHTypeName);
+      if (CGM.getTriple().isOSBinFormatCOFF()) {
+        if (ID->hasAttr<DLLExportAttr>())
+          Entry->setDLLStorageClass(llvm::GlobalValue::DLLExportStorageClass);
+        else if (ID->hasAttr<DLLImportAttr>())
+          Entry->setDLLStorageClass(llvm::GlobalValue::DLLImportStorageClass);
+      }
+      return Entry;
+    }
   }
 
-  // Otherwise we need to either make a new entry or fill in the
-  // initializer.
+  // Otherwise we need to either make a new entry or fill in the initializer.
   assert((!Entry || !Entry->hasInitializer()) && "Duplicate EHType definition");
-  llvm::SmallString<64> ClassName(getClassSymbolPrefix());
-  ClassName += ID->getObjCRuntimeNameAsString();
+
   std::string VTableName = "objc_ehtype_vtable";
-  llvm::GlobalVariable *VTableGV =
-    CGM.getModule().getGlobalVariable(VTableName);
-  if (!VTableGV)
-    VTableGV = new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.Int8PtrTy,
-                                        false,
-                                        llvm::GlobalValue::ExternalLinkage,
-                                        nullptr, VTableName);
+  auto *VTableGV = CGM.getModule().getGlobalVariable(VTableName);
+  if (!VTableGV) {
+    VTableGV =
+        new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.Int8PtrTy, false,
+                                 llvm::GlobalValue::ExternalLinkage, nullptr,
+                                 VTableName);
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      VTableGV->setDLLStorageClass(getStorage(CGM, VTableName));
+  }
 
   llvm::Value *VTableIdx = llvm::ConstantInt::get(CGM.Int32Ty, 2);
-
   llvm::Constant *Values[] = {
       llvm::ConstantExpr::getGetElementPtr(VTableGV->getValueType(), VTableGV,
                                            VTableIdx),
       GetClassName(ID->getObjCRuntimeNameAsString()),
-      GetClassGlobal(ClassName.str())};
-  llvm::Constant *Init =
-    llvm::ConstantStruct::get(ObjCTypes.EHTypeTy, Values);
+      GetClassGlobal((getClassSymbolPrefix() + ClassName).str()),
+  };
+  llvm::Constant *Init = llvm::ConstantStruct::get(ObjCTypes.EHTypeTy, Values);
 
   llvm::GlobalValue::LinkageTypes L = ForDefinition
                                           ? llvm::GlobalValue::ExternalLinkage
@@ -7201,24 +7346,25 @@ CGObjCNonFragileABIMac::GetInterfaceEHType(const ObjCInterfaceDecl *ID,
   if (Entry) {
     Entry->setInitializer(Init);
   } else {
-    llvm::SmallString<64> EHTYPEName("OBJC_EHTYPE_$_");
-    EHTYPEName += ID->getObjCRuntimeNameAsString();
-    Entry = new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.EHTypeTy, false,
-                                     L,
-                                     Init,
-                                     EHTYPEName.str());
+    Entry =
+        new llvm::GlobalVariable(CGM.getModule(), ObjCTypes.EHTypeTy, false, L,
+                                 Init, ("OBJC_EHTYPE_$_" + ClassName).str());
+    if (CGM.getTriple().isOSBinFormatCOFF())
+      if (hasObjCExceptionAttribute(CGM.getContext(), ID))
+        if (ID->hasAttr<DLLExportAttr>())
+          Entry->setDLLStorageClass(llvm::GlobalValue::DLLExportStorageClass);
   }
   assert(Entry->getLinkage() == L);
 
-  if (ID->getVisibility() == HiddenVisibility)
-    Entry->setVisibility(llvm::GlobalValue::HiddenVisibility);
-  Entry->setAlignment(CGM.getDataLayout().getABITypeAlignment(
-      ObjCTypes.EHTypeTy));
+  if (!CGM.getTriple().isOSBinFormatCOFF())
+    if (ID->getVisibility() == HiddenVisibility)
+      Entry->setVisibility(llvm::GlobalValue::HiddenVisibility);
+
+  const auto &DL = CGM.getDataLayout();
+  Entry->setAlignment(DL.getABITypeAlignment(ObjCTypes.EHTypeTy));
 
   if (ForDefinition)
     Entry->setSection("__DATA,__objc_const");
-  else
-    Entry->setSection("__DATA,__datacoal_nt,coalesced");
 
   return Entry;
 }
diff --git a/lib/CodeGen/CGObjCRuntime.cpp b/lib/CodeGen/CGObjCRuntime.cpp
index 7be9ae996040..0caf6d9f210a 100644
--- a/lib/CodeGen/CGObjCRuntime.cpp
+++ b/lib/CodeGen/CGObjCRuntime.cpp
@@ -120,9 +120,8 @@ LValue CGObjCRuntime::EmitValueForIvarAtOffset(CodeGen::CodeGenFunction &CGF,
   uint64_t BitOffset = FieldBitOffset % CGF.CGM.getContext().getCharWidth();
   uint64_t AlignmentBits = CGF.CGM.getTarget().getCharAlign();
   uint64_t BitFieldSize = Ivar->getBitWidthValue(CGF.getContext());
-  CharUnits StorageSize =
-    CGF.CGM.getContext().toCharUnitsFromBits(
-      llvm::RoundUpToAlignment(BitOffset + BitFieldSize, AlignmentBits));
+  CharUnits StorageSize = CGF.CGM.getContext().toCharUnitsFromBits(
+      llvm::alignTo(BitOffset + BitFieldSize, AlignmentBits));
   CharUnits Alignment = CGF.CGM.getContext().toCharUnitsFromBits(AlignmentBits);
 
   // Allocate a new CGBitFieldInfo object to describe this access.
@@ -364,25 +363,15 @@ CGObjCRuntime::getMessageSendInfo(const ObjCMethodDecl *method,
     llvm::PointerType *signatureType =
       CGM.getTypes().GetFunctionType(signature)->getPointerTo();
 
-    // If that's not variadic, there's no need to recompute the ABI
-    // arrangement.
-    if (!signature.isVariadic())
-      return MessageSendInfo(signature, signatureType);
+    const CGFunctionInfo &signatureForCall =
+      CGM.getTypes().arrangeCall(signature, callArgs);
 
-    // Otherwise, there is.
-    FunctionType::ExtInfo einfo = signature.getExtInfo();
-    const CGFunctionInfo &argsInfo =
-      CGM.getTypes().arrangeFreeFunctionCall(resultType, callArgs, einfo,
-                                             signature.getRequiredArgs());
-
-    return MessageSendInfo(argsInfo, signatureType);
+    return MessageSendInfo(signatureForCall, signatureType);
   }
 
   // There's no method;  just use a default CC.
   const CGFunctionInfo &argsInfo =
-    CGM.getTypes().arrangeFreeFunctionCall(resultType, callArgs, 
-                                           FunctionType::ExtInfo(),
-                                           RequiredArgs::All);
+    CGM.getTypes().arrangeUnprototypedObjCMessageSend(resultType, callArgs);
 
   // Derive the signature to call from that.
   llvm::PointerType *signatureType =
diff --git a/lib/CodeGen/CGObjCRuntime.h b/lib/CodeGen/CGObjCRuntime.h
index 28d88dd10be9..6c330590f7cd 100644
--- a/lib/CodeGen/CGObjCRuntime.h
+++ b/lib/CodeGen/CGObjCRuntime.h
@@ -280,7 +280,7 @@ public:
   virtual llvm::Constant *BuildByrefLayout(CodeGen::CodeGenModule &CGM,
                                            QualType T) = 0;
 
-  virtual llvm::GlobalVariable *GetClassGlobal(const std::string &Name,
+  virtual llvm::GlobalVariable *GetClassGlobal(StringRef Name,
                                                bool Weak = false) = 0;
 
   struct MessageSendInfo {
diff --git a/lib/CodeGen/CGOpenCLRuntime.cpp b/lib/CodeGen/CGOpenCLRuntime.cpp
index 686678962d3e..38aebea18ed3 100644
--- a/lib/CodeGen/CGOpenCLRuntime.cpp
+++ b/lib/CodeGen/CGOpenCLRuntime.cpp
@@ -40,46 +40,12 @@ llvm::Type *CGOpenCLRuntime::convertOpenCLSpecificType(const Type *T) {
   default: 
     llvm_unreachable("Unexpected opencl builtin type!");
     return nullptr;
-  case BuiltinType::OCLImage1d:
-    return llvm::PointerType::get(llvm::StructType::create(
-                           Ctx, "opencl.image1d_t"), ImgAddrSpc);
-  case BuiltinType::OCLImage1dArray:
-    return llvm::PointerType::get(llvm::StructType::create(
-                           Ctx, "opencl.image1d_array_t"), ImgAddrSpc);
-  case BuiltinType::OCLImage1dBuffer:
-    return llvm::PointerType::get(llvm::StructType::create(
-                           Ctx, "opencl.image1d_buffer_t"), ImgAddrSpc);
-  case BuiltinType::OCLImage2d:
-    return llvm::PointerType::get(llvm::StructType::create(
-                           Ctx, "opencl.image2d_t"), ImgAddrSpc);
-  case BuiltinType::OCLImage2dArray:
-    return llvm::PointerType::get(llvm::StructType::create(
-                           Ctx, "opencl.image2d_array_t"), ImgAddrSpc);
-  case BuiltinType::OCLImage2dDepth:
-    return llvm::PointerType::get(
-        llvm::StructType::create(Ctx, "opencl.image2d_depth_t"), ImgAddrSpc);
-  case BuiltinType::OCLImage2dArrayDepth:
-    return llvm::PointerType::get(
-        llvm::StructType::create(Ctx, "opencl.image2d_array_depth_t"),
+#define IMAGE_TYPE(ImgType, Id, SingletonId, Access, Suffix) \
+  case BuiltinType::Id: \
+    return llvm::PointerType::get( \
+        llvm::StructType::create(Ctx, "opencl." #ImgType "_" #Suffix "_t"), \
         ImgAddrSpc);
-  case BuiltinType::OCLImage2dMSAA:
-    return llvm::PointerType::get(
-        llvm::StructType::create(Ctx, "opencl.image2d_msaa_t"), ImgAddrSpc);
-  case BuiltinType::OCLImage2dArrayMSAA:
-    return llvm::PointerType::get(
-        llvm::StructType::create(Ctx, "opencl.image2d_array_msaa_t"),
-        ImgAddrSpc);
-  case BuiltinType::OCLImage2dMSAADepth:
-    return llvm::PointerType::get(
-        llvm::StructType::create(Ctx, "opencl.image2d_msaa_depth_t"),
-        ImgAddrSpc);
-  case BuiltinType::OCLImage2dArrayMSAADepth:
-    return llvm::PointerType::get(
-        llvm::StructType::create(Ctx, "opencl.image2d_array_msaa_depth_t"),
-        ImgAddrSpc);
-  case BuiltinType::OCLImage3d:
-    return llvm::PointerType::get(llvm::StructType::create(
-                           Ctx, "opencl.image3d_t"), ImgAddrSpc);
+#include "clang/Basic/OpenCLImageTypes.def"
   case BuiltinType::OCLSampler:
     return llvm::IntegerType::get(Ctx, 32);
   case BuiltinType::OCLEvent:
diff --git a/lib/CodeGen/CGOpenMPRuntime.cpp b/lib/CodeGen/CGOpenMPRuntime.cpp
index 5cfacacbe01a..6a0edbe0e7a9 100644
--- a/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -72,6 +72,8 @@ public:
   /// \return LValue for thread id variable. This LValue always has type int32*.
   virtual LValue getThreadIDVariableLValue(CodeGenFunction &CGF);
 
+  virtual void emitUntiedSwitch(CodeGenFunction & /*CGF*/) {}
+
   CGOpenMPRegionKind getRegionKind() const { return RegionKind; }
 
   OpenMPDirectiveKind getDirectiveKind() const { return Kind; }
@@ -82,6 +84,8 @@ public:
     return Info->getKind() == CR_OpenMP;
   }
 
+  ~CGOpenMPRegionInfo() override = default;
+
 protected:
   CGOpenMPRegionKind RegionKind;
   RegionCodeGenTy CodeGen;
@@ -90,7 +94,7 @@ protected:
 };
 
 /// \brief API for captured statement code generation in OpenMP constructs.
-class CGOpenMPOutlinedRegionInfo : public CGOpenMPRegionInfo {
+class CGOpenMPOutlinedRegionInfo final : public CGOpenMPRegionInfo {
 public:
   CGOpenMPOutlinedRegionInfo(const CapturedStmt &CS, const VarDecl *ThreadIDVar,
                              const RegionCodeGenTy &CodeGen,
@@ -100,6 +104,7 @@ public:
         ThreadIDVar(ThreadIDVar) {
     assert(ThreadIDVar != nullptr && "No ThreadID in OpenMP region.");
   }
+
   /// \brief Get a variable or parameter for storing global thread id
   /// inside OpenMP construct.
   const VarDecl *getThreadIDVariable() const override { return ThreadIDVar; }
@@ -120,16 +125,65 @@ private:
 };
 
 /// \brief API for captured statement code generation in OpenMP constructs.
-class CGOpenMPTaskOutlinedRegionInfo : public CGOpenMPRegionInfo {
+class CGOpenMPTaskOutlinedRegionInfo final : public CGOpenMPRegionInfo {
 public:
+  class UntiedTaskActionTy final : public PrePostActionTy {
+    bool Untied;
+    const VarDecl *PartIDVar;
+    const RegionCodeGenTy UntiedCodeGen;
+    llvm::SwitchInst *UntiedSwitch = nullptr;
+
+  public:
+    UntiedTaskActionTy(bool Tied, const VarDecl *PartIDVar,
+                       const RegionCodeGenTy &UntiedCodeGen)
+        : Untied(!Tied), PartIDVar(PartIDVar), UntiedCodeGen(UntiedCodeGen) {}
+    void Enter(CodeGenFunction &CGF) override {
+      if (Untied) {
+        // Emit task switching point.
+        auto PartIdLVal = CGF.EmitLoadOfPointerLValue(
+            CGF.GetAddrOfLocalVar(PartIDVar),
+            PartIDVar->getType()->castAs<PointerType>());
+        auto *Res = CGF.EmitLoadOfScalar(PartIdLVal, SourceLocation());
+        auto *DoneBB = CGF.createBasicBlock(".untied.done.");
+        UntiedSwitch = CGF.Builder.CreateSwitch(Res, DoneBB);
+        CGF.EmitBlock(DoneBB);
+        CGF.EmitBranchThroughCleanup(CGF.ReturnBlock);
+        CGF.EmitBlock(CGF.createBasicBlock(".untied.jmp."));
+        UntiedSwitch->addCase(CGF.Builder.getInt32(0),
+                              CGF.Builder.GetInsertBlock());
+        emitUntiedSwitch(CGF);
+      }
+    }
+    void emitUntiedSwitch(CodeGenFunction &CGF) const {
+      if (Untied) {
+        auto PartIdLVal = CGF.EmitLoadOfPointerLValue(
+            CGF.GetAddrOfLocalVar(PartIDVar),
+            PartIDVar->getType()->castAs<PointerType>());
+        CGF.EmitStoreOfScalar(CGF.Builder.getInt32(UntiedSwitch->getNumCases()),
+                              PartIdLVal);
+        UntiedCodeGen(CGF);
+        CodeGenFunction::JumpDest CurPoint =
+            CGF.getJumpDestInCurrentScope(".untied.next.");
+        CGF.EmitBranchThroughCleanup(CGF.ReturnBlock);
+        CGF.EmitBlock(CGF.createBasicBlock(".untied.jmp."));
+        UntiedSwitch->addCase(CGF.Builder.getInt32(UntiedSwitch->getNumCases()),
+                              CGF.Builder.GetInsertBlock());
+        CGF.EmitBranchThroughCleanup(CurPoint);
+        CGF.EmitBlock(CurPoint.getBlock());
+      }
+    }
+    unsigned getNumberOfParts() const { return UntiedSwitch->getNumCases(); }
+  };
   CGOpenMPTaskOutlinedRegionInfo(const CapturedStmt &CS,
                                  const VarDecl *ThreadIDVar,
                                  const RegionCodeGenTy &CodeGen,
-                                 OpenMPDirectiveKind Kind, bool HasCancel)
+                                 OpenMPDirectiveKind Kind, bool HasCancel,
+                                 const UntiedTaskActionTy &Action)
       : CGOpenMPRegionInfo(CS, TaskOutlinedRegion, CodeGen, Kind, HasCancel),
-        ThreadIDVar(ThreadIDVar) {
+        ThreadIDVar(ThreadIDVar), Action(Action) {
     assert(ThreadIDVar != nullptr && "No ThreadID in OpenMP region.");
   }
+
   /// \brief Get a variable or parameter for storing global thread id
   /// inside OpenMP construct.
   const VarDecl *getThreadIDVariable() const override { return ThreadIDVar; }
@@ -140,6 +194,10 @@ public:
   /// \brief Get the name of the capture helper.
   StringRef getHelperName() const override { return ".omp_outlined."; }
 
+  void emitUntiedSwitch(CodeGenFunction &CGF) override {
+    Action.emitUntiedSwitch(CGF);
+  }
+
   static bool classof(const CGCapturedStmtInfo *Info) {
     return CGOpenMPRegionInfo::classof(Info) &&
            cast<CGOpenMPRegionInfo>(Info)->getRegionKind() ==
@@ -150,6 +208,8 @@ private:
   /// \brief A variable or parameter storing global thread id for OpenMP
   /// constructs.
   const VarDecl *ThreadIDVar;
+  /// Action for emitting code for untied tasks.
+  const UntiedTaskActionTy &Action;
 };
 
 /// \brief API for inlined captured statement code generation in OpenMP
@@ -162,12 +222,14 @@ public:
       : CGOpenMPRegionInfo(InlinedRegion, CodeGen, Kind, HasCancel),
         OldCSI(OldCSI),
         OuterRegionInfo(dyn_cast_or_null<CGOpenMPRegionInfo>(OldCSI)) {}
+
   // \brief Retrieve the value of the context parameter.
   llvm::Value *getContextValue() const override {
     if (OuterRegionInfo)
       return OuterRegionInfo->getContextValue();
     llvm_unreachable("No context value for inlined OpenMP region");
   }
+
   void setContextValue(llvm::Value *V) override {
     if (OuterRegionInfo) {
       OuterRegionInfo->setContextValue(V);
@@ -175,6 +237,7 @@ public:
     }
     llvm_unreachable("No context value for inlined OpenMP region");
   }
+
   /// \brief Lookup the captured field decl for a variable.
   const FieldDecl *lookup(const VarDecl *VD) const override {
     if (OuterRegionInfo)
@@ -183,11 +246,13 @@ public:
     // captured variables, we can use the original one.
     return nullptr;
   }
+
   FieldDecl *getThisFieldDecl() const override {
     if (OuterRegionInfo)
       return OuterRegionInfo->getThisFieldDecl();
     return nullptr;
   }
+
   /// \brief Get a variable or parameter for storing global thread id
   /// inside OpenMP construct.
   const VarDecl *getThreadIDVariable() const override {
@@ -203,6 +268,11 @@ public:
     llvm_unreachable("No helper name for inlined OpenMP construct");
   }
 
+  void emitUntiedSwitch(CodeGenFunction &CGF) override {
+    if (OuterRegionInfo)
+      OuterRegionInfo->emitUntiedSwitch(CGF);
+  }
+
   CodeGenFunction::CGCapturedStmtInfo *getOldCSI() const { return OldCSI; }
 
   static bool classof(const CGCapturedStmtInfo *Info) {
@@ -210,6 +280,8 @@ public:
            cast<CGOpenMPRegionInfo>(Info)->getRegionKind() == InlinedRegion;
   }
 
+  ~CGOpenMPInlinedRegionInfo() override = default;
+
 private:
   /// \brief CodeGen info about outer OpenMP region.
   CodeGenFunction::CGCapturedStmtInfo *OldCSI;
@@ -221,7 +293,7 @@ private:
 /// captured fields. The name of the target region has to be unique in a given
 /// application so it is provided by the client, because only the client has
 /// the information to generate that.
-class CGOpenMPTargetRegionInfo : public CGOpenMPRegionInfo {
+class CGOpenMPTargetRegionInfo final : public CGOpenMPRegionInfo {
 public:
   CGOpenMPTargetRegionInfo(const CapturedStmt &CS,
                            const RegionCodeGenTy &CodeGen, StringRef HelperName)
@@ -245,9 +317,75 @@ private:
   StringRef HelperName;
 };
 
+static void EmptyCodeGen(CodeGenFunction &, PrePostActionTy &) {
+  llvm_unreachable("No codegen for expressions");
+}
+/// \brief API for generation of expressions captured in a innermost OpenMP
+/// region.
+class CGOpenMPInnerExprInfo final : public CGOpenMPInlinedRegionInfo {
+public:
+  CGOpenMPInnerExprInfo(CodeGenFunction &CGF, const CapturedStmt &CS)
+      : CGOpenMPInlinedRegionInfo(CGF.CapturedStmtInfo, EmptyCodeGen,
+                                  OMPD_unknown,
+                                  /*HasCancel=*/false),
+        PrivScope(CGF) {
+    // Make sure the globals captured in the provided statement are local by
+    // using the privatization logic. We assume the same variable is not
+    // captured more than once.
+    for (auto &C : CS.captures()) {
+      if (!C.capturesVariable() && !C.capturesVariableByCopy())
+        continue;
+
+      const VarDecl *VD = C.getCapturedVar();
+      if (VD->isLocalVarDeclOrParm())
+        continue;
+
+      DeclRefExpr DRE(const_cast<VarDecl *>(VD),
+                      /*RefersToEnclosingVariableOrCapture=*/false,
+                      VD->getType().getNonReferenceType(), VK_LValue,
+                      SourceLocation());
+      PrivScope.addPrivate(VD, [&CGF, &DRE]() -> Address {
+        return CGF.EmitLValue(&DRE).getAddress();
+      });
+    }
+    (void)PrivScope.Privatize();
+  }
+
+  /// \brief Lookup the captured field decl for a variable.
+  const FieldDecl *lookup(const VarDecl *VD) const override {
+    if (auto *FD = CGOpenMPInlinedRegionInfo::lookup(VD))
+      return FD;
+    return nullptr;
+  }
+
+  /// \brief Emit the captured statement body.
+  void EmitBody(CodeGenFunction &CGF, const Stmt *S) override {
+    llvm_unreachable("No body for expressions");
+  }
+
+  /// \brief Get a variable or parameter for storing global thread id
+  /// inside OpenMP construct.
+  const VarDecl *getThreadIDVariable() const override {
+    llvm_unreachable("No thread id for expressions");
+  }
+
+  /// \brief Get the name of the capture helper.
+  StringRef getHelperName() const override {
+    llvm_unreachable("No helper name for expressions");
+  }
+
+  static bool classof(const CGCapturedStmtInfo *Info) { return false; }
+
+private:
+  /// Private scope to capture global variables.
+  CodeGenFunction::OMPPrivateScope PrivScope;
+};
+
 /// \brief RAII for emitting code of OpenMP constructs.
 class InlinedOpenMPRegionRAII {
   CodeGenFunction &CGF;
+  llvm::DenseMap<const VarDecl *, FieldDecl *> LambdaCaptureFields;
+  FieldDecl *LambdaThisCaptureField = nullptr;
 
 public:
   /// \brief Constructs region for combined constructs.
@@ -260,30 +398,306 @@ public:
     // Start emission for the construct.
     CGF.CapturedStmtInfo = new CGOpenMPInlinedRegionInfo(
         CGF.CapturedStmtInfo, CodeGen, Kind, HasCancel);
+    std::swap(CGF.LambdaCaptureFields, LambdaCaptureFields);
+    LambdaThisCaptureField = CGF.LambdaThisCaptureField;
+    CGF.LambdaThisCaptureField = nullptr;
   }
+
   ~InlinedOpenMPRegionRAII() {
     // Restore original CapturedStmtInfo only if we're done with code emission.
     auto *OldCSI =
         cast<CGOpenMPInlinedRegionInfo>(CGF.CapturedStmtInfo)->getOldCSI();
     delete CGF.CapturedStmtInfo;
     CGF.CapturedStmtInfo = OldCSI;
+    std::swap(CGF.LambdaCaptureFields, LambdaCaptureFields);
+    CGF.LambdaThisCaptureField = LambdaThisCaptureField;
+  }
+};
+
+/// \brief Values for bit flags used in the ident_t to describe the fields.
+/// All enumeric elements are named and described in accordance with the code
+/// from http://llvm.org/svn/llvm-project/openmp/trunk/runtime/src/kmp.h
+enum OpenMPLocationFlags {
+  /// \brief Use trampoline for internal microtask.
+  OMP_IDENT_IMD = 0x01,
+  /// \brief Use c-style ident structure.
+  OMP_IDENT_KMPC = 0x02,
+  /// \brief Atomic reduction option for kmpc_reduce.
+  OMP_ATOMIC_REDUCE = 0x10,
+  /// \brief Explicit 'barrier' directive.
+  OMP_IDENT_BARRIER_EXPL = 0x20,
+  /// \brief Implicit barrier in code.
+  OMP_IDENT_BARRIER_IMPL = 0x40,
+  /// \brief Implicit barrier in 'for' directive.
+  OMP_IDENT_BARRIER_IMPL_FOR = 0x40,
+  /// \brief Implicit barrier in 'sections' directive.
+  OMP_IDENT_BARRIER_IMPL_SECTIONS = 0xC0,
+  /// \brief Implicit barrier in 'single' directive.
+  OMP_IDENT_BARRIER_IMPL_SINGLE = 0x140
+};
+
+/// \brief Describes ident structure that describes a source location.
+/// All descriptions are taken from
+/// http://llvm.org/svn/llvm-project/openmp/trunk/runtime/src/kmp.h
+/// Original structure:
+/// typedef struct ident {
+///    kmp_int32 reserved_1;   /**<  might be used in Fortran;
+///                                  see above  */
+///    kmp_int32 flags;        /**<  also f.flags; KMP_IDENT_xxx flags;
+///                                  KMP_IDENT_KMPC identifies this union
+///                                  member  */
+///    kmp_int32 reserved_2;   /**<  not really used in Fortran any more;
+///                                  see above */
+///#if USE_ITT_BUILD
+///                            /*  but currently used for storing
+///                                region-specific ITT */
+///                            /*  contextual information. */
+///#endif /* USE_ITT_BUILD */
+///    kmp_int32 reserved_3;   /**< source[4] in Fortran, do not use for
+///                                 C++  */
+///    char const *psource;    /**< String describing the source location.
+///                            The string is composed of semi-colon separated
+//                             fields which describe the source file,
+///                            the function and a pair of line numbers that
+///                            delimit the construct.
+///                             */
+/// } ident_t;
+enum IdentFieldIndex {
+  /// \brief might be used in Fortran
+  IdentField_Reserved_1,
+  /// \brief OMP_IDENT_xxx flags; OMP_IDENT_KMPC identifies this union member.
+  IdentField_Flags,
+  /// \brief Not really used in Fortran any more
+  IdentField_Reserved_2,
+  /// \brief Source[4] in Fortran, do not use for C++
+  IdentField_Reserved_3,
+  /// \brief String describing the source location. The string is composed of
+  /// semi-colon separated fields which describe the source file, the function
+  /// and a pair of line numbers that delimit the construct.
+  IdentField_PSource
+};
+
+/// \brief Schedule types for 'omp for' loops (these enumerators are taken from
+/// the enum sched_type in kmp.h).
+enum OpenMPSchedType {
+  /// \brief Lower bound for default (unordered) versions.
+  OMP_sch_lower = 32,
+  OMP_sch_static_chunked = 33,
+  OMP_sch_static = 34,
+  OMP_sch_dynamic_chunked = 35,
+  OMP_sch_guided_chunked = 36,
+  OMP_sch_runtime = 37,
+  OMP_sch_auto = 38,
+  /// static with chunk adjustment (e.g., simd)
+  OMP_sch_static_balanced_chunked   = 45,
+  /// \brief Lower bound for 'ordered' versions.
+  OMP_ord_lower = 64,
+  OMP_ord_static_chunked = 65,
+  OMP_ord_static = 66,
+  OMP_ord_dynamic_chunked = 67,
+  OMP_ord_guided_chunked = 68,
+  OMP_ord_runtime = 69,
+  OMP_ord_auto = 70,
+  OMP_sch_default = OMP_sch_static,
+  /// \brief dist_schedule types
+  OMP_dist_sch_static_chunked = 91,
+  OMP_dist_sch_static = 92,
+  /// Support for OpenMP 4.5 monotonic and nonmonotonic schedule modifiers.
+  /// Set if the monotonic schedule modifier was present.
+  OMP_sch_modifier_monotonic = (1 << 29),
+  /// Set if the nonmonotonic schedule modifier was present.
+  OMP_sch_modifier_nonmonotonic = (1 << 30),
+};
+
+enum OpenMPRTLFunction {
+  /// \brief Call to void __kmpc_fork_call(ident_t *loc, kmp_int32 argc,
+  /// kmpc_micro microtask, ...);
+  OMPRTL__kmpc_fork_call,
+  /// \brief Call to void *__kmpc_threadprivate_cached(ident_t *loc,
+  /// kmp_int32 global_tid, void *data, size_t size, void ***cache);
+  OMPRTL__kmpc_threadprivate_cached,
+  /// \brief Call to void __kmpc_threadprivate_register( ident_t *,
+  /// void *data, kmpc_ctor ctor, kmpc_cctor cctor, kmpc_dtor dtor);
+  OMPRTL__kmpc_threadprivate_register,
+  // Call to __kmpc_int32 kmpc_global_thread_num(ident_t *loc);
+  OMPRTL__kmpc_global_thread_num,
+  // Call to void __kmpc_critical(ident_t *loc, kmp_int32 global_tid,
+  // kmp_critical_name *crit);
+  OMPRTL__kmpc_critical,
+  // Call to void __kmpc_critical_with_hint(ident_t *loc, kmp_int32
+  // global_tid, kmp_critical_name *crit, uintptr_t hint);
+  OMPRTL__kmpc_critical_with_hint,
+  // Call to void __kmpc_end_critical(ident_t *loc, kmp_int32 global_tid,
+  // kmp_critical_name *crit);
+  OMPRTL__kmpc_end_critical,
+  // Call to kmp_int32 __kmpc_cancel_barrier(ident_t *loc, kmp_int32
+  // global_tid);
+  OMPRTL__kmpc_cancel_barrier,
+  // Call to void __kmpc_barrier(ident_t *loc, kmp_int32 global_tid);
+  OMPRTL__kmpc_barrier,
+  // Call to void __kmpc_for_static_fini(ident_t *loc, kmp_int32 global_tid);
+  OMPRTL__kmpc_for_static_fini,
+  // Call to void __kmpc_serialized_parallel(ident_t *loc, kmp_int32
+  // global_tid);
+  OMPRTL__kmpc_serialized_parallel,
+  // Call to void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32
+  // global_tid);
+  OMPRTL__kmpc_end_serialized_parallel,
+  // Call to void __kmpc_push_num_threads(ident_t *loc, kmp_int32 global_tid,
+  // kmp_int32 num_threads);
+  OMPRTL__kmpc_push_num_threads,
+  // Call to void __kmpc_flush(ident_t *loc);
+  OMPRTL__kmpc_flush,
+  // Call to kmp_int32 __kmpc_master(ident_t *, kmp_int32 global_tid);
+  OMPRTL__kmpc_master,
+  // Call to void __kmpc_end_master(ident_t *, kmp_int32 global_tid);
+  OMPRTL__kmpc_end_master,
+  // Call to kmp_int32 __kmpc_omp_taskyield(ident_t *, kmp_int32 global_tid,
+  // int end_part);
+  OMPRTL__kmpc_omp_taskyield,
+  // Call to kmp_int32 __kmpc_single(ident_t *, kmp_int32 global_tid);
+  OMPRTL__kmpc_single,
+  // Call to void __kmpc_end_single(ident_t *, kmp_int32 global_tid);
+  OMPRTL__kmpc_end_single,
+  // Call to kmp_task_t * __kmpc_omp_task_alloc(ident_t *, kmp_int32 gtid,
+  // kmp_int32 flags, size_t sizeof_kmp_task_t, size_t sizeof_shareds,
+  // kmp_routine_entry_t *task_entry);
+  OMPRTL__kmpc_omp_task_alloc,
+  // Call to kmp_int32 __kmpc_omp_task(ident_t *, kmp_int32 gtid, kmp_task_t *
+  // new_task);
+  OMPRTL__kmpc_omp_task,
+  // Call to void __kmpc_copyprivate(ident_t *loc, kmp_int32 global_tid,
+  // size_t cpy_size, void *cpy_data, void(*cpy_func)(void *, void *),
+  // kmp_int32 didit);
+  OMPRTL__kmpc_copyprivate,
+  // Call to kmp_int32 __kmpc_reduce(ident_t *loc, kmp_int32 global_tid,
+  // kmp_int32 num_vars, size_t reduce_size, void *reduce_data, void
+  // (*reduce_func)(void *lhs_data, void *rhs_data), kmp_critical_name *lck);
+  OMPRTL__kmpc_reduce,
+  // Call to kmp_int32 __kmpc_reduce_nowait(ident_t *loc, kmp_int32
+  // global_tid, kmp_int32 num_vars, size_t reduce_size, void *reduce_data,
+  // void (*reduce_func)(void *lhs_data, void *rhs_data), kmp_critical_name
+  // *lck);
+  OMPRTL__kmpc_reduce_nowait,
+  // Call to void __kmpc_end_reduce(ident_t *loc, kmp_int32 global_tid,
+  // kmp_critical_name *lck);
+  OMPRTL__kmpc_end_reduce,
+  // Call to void __kmpc_end_reduce_nowait(ident_t *loc, kmp_int32 global_tid,
+  // kmp_critical_name *lck);
+  OMPRTL__kmpc_end_reduce_nowait,
+  // Call to void __kmpc_omp_task_begin_if0(ident_t *, kmp_int32 gtid,
+  // kmp_task_t * new_task);
+  OMPRTL__kmpc_omp_task_begin_if0,
+  // Call to void __kmpc_omp_task_complete_if0(ident_t *, kmp_int32 gtid,
+  // kmp_task_t * new_task);
+  OMPRTL__kmpc_omp_task_complete_if0,
+  // Call to void __kmpc_ordered(ident_t *loc, kmp_int32 global_tid);
+  OMPRTL__kmpc_ordered,
+  // Call to void __kmpc_end_ordered(ident_t *loc, kmp_int32 global_tid);
+  OMPRTL__kmpc_end_ordered,
+  // Call to kmp_int32 __kmpc_omp_taskwait(ident_t *loc, kmp_int32
+  // global_tid);
+  OMPRTL__kmpc_omp_taskwait,
+  // Call to void __kmpc_taskgroup(ident_t *loc, kmp_int32 global_tid);
+  OMPRTL__kmpc_taskgroup,
+  // Call to void __kmpc_end_taskgroup(ident_t *loc, kmp_int32 global_tid);
+  OMPRTL__kmpc_end_taskgroup,
+  // Call to void __kmpc_push_proc_bind(ident_t *loc, kmp_int32 global_tid,
+  // int proc_bind);
+  OMPRTL__kmpc_push_proc_bind,
+  // Call to kmp_int32 __kmpc_omp_task_with_deps(ident_t *loc_ref, kmp_int32
+  // gtid, kmp_task_t * new_task, kmp_int32 ndeps, kmp_depend_info_t
+  // *dep_list, kmp_int32 ndeps_noalias, kmp_depend_info_t *noalias_dep_list);
+  OMPRTL__kmpc_omp_task_with_deps,
+  // Call to void __kmpc_omp_wait_deps(ident_t *loc_ref, kmp_int32
+  // gtid, kmp_int32 ndeps, kmp_depend_info_t *dep_list, kmp_int32
+  // ndeps_noalias, kmp_depend_info_t *noalias_dep_list);
+  OMPRTL__kmpc_omp_wait_deps,
+  // Call to kmp_int32 __kmpc_cancellationpoint(ident_t *loc, kmp_int32
+  // global_tid, kmp_int32 cncl_kind);
+  OMPRTL__kmpc_cancellationpoint,
+  // Call to kmp_int32 __kmpc_cancel(ident_t *loc, kmp_int32 global_tid,
+  // kmp_int32 cncl_kind);
+  OMPRTL__kmpc_cancel,
+  // Call to void __kmpc_push_num_teams(ident_t *loc, kmp_int32 global_tid,
+  // kmp_int32 num_teams, kmp_int32 thread_limit);
+  OMPRTL__kmpc_push_num_teams,
+  // Call to void __kmpc_fork_teams(ident_t *loc, kmp_int32 argc, kmpc_micro
+  // microtask, ...);
+  OMPRTL__kmpc_fork_teams,
+  // Call to void __kmpc_taskloop(ident_t *loc, int gtid, kmp_task_t *task, int
+  // if_val, kmp_uint64 *lb, kmp_uint64 *ub, kmp_int64 st, int nogroup, int
+  // sched, kmp_uint64 grainsize, void *task_dup);
+  OMPRTL__kmpc_taskloop,
+  // Call to void __kmpc_doacross_init(ident_t *loc, kmp_int32 gtid, kmp_int32
+  // num_dims, struct kmp_dim *dims);
+  OMPRTL__kmpc_doacross_init,
+  // Call to void __kmpc_doacross_fini(ident_t *loc, kmp_int32 gtid);
+  OMPRTL__kmpc_doacross_fini,
+  // Call to void __kmpc_doacross_post(ident_t *loc, kmp_int32 gtid, kmp_int64
+  // *vec);
+  OMPRTL__kmpc_doacross_post,
+  // Call to void __kmpc_doacross_wait(ident_t *loc, kmp_int32 gtid, kmp_int64
+  // *vec);
+  OMPRTL__kmpc_doacross_wait,
+
+  //
+  // Offloading related calls
+  //
+  // Call to int32_t __tgt_target(int32_t device_id, void *host_ptr, int32_t
+  // arg_num, void** args_base, void **args, size_t *arg_sizes, int32_t
+  // *arg_types);
+  OMPRTL__tgt_target,
+  // Call to int32_t __tgt_target_teams(int32_t device_id, void *host_ptr,
+  // int32_t arg_num, void** args_base, void **args, size_t *arg_sizes,
+  // int32_t *arg_types, int32_t num_teams, int32_t thread_limit);
+  OMPRTL__tgt_target_teams,
+  // Call to void __tgt_register_lib(__tgt_bin_desc *desc);
+  OMPRTL__tgt_register_lib,
+  // Call to void __tgt_unregister_lib(__tgt_bin_desc *desc);
+  OMPRTL__tgt_unregister_lib,
+  // Call to void __tgt_target_data_begin(int32_t device_id, int32_t arg_num,
+  // void** args_base, void **args, size_t *arg_sizes, int32_t *arg_types);
+  OMPRTL__tgt_target_data_begin,
+  // Call to void __tgt_target_data_end(int32_t device_id, int32_t arg_num,
+  // void** args_base, void **args, size_t *arg_sizes, int32_t *arg_types);
+  OMPRTL__tgt_target_data_end,
+  // Call to void __tgt_target_data_update(int32_t device_id, int32_t arg_num,
+  // void** args_base, void **args, size_t *arg_sizes, int32_t *arg_types);
+  OMPRTL__tgt_target_data_update,
+};
+
+/// A basic class for pre|post-action for advanced codegen sequence for OpenMP
+/// region.
+class CleanupTy final : public EHScopeStack::Cleanup {
+  PrePostActionTy *Action;
+
+public:
+  explicit CleanupTy(PrePostActionTy *Action) : Action(Action) {}
+  void Emit(CodeGenFunction &CGF, Flags /*flags*/) override {
+    if (!CGF.HaveInsertPoint())
+      return;
+    Action->Exit(CGF);
   }
 };
 
 } // anonymous namespace
 
-static LValue emitLoadOfPointerLValue(CodeGenFunction &CGF, Address PtrAddr,
-                                      QualType Ty) {
-  AlignmentSource Source;
-  CharUnits Align = CGF.getNaturalPointeeTypeAlignment(Ty, &Source);
-  return CGF.MakeAddrLValue(Address(CGF.Builder.CreateLoad(PtrAddr), Align),
-                            Ty->getPointeeType(), Source);
+void RegionCodeGenTy::operator()(CodeGenFunction &CGF) const {
+  CodeGenFunction::RunCleanupsScope Scope(CGF);
+  if (PrePostAction) {
+    CGF.EHStack.pushCleanup<CleanupTy>(NormalAndEHCleanup, PrePostAction);
+    Callback(CodeGen, CGF, *PrePostAction);
+  } else {
+    PrePostActionTy Action;
+    Callback(CodeGen, CGF, Action);
+  }
 }
 
 LValue CGOpenMPRegionInfo::getThreadIDVariableLValue(CodeGenFunction &CGF) {
-  return emitLoadOfPointerLValue(CGF,
-                                 CGF.GetAddrOfLocalVar(getThreadIDVariable()),
-                                 getThreadIDVariable()->getType());
+  return CGF.EmitLoadOfPointerLValue(
+      CGF.GetAddrOfLocalVar(getThreadIDVariable()),
+      getThreadIDVariable()->getType()->castAs<PointerType>());
 }
 
 void CGOpenMPRegionInfo::EmitBody(CodeGenFunction &CGF, const Stmt * /*S*/) {
@@ -295,10 +709,7 @@ void CGOpenMPRegionInfo::EmitBody(CodeGenFunction &CGF, const Stmt * /*S*/) {
   // The point of exit cannot be a branch out of the structured block.
   // longjmp() and throw() must not violate the entry/exit criteria.
   CGF.EHStack.pushTerminate();
-  {
-    CodeGenFunction::RunCleanupsScope Scope(CGF);
-    CodeGen(CGF);
-  }
+  CodeGen(CGF);
   CGF.EHStack.popTerminate();
 }
 
@@ -310,16 +721,11 @@ LValue CGOpenMPTaskOutlinedRegionInfo::getThreadIDVariableLValue(
 }
 
 CGOpenMPRuntime::CGOpenMPRuntime(CodeGenModule &CGM)
-    : CGM(CGM), DefaultOpenMPPSource(nullptr), KmpRoutineEntryPtrTy(nullptr),
-      OffloadEntriesInfoManager(CGM) {
+    : CGM(CGM), OffloadEntriesInfoManager(CGM) {
   IdentTy = llvm::StructType::create(
       "ident_t", CGM.Int32Ty /* reserved_1 */, CGM.Int32Ty /* flags */,
       CGM.Int32Ty /* reserved_2 */, CGM.Int32Ty /* reserved_3 */,
       CGM.Int8PtrTy /* psource */, nullptr);
-  // Build void (*kmpc_micro)(kmp_int32 *global_tid, kmp_int32 *bound_tid,...)
-  llvm::Type *MicroParams[] = {llvm::PointerType::getUnqual(CGM.Int32Ty),
-                               llvm::PointerType::getUnqual(CGM.Int32Ty)};
-  Kmpc_MicroTy = llvm::FunctionType::get(CGM.VoidTy, MicroParams, true);
   KmpCriticalNameTy = llvm::ArrayType::get(CGM.Int32Ty, /*NumElements*/ 8);
 
   loadOffloadInfoMetadata();
@@ -329,6 +735,90 @@ void CGOpenMPRuntime::clear() {
   InternalVars.clear();
 }
 
+static llvm::Function *
+emitCombinerOrInitializer(CodeGenModule &CGM, QualType Ty,
+                          const Expr *CombinerInitializer, const VarDecl *In,
+                          const VarDecl *Out, bool IsCombiner) {
+  // void .omp_combiner.(Ty *in, Ty *out);
+  auto &C = CGM.getContext();
+  QualType PtrTy = C.getPointerType(Ty).withRestrict();
+  FunctionArgList Args;
+  ImplicitParamDecl OmpOutParm(C, /*DC=*/nullptr, Out->getLocation(),
+                               /*Id=*/nullptr, PtrTy);
+  ImplicitParamDecl OmpInParm(C, /*DC=*/nullptr, In->getLocation(),
+                              /*Id=*/nullptr, PtrTy);
+  Args.push_back(&OmpOutParm);
+  Args.push_back(&OmpInParm);
+  auto &FnInfo =
+      CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
+  auto *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+  auto *Fn = llvm::Function::Create(
+      FnTy, llvm::GlobalValue::InternalLinkage,
+      IsCombiner ? ".omp_combiner." : ".omp_initializer.", &CGM.getModule());
+  CGM.SetInternalFunctionAttributes(/*D=*/nullptr, Fn, FnInfo);
+  Fn->addFnAttr(llvm::Attribute::AlwaysInline);
+  CodeGenFunction CGF(CGM);
+  // Map "T omp_in;" variable to "*omp_in_parm" value in all expressions.
+  // Map "T omp_out;" variable to "*omp_out_parm" value in all expressions.
+  CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args);
+  CodeGenFunction::OMPPrivateScope Scope(CGF);
+  Address AddrIn = CGF.GetAddrOfLocalVar(&OmpInParm);
+  Scope.addPrivate(In, [&CGF, AddrIn, PtrTy]() -> Address {
+    return CGF.EmitLoadOfPointerLValue(AddrIn, PtrTy->castAs<PointerType>())
+        .getAddress();
+  });
+  Address AddrOut = CGF.GetAddrOfLocalVar(&OmpOutParm);
+  Scope.addPrivate(Out, [&CGF, AddrOut, PtrTy]() -> Address {
+    return CGF.EmitLoadOfPointerLValue(AddrOut, PtrTy->castAs<PointerType>())
+        .getAddress();
+  });
+  (void)Scope.Privatize();
+  CGF.EmitIgnoredExpr(CombinerInitializer);
+  Scope.ForceCleanup();
+  CGF.FinishFunction();
+  return Fn;
+}
+
+void CGOpenMPRuntime::emitUserDefinedReduction(
+    CodeGenFunction *CGF, const OMPDeclareReductionDecl *D) {
+  if (UDRMap.count(D) > 0)
+    return;
+  auto &C = CGM.getContext();
+  if (!In || !Out) {
+    In = &C.Idents.get("omp_in");
+    Out = &C.Idents.get("omp_out");
+  }
+  llvm::Function *Combiner = emitCombinerOrInitializer(
+      CGM, D->getType(), D->getCombiner(), cast<VarDecl>(D->lookup(In).front()),
+      cast<VarDecl>(D->lookup(Out).front()),
+      /*IsCombiner=*/true);
+  llvm::Function *Initializer = nullptr;
+  if (auto *Init = D->getInitializer()) {
+    if (!Priv || !Orig) {
+      Priv = &C.Idents.get("omp_priv");
+      Orig = &C.Idents.get("omp_orig");
+    }
+    Initializer = emitCombinerOrInitializer(
+        CGM, D->getType(), Init, cast<VarDecl>(D->lookup(Orig).front()),
+        cast<VarDecl>(D->lookup(Priv).front()),
+        /*IsCombiner=*/false);
+  }
+  UDRMap.insert(std::make_pair(D, std::make_pair(Combiner, Initializer)));
+  if (CGF) {
+    auto &Decls = FunctionUDRMap.FindAndConstruct(CGF->CurFn);
+    Decls.second.push_back(D);
+  }
+}
+
+std::pair<llvm::Function *, llvm::Function *>
+CGOpenMPRuntime::getUserDefinedReduction(const OMPDeclareReductionDecl *D) {
+  auto I = UDRMap.find(D);
+  if (I != UDRMap.end())
+    return I->second;
+  emitUserDefinedReduction(/*CGF=*/nullptr, D);
+  return UDRMap.lookup(D);
+}
+
 // Layout information for ident_t.
 static CharUnits getIdentAlign(CodeGenModule &CGM) {
   return CGM.getPointerAlign();
@@ -337,18 +827,18 @@ static CharUnits getIdentSize(CodeGenModule &CGM) {
   assert((4 * CGM.getPointerSize()).isMultipleOf(CGM.getPointerAlign()));
   return CharUnits::fromQuantity(16) + CGM.getPointerSize();
 }
-static CharUnits getOffsetOfIdentField(CGOpenMPRuntime::IdentFieldIndex Field) {
+static CharUnits getOffsetOfIdentField(IdentFieldIndex Field) {
   // All the fields except the last are i32, so this works beautifully.
   return unsigned(Field) * CharUnits::fromQuantity(4);
 }
 static Address createIdentFieldGEP(CodeGenFunction &CGF, Address Addr,
-                                   CGOpenMPRuntime::IdentFieldIndex Field,
+                                   IdentFieldIndex Field,
                                    const llvm::Twine &Name = "") {
   auto Offset = getOffsetOfIdentField(Field);
   return CGF.Builder.CreateStructGEP(Addr, Field, Offset, Name);
 }
 
-llvm::Value *CGOpenMPRuntime::emitParallelOutlinedFunction(
+llvm::Value *CGOpenMPRuntime::emitParallelOrTeamsOutlinedFunction(
     const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
     OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen) {
   assert(ThreadIDVar->getType()->isPointerType() &&
@@ -370,19 +860,39 @@ llvm::Value *CGOpenMPRuntime::emitParallelOutlinedFunction(
 
 llvm::Value *CGOpenMPRuntime::emitTaskOutlinedFunction(
     const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
-    OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen) {
+    const VarDecl *PartIDVar, const VarDecl *TaskTVar,
+    OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen,
+    bool Tied, unsigned &NumberOfParts) {
+  auto &&UntiedCodeGen = [this, &D, TaskTVar](CodeGenFunction &CGF,
+                                              PrePostActionTy &) {
+    auto *ThreadID = getThreadID(CGF, D.getLocStart());
+    auto *UpLoc = emitUpdateLocation(CGF, D.getLocStart());
+    llvm::Value *TaskArgs[] = {
+        UpLoc, ThreadID,
+        CGF.EmitLoadOfPointerLValue(CGF.GetAddrOfLocalVar(TaskTVar),
+                                    TaskTVar->getType()->castAs<PointerType>())
+            .getPointer()};
+    CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_omp_task), TaskArgs);
+  };
+  CGOpenMPTaskOutlinedRegionInfo::UntiedTaskActionTy Action(Tied, PartIDVar,
+                                                            UntiedCodeGen);
+  CodeGen.setAction(Action);
   assert(!ThreadIDVar->getType()->isPointerType() &&
          "thread id variable must be of type kmp_int32 for tasks");
   auto *CS = cast<CapturedStmt>(D.getAssociatedStmt());
+  auto *TD = dyn_cast<OMPTaskDirective>(&D);
   CodeGenFunction CGF(CGM, true);
   CGOpenMPTaskOutlinedRegionInfo CGInfo(*CS, ThreadIDVar, CodeGen,
                                         InnermostKind,
-                                        cast<OMPTaskDirective>(D).hasCancel());
+                                        TD ? TD->hasCancel() : false, Action);
   CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CGInfo);
-  return CGF.GenerateCapturedStmtFunction(*CS);
+  auto *Res = CGF.GenerateCapturedStmtFunction(*CS);
+  if (!Tied)
+    NumberOfParts = Action.getNumberOfParts();
+  return Res;
 }
 
-Address CGOpenMPRuntime::getOrCreateDefaultLocation(OpenMPLocationFlags Flags) {
+Address CGOpenMPRuntime::getOrCreateDefaultLocation(unsigned Flags) {
   CharUnits Align = getIdentAlign(CGM);
   llvm::Value *Entry = OpenMPDefaultLocMap.lookup(Flags);
   if (!Entry) {
@@ -399,7 +909,7 @@ Address CGOpenMPRuntime::getOrCreateDefaultLocation(OpenMPLocationFlags Flags) {
     auto DefaultOpenMPLocation = new llvm::GlobalVariable(
         CGM.getModule(), IdentTy, /*isConstant*/ true,
         llvm::GlobalValue::PrivateLinkage, /*Initializer*/ nullptr);
-    DefaultOpenMPLocation->setUnnamedAddr(true);
+    DefaultOpenMPLocation->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
     DefaultOpenMPLocation->setAlignment(Align.getQuantity());
 
     llvm::Constant *Zero = llvm::ConstantInt::get(CGM.Int32Ty, 0, true);
@@ -415,9 +925,10 @@ Address CGOpenMPRuntime::getOrCreateDefaultLocation(OpenMPLocationFlags Flags) {
 
 llvm::Value *CGOpenMPRuntime::emitUpdateLocation(CodeGenFunction &CGF,
                                                  SourceLocation Loc,
-                                                 OpenMPLocationFlags Flags) {
+                                                 unsigned Flags) {
+  Flags |= OMP_IDENT_KMPC;
   // If no debug info is generated - return global default location.
-  if (CGM.getCodeGenOpts().getDebugInfo() == CodeGenOptions::NoDebugInfo ||
+  if (CGM.getCodeGenOpts().getDebugInfo() == codegenoptions::NoDebugInfo ||
       Loc.isInvalid())
     return getOrCreateDefaultLocation(Flags).getPointer();
 
@@ -517,20 +1028,34 @@ void CGOpenMPRuntime::functionFinished(CodeGenFunction &CGF) {
   assert(CGF.CurFn && "No function in current CodeGenFunction.");
   if (OpenMPLocThreadIDMap.count(CGF.CurFn))
     OpenMPLocThreadIDMap.erase(CGF.CurFn);
+  if (FunctionUDRMap.count(CGF.CurFn) > 0) {
+    for(auto *D : FunctionUDRMap[CGF.CurFn]) {
+      UDRMap.erase(D);
+    }
+    FunctionUDRMap.erase(CGF.CurFn);
+  }
 }
 
 llvm::Type *CGOpenMPRuntime::getIdentTyPointerTy() {
+  if (!IdentTy) {
+  }
   return llvm::PointerType::getUnqual(IdentTy);
 }
 
 llvm::Type *CGOpenMPRuntime::getKmpc_MicroPointerTy() {
+  if (!Kmpc_MicroTy) {
+    // Build void (*kmpc_micro)(kmp_int32 *global_tid, kmp_int32 *bound_tid,...)
+    llvm::Type *MicroParams[] = {llvm::PointerType::getUnqual(CGM.Int32Ty),
+                                 llvm::PointerType::getUnqual(CGM.Int32Ty)};
+    Kmpc_MicroTy = llvm::FunctionType::get(CGM.VoidTy, MicroParams, true);
+  }
   return llvm::PointerType::getUnqual(Kmpc_MicroTy);
 }
 
 llvm::Constant *
-CGOpenMPRuntime::createRuntimeFunction(OpenMPRTLFunction Function) {
+CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
   llvm::Constant *RTLFn = nullptr;
-  switch (Function) {
+  switch (static_cast<OpenMPRTLFunction>(Function)) {
   case OMPRTL__kmpc_fork_call: {
     // Build void __kmpc_fork_call(ident_t *loc, kmp_int32 argc, kmpc_micro
     // microtask, ...);
@@ -927,6 +1452,86 @@ CGOpenMPRuntime::createRuntimeFunction(OpenMPRTLFunction Function) {
     RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_cancel");
     break;
   }
+  case OMPRTL__kmpc_push_num_teams: {
+    // Build void kmpc_push_num_teams (ident_t loc, kmp_int32 global_tid,
+    // kmp_int32 num_teams, kmp_int32 num_threads)
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.Int32Ty,
+        CGM.Int32Ty};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_push_num_teams");
+    break;
+  }
+  case OMPRTL__kmpc_fork_teams: {
+    // Build void __kmpc_fork_teams(ident_t *loc, kmp_int32 argc, kmpc_micro
+    // microtask, ...);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
+                                getKmpc_MicroPointerTy()};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ true);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_fork_teams");
+    break;
+  }
+  case OMPRTL__kmpc_taskloop: {
+    // Build void __kmpc_taskloop(ident_t *loc, int gtid, kmp_task_t *task, int
+    // if_val, kmp_uint64 *lb, kmp_uint64 *ub, kmp_int64 st, int nogroup, int
+    // sched, kmp_uint64 grainsize, void *task_dup);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(),
+                                CGM.IntTy,
+                                CGM.VoidPtrTy,
+                                CGM.IntTy,
+                                CGM.Int64Ty->getPointerTo(),
+                                CGM.Int64Ty->getPointerTo(),
+                                CGM.Int64Ty,
+                                CGM.IntTy,
+                                CGM.IntTy,
+                                CGM.Int64Ty,
+                                CGM.VoidPtrTy};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_taskloop");
+    break;
+  }
+  case OMPRTL__kmpc_doacross_init: {
+    // Build void __kmpc_doacross_init(ident_t *loc, kmp_int32 gtid, kmp_int32
+    // num_dims, struct kmp_dim *dims);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(),
+                                CGM.Int32Ty,
+                                CGM.Int32Ty,
+                                CGM.VoidPtrTy};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_init");
+    break;
+  }
+  case OMPRTL__kmpc_doacross_fini: {
+    // Build void __kmpc_doacross_fini(ident_t *loc, kmp_int32 gtid);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_fini");
+    break;
+  }
+  case OMPRTL__kmpc_doacross_post: {
+    // Build void __kmpc_doacross_post(ident_t *loc, kmp_int32 gtid, kmp_int64
+    // *vec);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
+                                CGM.Int64Ty->getPointerTo()};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_post");
+    break;
+  }
+  case OMPRTL__kmpc_doacross_wait: {
+    // Build void __kmpc_doacross_wait(ident_t *loc, kmp_int32 gtid, kmp_int64
+    // *vec);
+    llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
+                                CGM.Int64Ty->getPointerTo()};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_wait");
+    break;
+  }
   case OMPRTL__tgt_target: {
     // Build int32_t __tgt_target(int32_t device_id, void *host_ptr, int32_t
     // arg_num, void** args_base, void **args, size_t *arg_sizes, int32_t
@@ -943,6 +1548,24 @@ CGOpenMPRuntime::createRuntimeFunction(OpenMPRTLFunction Function) {
     RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target");
     break;
   }
+  case OMPRTL__tgt_target_teams: {
+    // Build int32_t __tgt_target_teams(int32_t device_id, void *host_ptr,
+    // int32_t arg_num, void** args_base, void **args, size_t *arg_sizes,
+    // int32_t *arg_types, int32_t num_teams, int32_t thread_limit);
+    llvm::Type *TypeParams[] = {CGM.Int32Ty,
+                                CGM.VoidPtrTy,
+                                CGM.Int32Ty,
+                                CGM.VoidPtrPtrTy,
+                                CGM.VoidPtrPtrTy,
+                                CGM.SizeTy->getPointerTo(),
+                                CGM.Int32Ty->getPointerTo(),
+                                CGM.Int32Ty,
+                                CGM.Int32Ty};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_teams");
+    break;
+  }
   case OMPRTL__tgt_register_lib: {
     // Build void __tgt_register_lib(__tgt_bin_desc *desc);
     QualType ParamTy =
@@ -963,30 +1586,53 @@ CGOpenMPRuntime::createRuntimeFunction(OpenMPRTLFunction Function) {
     RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_unregister_lib");
     break;
   }
+  case OMPRTL__tgt_target_data_begin: {
+    // Build void __tgt_target_data_begin(int32_t device_id, int32_t arg_num,
+    // void** args_base, void **args, size_t *arg_sizes, int32_t *arg_types);
+    llvm::Type *TypeParams[] = {CGM.Int32Ty,
+                                CGM.Int32Ty,
+                                CGM.VoidPtrPtrTy,
+                                CGM.VoidPtrPtrTy,
+                                CGM.SizeTy->getPointerTo(),
+                                CGM.Int32Ty->getPointerTo()};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_data_begin");
+    break;
+  }
+  case OMPRTL__tgt_target_data_end: {
+    // Build void __tgt_target_data_end(int32_t device_id, int32_t arg_num,
+    // void** args_base, void **args, size_t *arg_sizes, int32_t *arg_types);
+    llvm::Type *TypeParams[] = {CGM.Int32Ty,
+                                CGM.Int32Ty,
+                                CGM.VoidPtrPtrTy,
+                                CGM.VoidPtrPtrTy,
+                                CGM.SizeTy->getPointerTo(),
+                                CGM.Int32Ty->getPointerTo()};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_data_end");
+    break;
+  }
+  case OMPRTL__tgt_target_data_update: {
+    // Build void __tgt_target_data_update(int32_t device_id, int32_t arg_num,
+    // void** args_base, void **args, size_t *arg_sizes, int32_t *arg_types);
+    llvm::Type *TypeParams[] = {CGM.Int32Ty,
+                                CGM.Int32Ty,
+                                CGM.VoidPtrPtrTy,
+                                CGM.VoidPtrPtrTy,
+                                CGM.SizeTy->getPointerTo(),
+                                CGM.Int32Ty->getPointerTo()};
+    llvm::FunctionType *FnTy =
+        llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
+    RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_data_update");
+    break;
+  }
   }
+  assert(RTLFn && "Unable to find OpenMP runtime function");
   return RTLFn;
 }
 
-static llvm::Value *getTypeSize(CodeGenFunction &CGF, QualType Ty) {
-  auto &C = CGF.getContext();
-  llvm::Value *Size = nullptr;
-  auto SizeInChars = C.getTypeSizeInChars(Ty);
-  if (SizeInChars.isZero()) {
-    // getTypeSizeInChars() returns 0 for a VLA.
-    while (auto *VAT = C.getAsVariableArrayType(Ty)) {
-      llvm::Value *ArraySize;
-      std::tie(ArraySize, Ty) = CGF.getVLASize(VAT);
-      Size = Size ? CGF.Builder.CreateNUWMul(Size, ArraySize) : ArraySize;
-    }
-    SizeInChars = C.getTypeSizeInChars(Ty);
-    assert(!SizeInChars.isZero());
-    Size = CGF.Builder.CreateNUWMul(
-        Size, llvm::ConstantInt::get(CGF.SizeTy, SizeInChars.getQuantity()));
-  } else
-    Size = llvm::ConstantInt::get(CGF.SizeTy, SizeInChars.getQuantity());
-  return Size;
-}
-
 llvm::Constant *CGOpenMPRuntime::createForStaticInitFunction(unsigned IVSize,
                                                              bool IVSigned) {
   assert((IVSize == 32 || IVSize == 64) &&
@@ -1144,9 +1790,8 @@ llvm::Function *CGOpenMPRuntime::emitThreadPrivateVarDefinition(
                             /*Id=*/nullptr, CGM.getContext().VoidPtrTy);
       Args.push_back(&Dst);
 
-      auto &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-          CGM.getContext().VoidPtrTy, Args, FunctionType::ExtInfo(),
-          /*isVariadic=*/false);
+      auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
+          CGM.getContext().VoidPtrTy, Args);
       auto FTy = CGM.getTypes().GetFunctionType(FI);
       auto Fn = CGM.CreateGlobalInitOrDestructFunction(
           FTy, ".__kmpc_global_ctor_.", FI, Loc);
@@ -1176,14 +1821,16 @@ llvm::Function *CGOpenMPRuntime::emitThreadPrivateVarDefinition(
                             /*Id=*/nullptr, CGM.getContext().VoidPtrTy);
       Args.push_back(&Dst);
 
-      auto &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-          CGM.getContext().VoidTy, Args, FunctionType::ExtInfo(),
-          /*isVariadic=*/false);
+      auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
+          CGM.getContext().VoidTy, Args);
       auto FTy = CGM.getTypes().GetFunctionType(FI);
       auto Fn = CGM.CreateGlobalInitOrDestructFunction(
           FTy, ".__kmpc_global_dtor_.", FI, Loc);
+      auto NL = ApplyDebugLocation::CreateEmpty(DtorCGF);
       DtorCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidTy, Fn, FI, Args,
                             SourceLocation());
+      // Create a scope with an artificial location for the body of this function.
+      auto AL = ApplyDebugLocation::CreateArtificial(DtorCGF);
       auto ArgVal = DtorCGF.EmitLoadOfScalar(
           DtorCGF.GetAddrOfLocalVar(&Dst),
           /*Volatile=*/false, CGM.getContext().VoidPtrTy, Dst.getLocation());
@@ -1251,12 +1898,10 @@ static void emitOMPIfClause(CodeGenFunction &CGF, const Expr *Cond,
   // the condition and the dead arm of the if/else.
   bool CondConstant;
   if (CGF.ConstantFoldsToSimpleInteger(Cond, CondConstant)) {
-    CodeGenFunction::RunCleanupsScope Scope(CGF);
-    if (CondConstant) {
+    if (CondConstant)
       ThenGen(CGF);
-    } else {
+    else
       ElseGen(CGF);
-    }
     return;
   }
 
@@ -1269,26 +1914,16 @@ static void emitOMPIfClause(CodeGenFunction &CGF, const Expr *Cond,
 
   // Emit the 'then' code.
   CGF.EmitBlock(ThenBlock);
-  {
-    CodeGenFunction::RunCleanupsScope ThenScope(CGF);
-    ThenGen(CGF);
-  }
+  ThenGen(CGF);
   CGF.EmitBranch(ContBlock);
   // Emit the 'else' code if present.
-  {
-    // There is no need to emit line number for unconditional branch.
-    auto NL = ApplyDebugLocation::CreateEmpty(CGF);
-    CGF.EmitBlock(ElseBlock);
-  }
-  {
-    CodeGenFunction::RunCleanupsScope ThenScope(CGF);
-    ElseGen(CGF);
-  }
-  {
-    // There is no need to emit line number for unconditional branch.
-    auto NL = ApplyDebugLocation::CreateEmpty(CGF);
-    CGF.EmitBranch(ContBlock);
-  }
+  // There is no need to emit line number for unconditional branch.
+  (void)ApplyDebugLocation::CreateEmpty(CGF);
+  CGF.EmitBlock(ElseBlock);
+  ElseGen(CGF);
+  // There is no need to emit line number for unconditional branch.
+  (void)ApplyDebugLocation::CreateEmpty(CGF);
+  CGF.EmitBranch(ContBlock);
   // Emit the continuation block for code after the if.
   CGF.EmitBlock(ContBlock, /*IsFinished=*/true);
 }
@@ -1300,34 +1935,36 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
   if (!CGF.HaveInsertPoint())
     return;
   auto *RTLoc = emitUpdateLocation(CGF, Loc);
-  auto &&ThenGen = [this, OutlinedFn, CapturedVars,
-                    RTLoc](CodeGenFunction &CGF) {
+  auto &&ThenGen = [OutlinedFn, CapturedVars, RTLoc](CodeGenFunction &CGF,
+                                                     PrePostActionTy &) {
     // Build call __kmpc_fork_call(loc, n, microtask, var1, .., varn);
+    auto &RT = CGF.CGM.getOpenMPRuntime();
     llvm::Value *Args[] = {
         RTLoc,
         CGF.Builder.getInt32(CapturedVars.size()), // Number of captured vars
-        CGF.Builder.CreateBitCast(OutlinedFn, getKmpc_MicroPointerTy())};
+        CGF.Builder.CreateBitCast(OutlinedFn, RT.getKmpc_MicroPointerTy())};
     llvm::SmallVector<llvm::Value *, 16> RealArgs;
     RealArgs.append(std::begin(Args), std::end(Args));
     RealArgs.append(CapturedVars.begin(), CapturedVars.end());
 
-    auto RTLFn = createRuntimeFunction(OMPRTL__kmpc_fork_call);
+    auto RTLFn = RT.createRuntimeFunction(OMPRTL__kmpc_fork_call);
     CGF.EmitRuntimeCall(RTLFn, RealArgs);
   };
-  auto &&ElseGen = [this, OutlinedFn, CapturedVars, RTLoc,
-                    Loc](CodeGenFunction &CGF) {
-    auto ThreadID = getThreadID(CGF, Loc);
+  auto &&ElseGen = [OutlinedFn, CapturedVars, RTLoc, Loc](CodeGenFunction &CGF,
+                                                          PrePostActionTy &) {
+    auto &RT = CGF.CGM.getOpenMPRuntime();
+    auto ThreadID = RT.getThreadID(CGF, Loc);
     // Build calls:
     // __kmpc_serialized_parallel(&Loc, GTid);
     llvm::Value *Args[] = {RTLoc, ThreadID};
-    CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_serialized_parallel),
-                        Args);
+    CGF.EmitRuntimeCall(
+        RT.createRuntimeFunction(OMPRTL__kmpc_serialized_parallel), Args);
 
     // OutlinedFn(&GTid, &zero, CapturedStruct);
-    auto ThreadIDAddr = emitThreadIDAddress(CGF, Loc);
+    auto ThreadIDAddr = RT.emitThreadIDAddress(CGF, Loc);
     Address ZeroAddr =
-      CGF.CreateTempAlloca(CGF.Int32Ty, CharUnits::fromQuantity(4),
-                           /*Name*/ ".zero.addr");
+        CGF.CreateTempAlloca(CGF.Int32Ty, CharUnits::fromQuantity(4),
+                             /*Name*/ ".zero.addr");
     CGF.InitTempAlloca(ZeroAddr, CGF.Builder.getInt32(/*C*/ 0));
     llvm::SmallVector<llvm::Value *, 16> OutlinedFnArgs;
     OutlinedFnArgs.push_back(ThreadIDAddr.getPointer());
@@ -1336,15 +1973,16 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
     CGF.EmitCallOrInvoke(OutlinedFn, OutlinedFnArgs);
 
     // __kmpc_end_serialized_parallel(&Loc, GTid);
-    llvm::Value *EndArgs[] = {emitUpdateLocation(CGF, Loc), ThreadID};
+    llvm::Value *EndArgs[] = {RT.emitUpdateLocation(CGF, Loc), ThreadID};
     CGF.EmitRuntimeCall(
-        createRuntimeFunction(OMPRTL__kmpc_end_serialized_parallel), EndArgs);
+        RT.createRuntimeFunction(OMPRTL__kmpc_end_serialized_parallel),
+        EndArgs);
   };
-  if (IfCond) {
+  if (IfCond)
     emitOMPIfClause(CGF, IfCond, ThenGen, ElseGen);
-  } else {
-    CodeGenFunction::RunCleanupsScope Scope(CGF);
-    ThenGen(CGF);
+  else {
+    RegionCodeGenTy ThenRCG(ThenGen);
+    ThenRCG(CGF);
   }
 }
 
@@ -1397,20 +2035,39 @@ llvm::Value *CGOpenMPRuntime::getCriticalRegionLock(StringRef CriticalName) {
 }
 
 namespace {
-template <size_t N> class CallEndCleanup final : public EHScopeStack::Cleanup {
-  llvm::Value *Callee;
-  llvm::Value *Args[N];
+/// Common pre(post)-action for different OpenMP constructs.
+class CommonActionTy final : public PrePostActionTy {
+  llvm::Value *EnterCallee;
+  ArrayRef<llvm::Value *> EnterArgs;
+  llvm::Value *ExitCallee;
+  ArrayRef<llvm::Value *> ExitArgs;
+  bool Conditional;
+  llvm::BasicBlock *ContBlock = nullptr;
 
 public:
-  CallEndCleanup(llvm::Value *Callee, ArrayRef<llvm::Value *> CleanupArgs)
-      : Callee(Callee) {
-    assert(CleanupArgs.size() == N);
-    std::copy(CleanupArgs.begin(), CleanupArgs.end(), std::begin(Args));
+  CommonActionTy(llvm::Value *EnterCallee, ArrayRef<llvm::Value *> EnterArgs,
+                 llvm::Value *ExitCallee, ArrayRef<llvm::Value *> ExitArgs,
+                 bool Conditional = false)
+      : EnterCallee(EnterCallee), EnterArgs(EnterArgs), ExitCallee(ExitCallee),
+        ExitArgs(ExitArgs), Conditional(Conditional) {}
+  void Enter(CodeGenFunction &CGF) override {
+    llvm::Value *EnterRes = CGF.EmitRuntimeCall(EnterCallee, EnterArgs);
+    if (Conditional) {
+      llvm::Value *CallBool = CGF.Builder.CreateIsNotNull(EnterRes);
+      auto *ThenBlock = CGF.createBasicBlock("omp_if.then");
+      ContBlock = CGF.createBasicBlock("omp_if.end");
+      // Generate the branch (If-stmt)
+      CGF.Builder.CreateCondBr(CallBool, ThenBlock, ContBlock);
+      CGF.EmitBlock(ThenBlock);
+    }
   }
-  void Emit(CodeGenFunction &CGF, Flags /*flags*/) override {
-    if (!CGF.HaveInsertPoint())
-      return;
-    CGF.EmitRuntimeCall(Callee, Args);
+  void Done(CodeGenFunction &CGF) {
+    // Emit the rest of blocks/branches
+    CGF.EmitBranch(ContBlock);
+    CGF.EmitBlock(ContBlock, true);
+  }
+  void Exit(CodeGenFunction &CGF) override {
+    CGF.EmitRuntimeCall(ExitCallee, ExitArgs);
   }
 };
 } // anonymous namespace
@@ -1425,45 +2082,22 @@ void CGOpenMPRuntime::emitCriticalRegion(CodeGenFunction &CGF,
   // Prepare arguments and build a call to __kmpc_critical
   if (!CGF.HaveInsertPoint())
     return;
-  CodeGenFunction::RunCleanupsScope Scope(CGF);
   llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
                          getCriticalRegionLock(CriticalName)};
+  llvm::SmallVector<llvm::Value *, 4> EnterArgs(std::begin(Args),
+                                                std::end(Args));
   if (Hint) {
-    llvm::SmallVector<llvm::Value *, 8> ArgsWithHint(std::begin(Args),
-                                                     std::end(Args));
-    auto *HintVal = CGF.EmitScalarExpr(Hint);
-    ArgsWithHint.push_back(
-        CGF.Builder.CreateIntCast(HintVal, CGM.IntPtrTy, /*isSigned=*/false));
-    CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_critical_with_hint),
-                        ArgsWithHint);
-  } else
-    CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_critical), Args);
-  // Build a call to __kmpc_end_critical
-  CGF.EHStack.pushCleanup<CallEndCleanup<std::extent<decltype(Args)>::value>>(
-      NormalAndEHCleanup, createRuntimeFunction(OMPRTL__kmpc_end_critical),
-      llvm::makeArrayRef(Args));
+    EnterArgs.push_back(CGF.Builder.CreateIntCast(
+        CGF.EmitScalarExpr(Hint), CGM.IntPtrTy, /*isSigned=*/false));
+  }
+  CommonActionTy Action(
+      createRuntimeFunction(Hint ? OMPRTL__kmpc_critical_with_hint
+                                 : OMPRTL__kmpc_critical),
+      EnterArgs, createRuntimeFunction(OMPRTL__kmpc_end_critical), Args);
+  CriticalOpGen.setAction(Action);
   emitInlinedDirective(CGF, OMPD_critical, CriticalOpGen);
 }
 
-static void emitIfStmt(CodeGenFunction &CGF, llvm::Value *IfCond,
-                       OpenMPDirectiveKind Kind, SourceLocation Loc,
-                       const RegionCodeGenTy &BodyOpGen) {
-  llvm::Value *CallBool = CGF.EmitScalarConversion(
-      IfCond,
-      CGF.getContext().getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/true),
-      CGF.getContext().BoolTy, Loc);
-
-  auto *ThenBlock = CGF.createBasicBlock("omp_if.then");
-  auto *ContBlock = CGF.createBasicBlock("omp_if.end");
-  // Generate the branch (If-stmt)
-  CGF.Builder.CreateCondBr(CallBool, ThenBlock, ContBlock);
-  CGF.EmitBlock(ThenBlock);
-  CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, Kind, BodyOpGen);
-  // Emit the rest of bblocks/branches
-  CGF.EmitBranch(ContBlock);
-  CGF.EmitBlock(ContBlock, true);
-}
-
 void CGOpenMPRuntime::emitMasterRegion(CodeGenFunction &CGF,
                                        const RegionCodeGenTy &MasterOpGen,
                                        SourceLocation Loc) {
@@ -1475,18 +2109,12 @@ void CGOpenMPRuntime::emitMasterRegion(CodeGenFunction &CGF,
   // }
   // Prepare arguments and build a call to __kmpc_master
   llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
-  auto *IsMaster =
-      CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_master), Args);
-  typedef CallEndCleanup<std::extent<decltype(Args)>::value>
-      MasterCallEndCleanup;
-  emitIfStmt(
-      CGF, IsMaster, OMPD_master, Loc, [&](CodeGenFunction &CGF) -> void {
-        CodeGenFunction::RunCleanupsScope Scope(CGF);
-        CGF.EHStack.pushCleanup<MasterCallEndCleanup>(
-            NormalAndEHCleanup, createRuntimeFunction(OMPRTL__kmpc_end_master),
-            llvm::makeArrayRef(Args));
-        MasterOpGen(CGF);
-      });
+  CommonActionTy Action(createRuntimeFunction(OMPRTL__kmpc_master), Args,
+                        createRuntimeFunction(OMPRTL__kmpc_end_master), Args,
+                        /*Conditional=*/true);
+  MasterOpGen.setAction(Action);
+  emitInlinedDirective(CGF, OMPD_master, MasterOpGen);
+  Action.Done(CGF);
 }
 
 void CGOpenMPRuntime::emitTaskyieldCall(CodeGenFunction &CGF,
@@ -1498,6 +2126,8 @@ void CGOpenMPRuntime::emitTaskyieldCall(CodeGenFunction &CGF,
       emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
       llvm::ConstantInt::get(CGM.IntTy, /*V=*/0, /*isSigned=*/true)};
   CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_omp_taskyield), Args);
+  if (auto *Region = dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo))
+    Region->emitUntiedSwitch(CGF);
 }
 
 void CGOpenMPRuntime::emitTaskgroupRegion(CodeGenFunction &CGF,
@@ -1509,16 +2139,12 @@ void CGOpenMPRuntime::emitTaskgroupRegion(CodeGenFunction &CGF,
   // TaskgroupOpGen();
   // __kmpc_end_taskgroup(ident_t *, gtid);
   // Prepare arguments and build a call to __kmpc_taskgroup
-  {
-    CodeGenFunction::RunCleanupsScope Scope(CGF);
-    llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
-    CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_taskgroup), Args);
-    // Build a call to __kmpc_end_taskgroup
-    CGF.EHStack.pushCleanup<CallEndCleanup<std::extent<decltype(Args)>::value>>(
-        NormalAndEHCleanup, createRuntimeFunction(OMPRTL__kmpc_end_taskgroup),
-        llvm::makeArrayRef(Args));
-    emitInlinedDirective(CGF, OMPD_taskgroup, TaskgroupOpGen);
-  }
+  llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
+  CommonActionTy Action(createRuntimeFunction(OMPRTL__kmpc_taskgroup), Args,
+                        createRuntimeFunction(OMPRTL__kmpc_end_taskgroup),
+                        Args);
+  TaskgroupOpGen.setAction(Action);
+  emitInlinedDirective(CGF, OMPD_taskgroup, TaskgroupOpGen);
 }
 
 /// Given an array of pointers to variables, project the address of a
@@ -1549,9 +2175,7 @@ static llvm::Value *emitCopyprivateCopyFunction(
                            C.VoidPtrTy);
   Args.push_back(&LHSArg);
   Args.push_back(&RHSArg);
-  FunctionType::ExtInfo EI;
-  auto &CGFI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      C.VoidTy, Args, EI, /*isVariadic=*/false);
+  auto &CGFI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
   auto *Fn = llvm::Function::Create(
       CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
       ".omp.copyprivate.copy_func", &CGM.getModule());
@@ -1616,22 +2240,16 @@ void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
   }
   // Prepare arguments and build a call to __kmpc_single
   llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
-  auto *IsSingle =
-      CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_single), Args);
-  typedef CallEndCleanup<std::extent<decltype(Args)>::value>
-      SingleCallEndCleanup;
-  emitIfStmt(
-      CGF, IsSingle, OMPD_single, Loc, [&](CodeGenFunction &CGF) -> void {
-        CodeGenFunction::RunCleanupsScope Scope(CGF);
-        CGF.EHStack.pushCleanup<SingleCallEndCleanup>(
-            NormalAndEHCleanup, createRuntimeFunction(OMPRTL__kmpc_end_single),
-            llvm::makeArrayRef(Args));
-        SingleOpGen(CGF);
-        if (DidIt.isValid()) {
-          // did_it = 1;
-          CGF.Builder.CreateStore(CGF.Builder.getInt32(1), DidIt);
-        }
-      });
+  CommonActionTy Action(createRuntimeFunction(OMPRTL__kmpc_single), Args,
+                        createRuntimeFunction(OMPRTL__kmpc_end_single), Args,
+                        /*Conditional=*/true);
+  SingleOpGen.setAction(Action);
+  emitInlinedDirective(CGF, OMPD_single, SingleOpGen);
+  if (DidIt.isValid()) {
+    // did_it = 1;
+    CGF.Builder.CreateStore(CGF.Builder.getInt32(1), DidIt);
+  }
+  Action.Done(CGF);
   // call __kmpc_copyprivate(ident_t *, gtid, <buf_size>, <copyprivate list>,
   // <copy_func>, did_it);
   if (DidIt.isValid()) {
@@ -1655,7 +2273,7 @@ void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
     auto *CpyFn = emitCopyprivateCopyFunction(
         CGM, CGF.ConvertTypeForMem(CopyprivateArrayTy)->getPointerTo(),
         CopyprivateVars, SrcExprs, DstExprs, AssignmentOps);
-    auto *BufSize = getTypeSize(CGF, CopyprivateArrayTy);
+    auto *BufSize = CGF.getTypeSize(CopyprivateArrayTy);
     Address CL =
       CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(CopyprivateList,
                                                       CGF.VoidPtrTy);
@@ -1681,14 +2299,14 @@ void CGOpenMPRuntime::emitOrderedRegion(CodeGenFunction &CGF,
   // OrderedOpGen();
   // __kmpc_end_ordered(ident_t *, gtid);
   // Prepare arguments and build a call to __kmpc_ordered
-  CodeGenFunction::RunCleanupsScope Scope(CGF);
   if (IsThreads) {
     llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
-    CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_ordered), Args);
-    // Build a call to __kmpc_end_ordered
-    CGF.EHStack.pushCleanup<CallEndCleanup<std::extent<decltype(Args)>::value>>(
-        NormalAndEHCleanup, createRuntimeFunction(OMPRTL__kmpc_end_ordered),
-        llvm::makeArrayRef(Args));
+    CommonActionTy Action(createRuntimeFunction(OMPRTL__kmpc_ordered), Args,
+                          createRuntimeFunction(OMPRTL__kmpc_end_ordered),
+                          Args);
+    OrderedOpGen.setAction(Action);
+    emitInlinedDirective(CGF, OMPD_ordered, OrderedOpGen);
+    return;
   }
   emitInlinedDirective(CGF, OMPD_ordered, OrderedOpGen);
 }
@@ -1700,21 +2318,17 @@ void CGOpenMPRuntime::emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
     return;
   // Build call __kmpc_cancel_barrier(loc, thread_id);
   // Build call __kmpc_barrier(loc, thread_id);
-  OpenMPLocationFlags Flags = OMP_IDENT_KMPC;
-  if (Kind == OMPD_for) {
-    Flags =
-        static_cast<OpenMPLocationFlags>(Flags | OMP_IDENT_BARRIER_IMPL_FOR);
-  } else if (Kind == OMPD_sections) {
-    Flags = static_cast<OpenMPLocationFlags>(Flags |
-                                             OMP_IDENT_BARRIER_IMPL_SECTIONS);
-  } else if (Kind == OMPD_single) {
-    Flags =
-        static_cast<OpenMPLocationFlags>(Flags | OMP_IDENT_BARRIER_IMPL_SINGLE);
-  } else if (Kind == OMPD_barrier) {
-    Flags = static_cast<OpenMPLocationFlags>(Flags | OMP_IDENT_BARRIER_EXPL);
-  } else {
-    Flags = static_cast<OpenMPLocationFlags>(Flags | OMP_IDENT_BARRIER_IMPL);
-  }
+  unsigned Flags;
+  if (Kind == OMPD_for)
+    Flags = OMP_IDENT_BARRIER_IMPL_FOR;
+  else if (Kind == OMPD_sections)
+    Flags = OMP_IDENT_BARRIER_IMPL_SECTIONS;
+  else if (Kind == OMPD_single)
+    Flags = OMP_IDENT_BARRIER_IMPL_SINGLE;
+  else if (Kind == OMPD_barrier)
+    Flags = OMP_IDENT_BARRIER_EXPL;
+  else
+    Flags = OMP_IDENT_BARRIER_IMPL;
   // Build call __kmpc_cancel_barrier(loc, thread_id) or __kmpc_barrier(loc,
   // thread_id);
   llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc, Flags),
@@ -1745,28 +2359,6 @@ void CGOpenMPRuntime::emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
   CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_barrier), Args);
 }
 
-/// \brief Schedule types for 'omp for' loops (these enumerators are taken from
-/// the enum sched_type in kmp.h).
-enum OpenMPSchedType {
-  /// \brief Lower bound for default (unordered) versions.
-  OMP_sch_lower = 32,
-  OMP_sch_static_chunked = 33,
-  OMP_sch_static = 34,
-  OMP_sch_dynamic_chunked = 35,
-  OMP_sch_guided_chunked = 36,
-  OMP_sch_runtime = 37,
-  OMP_sch_auto = 38,
-  /// \brief Lower bound for 'ordered' versions.
-  OMP_ord_lower = 64,
-  OMP_ord_static_chunked = 65,
-  OMP_ord_static = 66,
-  OMP_ord_dynamic_chunked = 67,
-  OMP_ord_guided_chunked = 68,
-  OMP_ord_runtime = 69,
-  OMP_ord_auto = 70,
-  OMP_sch_default = OMP_sch_static,
-};
-
 /// \brief Map the OpenMP loop schedule to the runtime enumeration.
 static OpenMPSchedType getRuntimeSchedule(OpenMPScheduleClauseKind ScheduleKind,
                                           bool Chunked, bool Ordered) {
@@ -1789,12 +2381,26 @@ static OpenMPSchedType getRuntimeSchedule(OpenMPScheduleClauseKind ScheduleKind,
   llvm_unreachable("Unexpected runtime schedule");
 }
 
+/// \brief Map the OpenMP distribute schedule to the runtime enumeration.
+static OpenMPSchedType
+getRuntimeSchedule(OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) {
+  // only static is allowed for dist_schedule
+  return Chunked ? OMP_dist_sch_static_chunked : OMP_dist_sch_static;
+}
+
 bool CGOpenMPRuntime::isStaticNonchunked(OpenMPScheduleClauseKind ScheduleKind,
                                          bool Chunked) const {
   auto Schedule = getRuntimeSchedule(ScheduleKind, Chunked, /*Ordered=*/false);
   return Schedule == OMP_sch_static;
 }
 
+bool CGOpenMPRuntime::isStaticNonchunked(
+    OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) const {
+  auto Schedule = getRuntimeSchedule(ScheduleKind, Chunked);
+  return Schedule == OMP_dist_sch_static;
+}
+
+
 bool CGOpenMPRuntime::isDynamic(OpenMPScheduleClauseKind ScheduleKind) const {
   auto Schedule =
       getRuntimeSchedule(ScheduleKind, /*Chunked=*/false, /*Ordered=*/false);
@@ -1802,19 +2408,57 @@ bool CGOpenMPRuntime::isDynamic(OpenMPScheduleClauseKind ScheduleKind) const {
   return Schedule != OMP_sch_static;
 }
 
+static int addMonoNonMonoModifier(OpenMPSchedType Schedule,
+                                  OpenMPScheduleClauseModifier M1,
+                                  OpenMPScheduleClauseModifier M2) {
+  int Modifier = 0;
+  switch (M1) {
+  case OMPC_SCHEDULE_MODIFIER_monotonic:
+    Modifier = OMP_sch_modifier_monotonic;
+    break;
+  case OMPC_SCHEDULE_MODIFIER_nonmonotonic:
+    Modifier = OMP_sch_modifier_nonmonotonic;
+    break;
+  case OMPC_SCHEDULE_MODIFIER_simd:
+    if (Schedule == OMP_sch_static_chunked)
+      Schedule = OMP_sch_static_balanced_chunked;
+    break;
+  case OMPC_SCHEDULE_MODIFIER_last:
+  case OMPC_SCHEDULE_MODIFIER_unknown:
+    break;
+  }
+  switch (M2) {
+  case OMPC_SCHEDULE_MODIFIER_monotonic:
+    Modifier = OMP_sch_modifier_monotonic;
+    break;
+  case OMPC_SCHEDULE_MODIFIER_nonmonotonic:
+    Modifier = OMP_sch_modifier_nonmonotonic;
+    break;
+  case OMPC_SCHEDULE_MODIFIER_simd:
+    if (Schedule == OMP_sch_static_chunked)
+      Schedule = OMP_sch_static_balanced_chunked;
+    break;
+  case OMPC_SCHEDULE_MODIFIER_last:
+  case OMPC_SCHEDULE_MODIFIER_unknown:
+    break;
+  }
+  return Schedule | Modifier;
+}
+
 void CGOpenMPRuntime::emitForDispatchInit(CodeGenFunction &CGF,
                                           SourceLocation Loc,
-                                          OpenMPScheduleClauseKind ScheduleKind,
+                                          const OpenMPScheduleTy &ScheduleKind,
                                           unsigned IVSize, bool IVSigned,
                                           bool Ordered, llvm::Value *UB,
                                           llvm::Value *Chunk) {
   if (!CGF.HaveInsertPoint())
     return;
   OpenMPSchedType Schedule =
-      getRuntimeSchedule(ScheduleKind, Chunk != nullptr, Ordered);
+      getRuntimeSchedule(ScheduleKind.Schedule, Chunk != nullptr, Ordered);
   assert(Ordered ||
          (Schedule != OMP_sch_static && Schedule != OMP_sch_static_chunked &&
-          Schedule != OMP_ord_static && Schedule != OMP_ord_static_chunked));
+          Schedule != OMP_ord_static && Schedule != OMP_ord_static_chunked &&
+          Schedule != OMP_sch_static_balanced_chunked));
   // Call __kmpc_dispatch_init(
   //          ident_t *loc, kmp_int32 tid, kmp_int32 schedule,
   //          kmp_int[32|64] lower, kmp_int[32|64] upper,
@@ -1824,59 +2468,94 @@ void CGOpenMPRuntime::emitForDispatchInit(CodeGenFunction &CGF,
   if (Chunk == nullptr)
     Chunk = CGF.Builder.getIntN(IVSize, 1);
   llvm::Value *Args[] = {
-    emitUpdateLocation(CGF, Loc, OMP_IDENT_KMPC),
-    getThreadID(CGF, Loc),
-    CGF.Builder.getInt32(Schedule), // Schedule type
-    CGF.Builder.getIntN(IVSize, 0), // Lower
-    UB,                             // Upper
-    CGF.Builder.getIntN(IVSize, 1), // Stride
-    Chunk                           // Chunk
+      emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
+      CGF.Builder.getInt32(addMonoNonMonoModifier(
+          Schedule, ScheduleKind.M1, ScheduleKind.M2)), // Schedule type
+      CGF.Builder.getIntN(IVSize, 0),                   // Lower
+      UB,                                               // Upper
+      CGF.Builder.getIntN(IVSize, 1),                   // Stride
+      Chunk                                             // Chunk
   };
   CGF.EmitRuntimeCall(createDispatchInitFunction(IVSize, IVSigned), Args);
 }
 
+static void emitForStaticInitCall(
+    CodeGenFunction &CGF, llvm::Value *UpdateLocation, llvm::Value *ThreadId,
+    llvm::Constant *ForStaticInitFunction, OpenMPSchedType Schedule,
+    OpenMPScheduleClauseModifier M1, OpenMPScheduleClauseModifier M2,
+    unsigned IVSize, bool Ordered, Address IL, Address LB, Address UB,
+    Address ST, llvm::Value *Chunk) {
+  if (!CGF.HaveInsertPoint())
+     return;
+
+   assert(!Ordered);
+   assert(Schedule == OMP_sch_static || Schedule == OMP_sch_static_chunked ||
+          Schedule == OMP_sch_static_balanced_chunked ||
+          Schedule == OMP_ord_static || Schedule == OMP_ord_static_chunked ||
+          Schedule == OMP_dist_sch_static ||
+          Schedule == OMP_dist_sch_static_chunked);
+
+   // Call __kmpc_for_static_init(
+   //          ident_t *loc, kmp_int32 tid, kmp_int32 schedtype,
+   //          kmp_int32 *p_lastiter, kmp_int[32|64] *p_lower,
+   //          kmp_int[32|64] *p_upper, kmp_int[32|64] *p_stride,
+   //          kmp_int[32|64] incr, kmp_int[32|64] chunk);
+   if (Chunk == nullptr) {
+     assert((Schedule == OMP_sch_static || Schedule == OMP_ord_static ||
+             Schedule == OMP_dist_sch_static) &&
+            "expected static non-chunked schedule");
+     // If the Chunk was not specified in the clause - use default value 1.
+       Chunk = CGF.Builder.getIntN(IVSize, 1);
+   } else {
+     assert((Schedule == OMP_sch_static_chunked ||
+             Schedule == OMP_sch_static_balanced_chunked ||
+             Schedule == OMP_ord_static_chunked ||
+             Schedule == OMP_dist_sch_static_chunked) &&
+            "expected static chunked schedule");
+   }
+   llvm::Value *Args[] = {
+       UpdateLocation, ThreadId, CGF.Builder.getInt32(addMonoNonMonoModifier(
+                                     Schedule, M1, M2)), // Schedule type
+       IL.getPointer(),                                  // &isLastIter
+       LB.getPointer(),                                  // &LB
+       UB.getPointer(),                                  // &UB
+       ST.getPointer(),                                  // &Stride
+       CGF.Builder.getIntN(IVSize, 1),                   // Incr
+       Chunk                                             // Chunk
+   };
+   CGF.EmitRuntimeCall(ForStaticInitFunction, Args);
+}
+
 void CGOpenMPRuntime::emitForStaticInit(CodeGenFunction &CGF,
                                         SourceLocation Loc,
-                                        OpenMPScheduleClauseKind ScheduleKind,
+                                        const OpenMPScheduleTy &ScheduleKind,
                                         unsigned IVSize, bool IVSigned,
                                         bool Ordered, Address IL, Address LB,
                                         Address UB, Address ST,
                                         llvm::Value *Chunk) {
-  if (!CGF.HaveInsertPoint())
-    return;
-  OpenMPSchedType Schedule =
-    getRuntimeSchedule(ScheduleKind, Chunk != nullptr, Ordered);
-  assert(!Ordered);
-  assert(Schedule == OMP_sch_static || Schedule == OMP_sch_static_chunked ||
-         Schedule == OMP_ord_static || Schedule == OMP_ord_static_chunked);
-
-  // Call __kmpc_for_static_init(
-  //          ident_t *loc, kmp_int32 tid, kmp_int32 schedtype,
-  //          kmp_int32 *p_lastiter, kmp_int[32|64] *p_lower,
-  //          kmp_int[32|64] *p_upper, kmp_int[32|64] *p_stride,
-  //          kmp_int[32|64] incr, kmp_int[32|64] chunk);
-  if (Chunk == nullptr) {
-    assert((Schedule == OMP_sch_static || Schedule == OMP_ord_static) &&
-           "expected static non-chunked schedule");
-    // If the Chunk was not specified in the clause - use default value 1.
-      Chunk = CGF.Builder.getIntN(IVSize, 1);
-  } else {
-    assert((Schedule == OMP_sch_static_chunked ||
-            Schedule == OMP_ord_static_chunked) &&
-           "expected static chunked schedule");
-  }
-  llvm::Value *Args[] = {
-    emitUpdateLocation(CGF, Loc, OMP_IDENT_KMPC),
-    getThreadID(CGF, Loc),
-    CGF.Builder.getInt32(Schedule), // Schedule type
-    IL.getPointer(),                // &isLastIter
-    LB.getPointer(),                // &LB
-    UB.getPointer(),                // &UB
-    ST.getPointer(),                // &Stride
-    CGF.Builder.getIntN(IVSize, 1), // Incr
-    Chunk                           // Chunk
-  };
-  CGF.EmitRuntimeCall(createForStaticInitFunction(IVSize, IVSigned), Args);
+  OpenMPSchedType ScheduleNum =
+      getRuntimeSchedule(ScheduleKind.Schedule, Chunk != nullptr, Ordered);
+  auto *UpdatedLocation = emitUpdateLocation(CGF, Loc);
+  auto *ThreadId = getThreadID(CGF, Loc);
+  auto *StaticInitFunction = createForStaticInitFunction(IVSize, IVSigned);
+  emitForStaticInitCall(CGF, UpdatedLocation, ThreadId, StaticInitFunction,
+                        ScheduleNum, ScheduleKind.M1, ScheduleKind.M2, IVSize,
+                        Ordered, IL, LB, UB, ST, Chunk);
+}
+
+void CGOpenMPRuntime::emitDistributeStaticInit(
+    CodeGenFunction &CGF, SourceLocation Loc,
+    OpenMPDistScheduleClauseKind SchedKind, unsigned IVSize, bool IVSigned,
+    bool Ordered, Address IL, Address LB, Address UB, Address ST,
+    llvm::Value *Chunk) {
+  OpenMPSchedType ScheduleNum = getRuntimeSchedule(SchedKind, Chunk != nullptr);
+  auto *UpdatedLocation = emitUpdateLocation(CGF, Loc);
+  auto *ThreadId = getThreadID(CGF, Loc);
+  auto *StaticInitFunction = createForStaticInitFunction(IVSize, IVSigned);
+  emitForStaticInitCall(CGF, UpdatedLocation, ThreadId, StaticInitFunction,
+                        ScheduleNum, OMPC_SCHEDULE_MODIFIER_unknown,
+                        OMPC_SCHEDULE_MODIFIER_unknown, IVSize, Ordered, IL, LB,
+                        UB, ST, Chunk);
 }
 
 void CGOpenMPRuntime::emitForStaticFinish(CodeGenFunction &CGF,
@@ -1884,8 +2563,7 @@ void CGOpenMPRuntime::emitForStaticFinish(CodeGenFunction &CGF,
   if (!CGF.HaveInsertPoint())
     return;
   // Call __kmpc_for_static_fini(ident_t *loc, kmp_int32 tid);
-  llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc, OMP_IDENT_KMPC),
-                         getThreadID(CGF, Loc)};
+  llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
   CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_for_static_fini),
                       Args);
 }
@@ -1897,8 +2575,7 @@ void CGOpenMPRuntime::emitForOrderedIterationEnd(CodeGenFunction &CGF,
   if (!CGF.HaveInsertPoint())
     return;
   // Call __kmpc_for_dynamic_fini_(4|8)[u](ident_t *loc, kmp_int32 tid);
-  llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc, OMP_IDENT_KMPC),
-                         getThreadID(CGF, Loc)};
+  llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
   CGF.EmitRuntimeCall(createDispatchFiniFunction(IVSize, IVSigned), Args);
 }
 
@@ -1912,7 +2589,8 @@ llvm::Value *CGOpenMPRuntime::emitForNext(CodeGenFunction &CGF,
   //          kmp_int[32|64] *p_lower, kmp_int[32|64] *p_upper,
   //          kmp_int[32|64] *p_stride);
   llvm::Value *Args[] = {
-      emitUpdateLocation(CGF, Loc, OMP_IDENT_KMPC), getThreadID(CGF, Loc),
+      emitUpdateLocation(CGF, Loc),
+      getThreadID(CGF, Loc),
       IL.getPointer(), // &isLastIter
       LB.getPointer(), // &Lower
       UB.getPointer(), // &Upper
@@ -1991,8 +2669,18 @@ enum KmpTaskTFields {
   KmpTaskTRoutine,
   /// \brief Partition id for the untied tasks.
   KmpTaskTPartId,
-  /// \brief Function with call of destructors for private variables.
-  KmpTaskTDestructors,
+  /// Function with call of destructors for private variables.
+  Data1,
+  /// Task priority.
+  Data2,
+  /// (Taskloops only) Lower bound.
+  KmpTaskTLowerBound,
+  /// (Taskloops only) Upper bound.
+  KmpTaskTUpperBound,
+  /// (Taskloops only) Stride.
+  KmpTaskTStride,
+  /// (Taskloops only) Is last iteration flag.
+  KmpTaskTLastIter,
 };
 } // anonymous namespace
 
@@ -2005,11 +2693,11 @@ bool CGOpenMPRuntime::OffloadEntriesInfoManagerTy::empty() const {
 void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
     initializeTargetRegionEntryInfo(unsigned DeviceID, unsigned FileID,
                                     StringRef ParentName, unsigned LineNum,
-                                    unsigned ColNum, unsigned Order) {
+                                    unsigned Order) {
   assert(CGM.getLangOpts().OpenMPIsDevice && "Initialization of entries is "
                                              "only required for the device "
                                              "code generation.");
-  OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum][ColNum] =
+  OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum] =
       OffloadEntryInfoTargetRegion(Order, /*Addr=*/nullptr, /*ID=*/nullptr);
   ++OffloadingEntriesNum;
 }
@@ -2017,30 +2705,27 @@ void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
 void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
     registerTargetRegionEntryInfo(unsigned DeviceID, unsigned FileID,
                                   StringRef ParentName, unsigned LineNum,
-                                  unsigned ColNum, llvm::Constant *Addr,
-                                  llvm::Constant *ID) {
+                                  llvm::Constant *Addr, llvm::Constant *ID) {
   // If we are emitting code for a target, the entry is already initialized,
   // only has to be registered.
   if (CGM.getLangOpts().OpenMPIsDevice) {
-    assert(hasTargetRegionEntryInfo(DeviceID, FileID, ParentName, LineNum,
-                                    ColNum) &&
+    assert(hasTargetRegionEntryInfo(DeviceID, FileID, ParentName, LineNum) &&
            "Entry must exist.");
-    auto &Entry = OffloadEntriesTargetRegion[DeviceID][FileID][ParentName]
-                                            [LineNum][ColNum];
+    auto &Entry =
+        OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum];
     assert(Entry.isValid() && "Entry not initialized!");
     Entry.setAddress(Addr);
     Entry.setID(ID);
     return;
   } else {
     OffloadEntryInfoTargetRegion Entry(OffloadingEntriesNum++, Addr, ID);
-    OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum][ColNum] =
-        Entry;
+    OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum] = Entry;
   }
 }
 
 bool CGOpenMPRuntime::OffloadEntriesInfoManagerTy::hasTargetRegionEntryInfo(
-    unsigned DeviceID, unsigned FileID, StringRef ParentName, unsigned LineNum,
-    unsigned ColNum) const {
+    unsigned DeviceID, unsigned FileID, StringRef ParentName,
+    unsigned LineNum) const {
   auto PerDevice = OffloadEntriesTargetRegion.find(DeviceID);
   if (PerDevice == OffloadEntriesTargetRegion.end())
     return false;
@@ -2053,11 +2738,8 @@ bool CGOpenMPRuntime::OffloadEntriesInfoManagerTy::hasTargetRegionEntryInfo(
   auto PerLine = PerParentName->second.find(LineNum);
   if (PerLine == PerParentName->second.end())
     return false;
-  auto PerColumn = PerLine->second.find(ColNum);
-  if (PerColumn == PerLine->second.end())
-    return false;
   // Fail if this entry is already registered.
-  if (PerColumn->second.getAddress() || PerColumn->second.getID())
+  if (PerLine->second.getAddress() || PerLine->second.getID())
     return false;
   return true;
 }
@@ -2069,8 +2751,7 @@ void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::actOnTargetRegionEntriesInfo(
     for (auto &F : D.second)
       for (auto &P : F.second)
         for (auto &L : P.second)
-          for (auto &C : L.second)
-            Action(D.first, F.first, P.first(), L.first, C.first, C.second);
+          Action(D.first, F.first, P.first(), L.first, L.second);
 }
 
 /// \brief Create a Ctor/Dtor-like function whose body is emitted through
@@ -2087,9 +2768,7 @@ createOffloadingBinaryDescriptorFunction(CodeGenModule &CGM, StringRef Name,
 
   CodeGenFunction CGF(CGM);
   GlobalDecl();
-  auto &FI = CGM.getTypes().arrangeFreeFunctionDeclaration(
-      C.VoidTy, Args, FunctionType::ExtInfo(),
-      /*isVariadic=*/false);
+  auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
   auto FTy = CGM.getTypes().GetFunctionType(FI);
   auto *Fn =
       CGM.CreateGlobalInitOrDestructFunction(FTy, Name, FI, SourceLocation());
@@ -2123,11 +2802,11 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
       CGM.getTypes().ConvertTypeForMem(getTgtOffloadEntryQTy());
   llvm::GlobalVariable *HostEntriesBegin = new llvm::GlobalVariable(
       M, OffloadEntryTy, /*isConstant=*/true,
-      llvm::GlobalValue::ExternalLinkage, /*Initializer=*/0,
+      llvm::GlobalValue::ExternalLinkage, /*Initializer=*/nullptr,
       ".omp_offloading.entries_begin");
   llvm::GlobalVariable *HostEntriesEnd = new llvm::GlobalVariable(
       M, OffloadEntryTy, /*isConstant=*/true,
-      llvm::GlobalValue::ExternalLinkage, /*Initializer=*/0,
+      llvm::GlobalValue::ExternalLinkage, /*Initializer=*/nullptr,
       ".omp_offloading.entries_end");
 
   // Create all device images
@@ -2139,10 +2818,11 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
     StringRef T = Devices[i].getTriple();
     auto *ImgBegin = new llvm::GlobalVariable(
         M, CGM.Int8Ty, /*isConstant=*/true, llvm::GlobalValue::ExternalLinkage,
-        /*Initializer=*/0, Twine(".omp_offloading.img_start.") + Twine(T));
+        /*Initializer=*/nullptr,
+        Twine(".omp_offloading.img_start.") + Twine(T));
     auto *ImgEnd = new llvm::GlobalVariable(
         M, CGM.Int8Ty, /*isConstant=*/true, llvm::GlobalValue::ExternalLinkage,
-        /*Initializer=*/0, Twine(".omp_offloading.img_end.") + Twine(T));
+        /*Initializer=*/nullptr, Twine(".omp_offloading.img_end.") + Twine(T));
 
     llvm::Constant *Dev =
         llvm::ConstantStruct::get(DeviceImageTy, ImgBegin, ImgEnd,
@@ -2160,7 +2840,7 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
       M, DeviceImagesInitTy, /*isConstant=*/true,
       llvm::GlobalValue::InternalLinkage, DeviceImagesInit,
       ".omp_offloading.device_images");
-  DeviceImages->setUnnamedAddr(true);
+  DeviceImages->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
 
   // This is a Zero array to be used in the creation of the constant expressions
   llvm::Constant *Index[] = {llvm::Constant::getNullValue(CGM.Int32Ty),
@@ -2190,12 +2870,14 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
                                 IdentInfo, C.CharTy);
 
   auto *UnRegFn = createOffloadingBinaryDescriptorFunction(
-      CGM, ".omp_offloading.descriptor_unreg", [&](CodeGenFunction &CGF) {
+      CGM, ".omp_offloading.descriptor_unreg",
+      [&](CodeGenFunction &CGF, PrePostActionTy &) {
         CGF.EmitCallOrInvoke(createRuntimeFunction(OMPRTL__tgt_unregister_lib),
                              Desc);
       });
   auto *RegFn = createOffloadingBinaryDescriptorFunction(
-      CGM, ".omp_offloading.descriptor_reg", [&](CodeGenFunction &CGF) {
+      CGM, ".omp_offloading.descriptor_reg",
+      [&](CodeGenFunction &CGF, PrePostActionTy &) {
         CGF.EmitCallOrInvoke(createRuntimeFunction(OMPRTL__tgt_register_lib),
                              Desc);
         CGM.getCXXABI().registerGlobalDtor(CGF, RegUnregVar, UnRegFn, Desc);
@@ -2203,15 +2885,16 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
   return RegFn;
 }
 
-void CGOpenMPRuntime::createOffloadEntry(llvm::Constant *Addr, StringRef Name,
-                                         uint64_t Size) {
+void CGOpenMPRuntime::createOffloadEntry(llvm::Constant *ID,
+                                         llvm::Constant *Addr, uint64_t Size) {
+  StringRef Name = Addr->getName();
   auto *TgtOffloadEntryType = cast<llvm::StructType>(
       CGM.getTypes().ConvertTypeForMem(getTgtOffloadEntryQTy()));
   llvm::LLVMContext &C = CGM.getModule().getContext();
   llvm::Module &M = CGM.getModule();