diff --git a/clang/lib/AST/Interp/Interp.h b/clang/lib/AST/Interp/Interp.h
@@ -572,7 +572,8 @@ bool IncDecHelper(InterpState &S, CodePtr OpPC, const Pointer &Ptr) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool Inc(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -585,7 +586,8 @@ bool Inc(InterpState &S, CodePtr OpPC) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool IncPop(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -599,7 +601,8 @@ bool IncPop(InterpState &S, CodePtr OpPC) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool Dec(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -612,7 +615,8 @@ bool Dec(InterpState &S, CodePtr OpPC) {
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool DecPop(InterpState &S, CodePtr OpPC) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -641,7 +645,8 @@ bool IncDecFloatHelper(InterpState &S, CodePtr OpPC, const Pointer &Ptr,
 
 inline bool Incf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -650,7 +655,8 @@ inline bool Incf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
 
 inline bool IncfPop(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
-
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Increment))
     return false;
 
@@ -660,6 +666,9 @@ inline bool IncfPop(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
 inline bool Decf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
 
+  if (Ptr.isDummy())
+    return false;
+
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -669,6 +678,8 @@ inline bool Decf(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
 inline bool DecfPop(InterpState &S, CodePtr OpPC, llvm::RoundingMode RM) {
   const Pointer &Ptr = S.Stk.pop<Pointer>();
 
+  if (Ptr.isDummy())
+    return false;
   if (!CheckInitialized(S, OpPC, Ptr, AK_Decrement))
     return false;
 
@@ -774,9 +785,9 @@ inline bool CmpHelperEQ<Pointer>(InterpState &S, CodePtr OpPC, CompareFn Fn) {
     // element in the same array are NOT equal. They have the same Base value,
     // but a different Offset. This is a pretty rare case, so we fix this here
     // by comparing pointers to the first elements.
-    if (LHS.isArrayRoot())
+    if (!LHS.isDummy() && LHS.isArrayRoot())
       VL = LHS.atIndex(0).getByteOffset();
-    if (RHS.isArrayRoot())
+    if (!RHS.isDummy() && RHS.isArrayRoot())
       VR = RHS.atIndex(0).getByteOffset();
 
     S.Stk.push<BoolT>(BoolT::from(Fn(Compare(VL, VR))));
@@ -1895,7 +1906,7 @@ inline bool ArrayElemPtr(InterpState &S, CodePtr OpPC) {
   const T &Offset = S.Stk.pop<T>();
   const Pointer &Ptr = S.Stk.peek<Pointer>();
 
-  if (!CheckDummy(S, OpPC, Ptr))
+  if (Ptr.isDummy())
     return true;
 
   if (!OffsetHelper<T, ArithOp::Add>(S, OpPC, Offset, Ptr))
@@ -1909,7 +1920,7 @@ inline bool ArrayElemPtrPop(InterpState &S, CodePtr OpPC) {
   const T &Offset = S.Stk.pop<T>();
   const Pointer &Ptr = S.Stk.pop<Pointer>();
 
-  if (!CheckDummy(S, OpPC, Ptr)) {
+  if (Ptr.isDummy()) {
     S.Stk.push<Pointer>(Ptr);
     return true;
   }

diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp
@@ -186,6 +186,14 @@ class EmitAssemblyHelper {
            TargetTriple.getVendor() != llvm::Triple::Apple;
   }
 
+  /// Check whether we should emit a flag for UnifiedLTO.
+  /// The UnifiedLTO module flag should be set when UnifiedLTO is enabled for
+  /// ThinLTO or Full LTO with module summaries.
+  bool shouldEmitUnifiedLTOModueFlag() const {
+    return CodeGenOpts.UnifiedLTO &&
+           (CodeGenOpts.PrepareForThinLTO || shouldEmitRegularLTOSummary());
+  }
+
 public:
   EmitAssemblyHelper(DiagnosticsEngine &_Diags,
                      const HeaderSearchOptions &HeaderSearchOpts,
@@ -1036,7 +1044,8 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
   if (!actionRequiresCodeGen(Action) && CodeGenOpts.VerifyModule)
     MPM.addPass(VerifierPass());
 
-  if (Action == Backend_EmitBC || Action == Backend_EmitLL) {
+  if (Action == Backend_EmitBC || Action == Backend_EmitLL ||
+      CodeGenOpts.FatLTO) {
     if (CodeGenOpts.PrepareForThinLTO && !CodeGenOpts.DisableLLVMPasses) {
       if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
         TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
@@ -1047,11 +1056,9 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
           if (!ThinLinkOS)
             return;
         }
-        if (CodeGenOpts.UnifiedLTO)
-          TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
         MPM.addPass(ThinLTOBitcodeWriterPass(
             *OS, ThinLinkOS ? &ThinLinkOS->os() : nullptr));
-      } else {
+      } else if (Action == Backend_EmitLL) {
         MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
                                     /*EmitLTOSummary=*/true));
       }
@@ -1065,24 +1072,17 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
         if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
           TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
                                    uint32_t(1));
-        if (CodeGenOpts.UnifiedLTO)
-          TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
       }
-      if (Action == Backend_EmitBC)
+      if (Action == Backend_EmitBC) {
         MPM.addPass(BitcodeWriterPass(*OS, CodeGenOpts.EmitLLVMUseLists,
                                       EmitLTOSummary));
-      else
+      } else if (Action == Backend_EmitLL) {
         MPM.addPass(PrintModulePass(*OS, "", CodeGenOpts.EmitLLVMUseLists,
                                     EmitLTOSummary));
+      }
     }
-  }
-  if (CodeGenOpts.FatLTO) {
-    // Set the EnableSplitLTOUnit and UnifiedLTO module flags, since FatLTO
-    // uses a different action than Backend_EmitBC or Backend_EmitLL.
-    if (!TheModule->getModuleFlag("EnableSplitLTOUnit"))
-      TheModule->addModuleFlag(llvm::Module::Error, "EnableSplitLTOUnit",
-                               uint32_t(CodeGenOpts.EnableSplitLTOUnit));
-    if (CodeGenOpts.UnifiedLTO && !TheModule->getModuleFlag("UnifiedLTO"))
+
+    if (shouldEmitUnifiedLTOModueFlag())
       TheModule->addModuleFlag(llvm::Module::Error, "UnifiedLTO", uint32_t(1));
   }
 

diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -7023,19 +7023,25 @@ void CodeGenFunction::EmitOMPInteropDirective(const OMPInteropDirective &S) {
                                      S.getSingleClause<OMPUseClause>())) &&
          "OMPNowaitClause clause is used separately in OMPInteropDirective.");
 
-  if (const auto *C = S.getSingleClause<OMPInitClause>()) {
-    llvm::Value *InteropvarPtr =
-        EmitLValue(C->getInteropVar()).getPointer(*this);
-    llvm::omp::OMPInteropType InteropType = llvm::omp::OMPInteropType::Unknown;
-    if (C->getIsTarget()) {
-      InteropType = llvm::omp::OMPInteropType::Target;
-    } else {
-      assert(C->getIsTargetSync() && "Expected interop-type target/targetsync");
-      InteropType = llvm::omp::OMPInteropType::TargetSync;
+  auto ItOMPInitClause = S.getClausesOfKind<OMPInitClause>();
+  if (!ItOMPInitClause.empty()) {
+    // Look at the multiple init clauses
+    for (const OMPInitClause *C : ItOMPInitClause) {
+      llvm::Value *InteropvarPtr =
+          EmitLValue(C->getInteropVar()).getPointer(*this);
+      llvm::omp::OMPInteropType InteropType =
+          llvm::omp::OMPInteropType::Unknown;
+      if (C->getIsTarget()) {
+        InteropType = llvm::omp::OMPInteropType::Target;
+      } else {
+        assert(C->getIsTargetSync() &&
+               "Expected interop-type target/targetsync");
+        InteropType = llvm::omp::OMPInteropType::TargetSync;
+      }
+      OMPBuilder.createOMPInteropInit(Builder, InteropvarPtr, InteropType,
+                                      Device, NumDependences, DependenceList,
+                                      Data.HasNowaitClause);
     }
-    OMPBuilder.createOMPInteropInit(Builder, InteropvarPtr, InteropType, Device,
-                                    NumDependences, DependenceList,
-                                    Data.HasNowaitClause);
   } else if (const auto *C = S.getSingleClause<OMPDestroyClause>()) {
     llvm::Value *InteropvarPtr =
         EmitLValue(C->getInteropVar()).getPointer(*this);

diff --git a/clang/lib/CodeGen/Targets/AArch64.cpp b/clang/lib/CodeGen/Targets/AArch64.cpp
@@ -132,8 +132,7 @@ class AArch64TargetCodeGenInfo : public TargetCodeGenInfo {
     assert(Error.empty());
 
     auto *Fn = cast<llvm::Function>(GV);
-    static const char *SignReturnAddrStr[] = {"none", "non-leaf", "all"};
-    Fn->addFnAttr("sign-return-address", SignReturnAddrStr[static_cast<int>(BPI.SignReturnAddr)]);
+    Fn->addFnAttr("sign-return-address", BPI.getSignReturnAddrStr());
 
     if (BPI.SignReturnAddr != LangOptions::SignReturnAddressScopeKind::None) {
       Fn->addFnAttr("sign-return-address-key",

diff --git a/clang/lib/CodeGen/Targets/ARM.cpp b/clang/lib/CodeGen/Targets/ARM.cpp
@@ -152,13 +152,7 @@ class ARMTargetCodeGenInfo : public TargetCodeGenInfo {
               diag::warn_target_unsupported_branch_protection_attribute)
               << Arch;
         } else {
-          static const char *SignReturnAddrStr[] = {"none", "non-leaf", "all"};
-          assert(static_cast<unsigned>(BPI.SignReturnAddr) <= 2 &&
-                 "Unexpected SignReturnAddressScopeKind");
-          Fn->addFnAttr(
-              "sign-return-address",
-              SignReturnAddrStr[static_cast<int>(BPI.SignReturnAddr)]);
-
+          Fn->addFnAttr("sign-return-address", BPI.getSignReturnAddrStr());
           Fn->addFnAttr("branch-target-enforcement",
                         BPI.BranchTargetEnforcement ? "true" : "false");
         }

diff --git a/clang/lib/Driver/ToolChains/CommonArgs.cpp b/clang/lib/Driver/ToolChains/CommonArgs.cpp
@@ -2763,13 +2763,13 @@ void tools::addOpenMPDeviceRTL(const Driver &D,
                                const llvm::opt::ArgList &DriverArgs,
                                llvm::opt::ArgStringList &CC1Args,
                                StringRef BitcodeSuffix,
-                               const llvm::Triple &Triple) {
+                               const llvm::Triple &Triple,
+                               const ToolChain &HostTC) {
   SmallVector<StringRef, 8> LibraryPaths;
 
-  // Add path to clang lib / lib64 folder.
-  SmallString<256> DefaultLibPath = llvm::sys::path::parent_path(D.Dir);
-  llvm::sys::path::append(DefaultLibPath, CLANG_INSTALL_LIBDIR_BASENAME);
-  LibraryPaths.emplace_back(DefaultLibPath.c_str());
+  // Check all of the standard library search paths used by the compiler.
+  for (const auto &LibPath : HostTC.getFilePaths())
+    LibraryPaths.emplace_back(LibPath);
 
   // Add user defined library paths from LIBRARY_PATH.
   std::optional<std::string> LibPath =

diff --git a/clang/lib/Driver/ToolChains/CommonArgs.h b/clang/lib/Driver/ToolChains/CommonArgs.h
@@ -214,7 +214,8 @@ void addMachineOutlinerArgs(const Driver &D, const llvm::opt::ArgList &Args,
 
 void addOpenMPDeviceRTL(const Driver &D, const llvm::opt::ArgList &DriverArgs,
                         llvm::opt::ArgStringList &CC1Args,
-                        StringRef BitcodeSuffix, const llvm::Triple &Triple);
+                        StringRef BitcodeSuffix, const llvm::Triple &Triple,
+                        const ToolChain &HostTC);
 
 void addOutlineAtomicsArgs(const Driver &D, const ToolChain &TC,
                            const llvm::opt::ArgList &Args,

diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -903,7 +903,7 @@ void CudaToolChain::addClangTargetOptions(
       return;
 
     addOpenMPDeviceRTL(getDriver(), DriverArgs, CC1Args, GpuArch.str(),
-                       getTriple());
+                       getTriple(), HostTC);
   }
 }
 

diff --git a/clang/lib/InstallAPI/Visitor.cpp b/clang/lib/InstallAPI/Visitor.cpp
@@ -19,7 +19,7 @@ using namespace llvm::MachO;
 
 namespace clang::installapi {
 
-// Exported NamedDecl needs to have externally visibiliy linkage and
+// Exported NamedDecl needs to have external linkage and
 // default visibility from LinkageComputer.
 static bool isExported(const NamedDecl *D) {
   auto LV = D->getLinkageAndVisibility();

diff --git a/clang/lib/Sema/JumpDiagnostics.cpp b/clang/lib/Sema/JumpDiagnostics.cpp
@@ -604,6 +604,16 @@ void JumpScopeChecker::BuildScopeInformation(Stmt *S,
     break;
   }
 
+  case Stmt::OpenACCComputeConstructClass: {
+    unsigned NewParentScope = Scopes.size();
+    OpenACCComputeConstruct *CC = cast<OpenACCComputeConstruct>(S);
+    Scopes.push_back(GotoScope(
+        ParentScope, diag::note_acc_branch_into_compute_construct,
+        diag::note_acc_branch_out_of_compute_construct, CC->getBeginLoc()));
+    BuildScopeInformation(CC->getStructuredBlock(), NewParentScope);
+    return;
+  }
+
   default:
     if (auto *ED = dyn_cast<OMPExecutableDirective>(S)) {
       if (!ED->isStandaloneDirective()) {
@@ -936,11 +946,16 @@ void JumpScopeChecker::CheckJump(Stmt *From, Stmt *To, SourceLocation DiagLoc,
       if (Scopes[I].InDiag == diag::note_protected_by_seh_finally) {
         S.Diag(From->getBeginLoc(), diag::warn_jump_out_of_seh_finally);
         break;
-      }
-      if (Scopes[I].InDiag == diag::note_omp_protected_structured_block) {
+      } else if (Scopes[I].InDiag ==
+                 diag::note_omp_protected_structured_block) {
         S.Diag(From->getBeginLoc(), diag::err_goto_into_protected_scope);
         S.Diag(To->getBeginLoc(), diag::note_omp_exits_structured_block);
         break;
+      } else if (Scopes[I].InDiag ==
+                 diag::note_acc_branch_into_compute_construct) {
+        S.Diag(From->getBeginLoc(), diag::err_goto_into_protected_scope);
+        S.Diag(Scopes[I].Loc, diag::note_acc_branch_out_of_compute_construct);
+        return;
       }
     }
   }

diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
@@ -2190,17 +2190,23 @@ static bool SemaBuiltinCpu(Sema &S, const TargetInfo &TI, CallExpr *TheCall,
 }
 
 /// Checks that __builtin_popcountg was called with a single argument, which is
-/// an integer.
+/// an unsigned integer.
 static bool SemaBuiltinPopcountg(Sema &S, CallExpr *TheCall) {
   if (checkArgCount(S, TheCall, 1))
     return true;
 
-  Expr *Arg = TheCall->getArg(0);
+  ExprResult ArgRes = S.DefaultLvalueConversion(TheCall->getArg(0));
+  if (ArgRes.isInvalid())
+    return true;
+
+  Expr *Arg = ArgRes.get();
+  TheCall->setArg(0, Arg);
+
   QualType ArgTy = Arg->getType();
 
-  if (!ArgTy->isIntegerType()) {
+  if (!ArgTy->isUnsignedIntegerType()) {
     S.Diag(Arg->getBeginLoc(), diag::err_builtin_invalid_arg_type)
-        << 1 << /*integer ty*/ 7 << ArgTy;
+        << 1 << /*unsigned integer ty*/ 7 << ArgTy;
     return true;
   }
   return false;
@@ -16538,6 +16544,17 @@ void Sema::DiagnoseAlwaysNonNullPointer(Expr *E,
     }
   }
 
+  // Complain if we are converting a lambda expression to a boolean value
+  if (const auto *MCallExpr = dyn_cast<CXXMemberCallExpr>(E)) {
+    if (const auto *MRecordDecl = MCallExpr->getRecordDecl();
+        MRecordDecl && MRecordDecl->isLambda()) {
+      Diag(E->getExprLoc(), diag::warn_impcast_pointer_to_bool)
+          << /*LambdaPointerConversionOperatorType=*/3
+          << MRecordDecl->getSourceRange() << Range << IsEqual;
+      return;
+    }
+  }
+
   // Expect to find a single Decl.  Skip anything more complicated.
   ValueDecl *D = nullptr;
   if (DeclRefExpr *R = dyn_cast<DeclRefExpr>(E)) {

diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
@@ -4962,7 +4962,8 @@ StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
           if (RC->getModifier() != OMPC_REDUCTION_inscan)
             continue;
           for (Expr *E : RC->copy_array_temps())
-            MarkDeclarationsReferencedInExpr(E);
+            if (E)
+              MarkDeclarationsReferencedInExpr(E);
         }
         if (auto *AC = dyn_cast<OMPAlignedClause>(C)) {
           for (Expr *E : AC->varlists())

diff --git a/clang/lib/Sema/SemaStmt.cpp b/clang/lib/Sema/SemaStmt.cpp
@@ -567,6 +567,11 @@ Sema::ActOnLabelStmt(SourceLocation IdentLoc, LabelDecl *TheDecl,
     Diag(IdentLoc, diag::warn_reserved_extern_symbol)
         << TheDecl << static_cast<int>(Status);
 
+  // If this label is in a compute construct scope, we need to make sure we
+  // check gotos in/out.
+  if (getCurScope()->isInOpenACCComputeConstructScope())
+    setFunctionHasBranchProtectedScope();
+
   // Otherwise, things are good.  Fill in the declaration and return it.
   LabelStmt *LS = new (Context) LabelStmt(IdentLoc, TheDecl, SubStmt);
   TheDecl->setStmt(LS);
@@ -3304,6 +3309,12 @@ StmtResult Sema::ActOnGotoStmt(SourceLocation GotoLoc,
                                SourceLocation LabelLoc,
                                LabelDecl *TheDecl) {
   setFunctionHasBranchIntoScope();
+
+  // If this goto is in a compute construct scope, we need to make sure we check
+  // gotos in/out.
+  if (getCurScope()->isInOpenACCComputeConstructScope())
+    setFunctionHasBranchProtectedScope();
+
   TheDecl->markUsed(Context);
   return new (Context) GotoStmt(TheDecl, GotoLoc, LabelLoc);
 }
@@ -3332,6 +3343,11 @@ Sema::ActOnIndirectGotoStmt(SourceLocation GotoLoc, SourceLocation StarLoc,
 
   setFunctionHasIndirectGoto();
 
+  // If this goto is in a compute construct scope, we need to make sure we
+  // check gotos in/out.
+  if (getCurScope()->isInOpenACCComputeConstructScope())
+    setFunctionHasBranchProtectedScope();
+
   return new (Context) IndirectGotoStmt(GotoLoc, StarLoc, E);
 }
 

diff --git a/clang/test/AST/Interp/arrays.cpp b/clang/test/AST/Interp/arrays.cpp
@@ -564,3 +564,8 @@ namespace LocalVLA {
 #endif
   }
 }
+
+char melchizedek[2200000000];
+typedef decltype(melchizedek[1] - melchizedek[0]) ptrdiff_t;
+constexpr ptrdiff_t d1 = &melchizedek[0x7fffffff] - &melchizedek[0]; // ok
+constexpr ptrdiff_t d3 = &melchizedek[0] - &melchizedek[0x80000000u]; // ok
diff --git a/clang/test/AST/Interp/c.c b/clang/test/AST/Interp/c.c
@@ -33,15 +33,15 @@ const int b = 3;
 _Static_assert(b == 3, ""); // pedantic-ref-warning {{not an integer constant expression}} \
                             // pedantic-expected-warning {{not an integer constant expression}}
 
-/// FIXME: The new interpreter is missing the "initializer of 'c' unknown" diagnostics.
-const int c; // ref-note {{declared here}} \
-             // pedantic-ref-note {{declared here}}
+const int c; // all-note {{declared here}}
 _Static_assert(c == 0, ""); // ref-error {{not an integral constant expression}} \
                             // ref-note {{initializer of 'c' is unknown}} \
                             // pedantic-ref-error {{not an integral constant expression}} \
                             // pedantic-ref-note {{initializer of 'c' is unknown}} \
                             // expected-error {{not an integral constant expression}} \
-                            // pedantic-expected-error {{not an integral constant expression}}
+                            // expected-note {{initializer of 'c' is unknown}} \
+                            // pedantic-expected-error {{not an integral constant expression}} \
+                            // pedantic-expected-note {{initializer of 'c' is unknown}}
 
 _Static_assert(&c != 0, ""); // ref-warning {{always true}} \
                              // pedantic-ref-warning {{always true}} \

diff --git a/clang/test/AST/Interp/cxx98.cpp b/clang/test/AST/Interp/cxx98.cpp
@@ -18,12 +18,13 @@ template struct C<cval>;
 
 /// FIXME: This example does not get properly diagnosed in the new interpreter.
 extern const int recurse1;
-const int recurse2 = recurse1; // ref-note {{here}}
+const int recurse2 = recurse1; // both-note {{declared here}}
 const int recurse1 = 1;
 int array1[recurse1];
 int array2[recurse2]; // ref-warning 2{{variable length array}} \
                       // ref-note {{initializer of 'recurse2' is not a constant expression}} \
                       // expected-warning {{variable length array}} \
+                      // expected-note {{read of non-const variable 'recurse2'}} \
                       // expected-error {{variable length array}}
 
 int NCI; // both-note {{declared here}}

diff --git a/clang/test/CXX/drs/dr18xx.cpp b/clang/test/CXX/drs/dr18xx.cpp
@@ -282,6 +282,7 @@ namespace dr1837 { // dr1837: 3.3
   struct A {
     int f();
     bool b = [] {
+      // since-cxx11-warning@-1 {{address of lambda function pointer conversion operator will always evaluate to 'true'}}
       struct Local {
         static_assert(sizeof(this->f()) == sizeof(int), "");
       };

diff --git a/clang/test/CXX/expr/expr.prim/expr.prim.lambda/blocks.mm b/clang/test/CXX/expr/expr.prim/expr.prim.lambda/blocks.mm
@@ -65,10 +65,10 @@ void nesting() {
 
 namespace overloading {
   void bool_conversion() {
-    if ([](){}) {
+    if ([](){}) { // expected-warning{{address of lambda function pointer conversion operator will always evaluate to 'true'}}
     }
 
-    bool b = []{};
+    bool b = []{}; // expected-warning{{address of lambda function pointer conversion operator will always evaluate to 'true'}}
     b = (bool)[]{};
   }
 
@@ -108,8 +108,9 @@ void call_with_lambda() {
     using decltype(a)::operator id<void(*)()>; // expected-note {{here}}
   } extern d;
 
-  bool r1 = c;
-  bool r2 = d; // expected-error {{private}}
+  bool r1 = c; // expected-warning{{address of lambda function pointer conversion operator will always evaluate to 'true'}}
+  bool r2 = d; // expected-error {{private}} \
+                  expected-warning{{address of lambda function pointer conversion operator will always evaluate to 'true'}}
 }
 
 namespace PR13117 {

diff --git a/clang/test/CodeGen/builtins.c b/clang/test/CodeGen/builtins.c
@@ -948,38 +948,38 @@ void test_builtin_popcountg(unsigned char uc, unsigned short us,
   volatile int pop;
   pop = __builtin_popcountg(uc);
   // CHECK: %1 = load i8, ptr %uc.addr, align 1
-  // CHECK-NEXT: %conv = zext i8 %1 to i32
-  // CHECK-NEXT: %2 = call i32 @llvm.ctpop.i32(i32 %conv)
-  // CHECK-NEXT: store volatile i32 %2, ptr %pop, align 4
+  // CHECK-NEXT: %2 = call i8 @llvm.ctpop.i8(i8 %1)
+  // CHECK-NEXT: %cast = sext i8 %2 to i32
+  // CHECK-NEXT: store volatile i32 %cast, ptr %pop, align 4
   pop = __builtin_popcountg(us);
   // CHECK-NEXT: %3 = load i16, ptr %us.addr, align 2
-  // CHECK-NEXT: %conv1 = zext i16 %3 to i32
-  // CHECK-NEXT: %4 = call i32 @llvm.ctpop.i32(i32 %conv1)
-  // CHECK-NEXT: store volatile i32 %4, ptr %pop, align 4
+  // CHECK-NEXT: %4 = call i16 @llvm.ctpop.i16(i16 %3)
+  // CHECK-NEXT: %cast1 = sext i16 %4 to i32
+  // CHECK-NEXT: store volatile i32 %cast1, ptr %pop, align 4
   pop = __builtin_popcountg(ui);
   // CHECK-NEXT: %5 = load i32, ptr %ui.addr, align 4
   // CHECK-NEXT: %6 = call i32 @llvm.ctpop.i32(i32 %5)
   // CHECK-NEXT: store volatile i32 %6, ptr %pop, align 4
   pop = __builtin_popcountg(ul);
   // CHECK-NEXT: %7 = load i64, ptr %ul.addr, align 8
   // CHECK-NEXT: %8 = call i64 @llvm.ctpop.i64(i64 %7)
-  // CHECK-NEXT: %cast = trunc i64 %8 to i32
-  // CHECK-NEXT: store volatile i32 %cast, ptr %pop, align 4
+  // CHECK-NEXT: %cast2 = trunc i64 %8 to i32
+  // CHECK-NEXT: store volatile i32 %cast2, ptr %pop, align 4
   pop = __builtin_popcountg(ull);
   // CHECK-NEXT: %9 = load i64, ptr %ull.addr, align 8
   // CHECK-NEXT: %10 = call i64 @llvm.ctpop.i64(i64 %9)
-  // CHECK-NEXT: %cast2 = trunc i64 %10 to i32
-  // CHECK-NEXT: store volatile i32 %cast2, ptr %pop, align 4
+  // CHECK-NEXT: %cast3 = trunc i64 %10 to i32
+  // CHECK-NEXT: store volatile i32 %cast3, ptr %pop, align 4
   pop = __builtin_popcountg(ui128);
   // CHECK-NEXT: %11 = load i128, ptr %ui128.addr, align 16
   // CHECK-NEXT: %12 = call i128 @llvm.ctpop.i128(i128 %11)
-  // CHECK-NEXT: %cast3 = trunc i128 %12 to i32
-  // CHECK-NEXT: store volatile i32 %cast3, ptr %pop, align 4
+  // CHECK-NEXT: %cast4 = trunc i128 %12 to i32
+  // CHECK-NEXT: store volatile i32 %cast4, ptr %pop, align 4
   pop = __builtin_popcountg(ubi128);
   // CHECK-NEXT: %13 = load i128, ptr %ubi128.addr, align 8
   // CHECK-NEXT: %14 = call i128 @llvm.ctpop.i128(i128 %13)
-  // CHECK-NEXT: %cast4 = trunc i128 %14 to i32
-  // CHECK-NEXT: store volatile i32 %cast4, ptr %pop, align 4
+  // CHECK-NEXT: %cast5 = trunc i128 %14 to i32
+  // CHECK-NEXT: store volatile i32 %cast5, ptr %pop, align 4
   // CHECK-NEXT: ret void
 }
 

diff --git a/clang/test/CodeGen/fat-lto-objects.c b/clang/test/CodeGen/fat-lto-objects.c
@@ -11,10 +11,11 @@
 // RUN: llvm-objcopy --dump-section=.llvm.lto=%t.full.split.bc %t.full.split.o
 // RUN: llvm-dis %t.full.split.bc -o - | FileCheck %s --check-prefixes=FULL,SPLIT,NOUNIFIED
 
+/// Full LTO always sets EnableSplitLTOUnit when the summary is used.
 // RUN: %clang -cc1 -triple x86_64-unknown-linux-gnu -flto=full -ffat-lto-objects -emit-obj < %s -o %t.full.nosplit.o
 // RUN: llvm-readelf -S %t.full.nosplit.o | FileCheck %s --check-prefixes=ELF
 // RUN: llvm-objcopy --dump-section=.llvm.lto=%t.full.nosplit.bc %t.full.nosplit.o
-// RUN: llvm-dis %t.full.nosplit.bc -o - | FileCheck %s --check-prefixes=FULL,NOSPLIT,NOUNIFIED
+// RUN: llvm-dis %t.full.nosplit.bc -o - | FileCheck %s --check-prefixes=FULL,SPLIT,NOUNIFIED
 
 // RUN: %clang -cc1 -triple x86_64-unknown-linux-gnu -flto=thin -fsplit-lto-unit -ffat-lto-objects -emit-obj < %s -o %t.thin.split.o
 // RUN: llvm-readelf -S %t.thin.split.o | FileCheck %s --check-prefixes=ELF
@@ -34,6 +35,21 @@
 // RUN: %clang -cc1 -triple x86_64-unknown-linux-gnu -flto=full -ffat-lto-objects -fsplit-lto-unit -S < %s -o - \
 // RUN: | FileCheck %s --check-prefixes=ASM
 
+/// Make sure that FatLTO generates .llvm.lto sections that are the same as the output from normal LTO compilations
+// RUN: %clang -O2 --target=x86_64-unknown-linux-gnu -fPIE -flto=full -ffat-lto-objects -c %s -o %t.fatlto.full.o
+// RUN: llvm-objcopy --dump-section=.llvm.lto=%t.fatlto.full.bc %t.fatlto.full.o
+// RUN: llvm-dis < %t.fatlto.full.bc -o %t.fatlto.full.ll
+// RUN: %clang -O2 --target=x86_64-unknown-linux-gnu -fPIE -flto=full -c %s -o %t.nofat.full.bc
+// RUN: llvm-dis < %t.nofat.full.bc -o %t.nofat.full.ll
+// RUN: diff %t.fatlto.full.ll %t.nofat.full.ll
+
+// RUN: %clang -O2 --target=x86_64-unknown-linux-gnu -fPIE -flto=thin -ffat-lto-objects -c %s -o %t.fatlto.thin.o
+// RUN: llvm-objcopy --dump-section=.llvm.lto=%t.fatlto.thin.bc %t.fatlto.thin.o
+// RUN: llvm-dis < %t.fatlto.thin.bc -o %t.fatlto.thin.ll
+// RUN: %clang -O2 --target=x86_64-unknown-linux-gnu -fPIE -flto=thin -c %s -o %t.nofat.thin.bc
+// RUN: llvm-dis < %t.nofat.thin.bc -o %t.nofat.thin.ll
+// RUN: diff %t.fatlto.thin.ll %t.nofat.thin.ll
+
 /// Be sure we enable split LTO units correctly under -ffat-lto-objects.
 //   SPLIT: ![[#]] = !{i32 1, !"EnableSplitLTOUnit", i32 1}
 // NOSPLIT: ![[#]] = !{i32 1, !"EnableSplitLTOUnit", i32 0}
@@ -51,6 +67,9 @@
 // ASM-NEXT:        .asciz  "BC
 // ASM-NEXT: .size   .Lllvm.embedded.object
 
+const char* foo = "foo";
+
 int test(void) {
+  const char* bar = "bar";
   return 0xabcd;
 }
diff --git a/clang/test/Driver/darwin-header-search-libcxx.cpp b/clang/test/Driver/darwin-header-search-libcxx.cpp
@@ -193,7 +193,7 @@
 // RUN: ln -sf %t/install/bin/clang %t/symlinked1/bin/clang
 // RUN: mkdir -p %t/symlinked1/include/c++/v1
 
-// RUN: %t/symlinked1/bin/clang -### %s -fsyntax-only 2>&1 \
+// RUN: %t/symlinked1/bin/clang -### %s -no-canonical-prefixes -fsyntax-only 2>&1 \
 // RUN:     --target=x86_64-apple-darwin \
 // RUN:     -stdlib=libc++ \
 // RUN:     -isysroot %S/Inputs/basic_darwin_sdk_usr_cxx_v1 \

diff --git a/clang/test/Driver/mingw-sysroot.cpp b/clang/test/Driver/mingw-sysroot.cpp
@@ -50,10 +50,12 @@
 // CHECK_TESTROOT_GCC_EXPLICIT: "-internal-isystem" "{{[^"]+}}/testroot-gcc{{/|\\\\}}include"
 
 
-// If there's a matching sysroot next to the clang binary itself, prefer that
+// If -no-canonical-prefixes and there's a matching sysroot next to the clang binary itself, prefer that
 // over a gcc in the path:
 
-// RUN: env "PATH=%T/testroot-gcc/bin:%PATH%" %T/testroot-clang/bin/x86_64-w64-mingw32-clang -target x86_64-w64-mingw32 -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CLANG %s
+// RUN: env "PATH=%T/testroot-gcc/bin:%PATH%" %T/testroot-clang/bin/x86_64-w64-mingw32-clang --target=x86_64-w64-mingw32 -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_GCC2 %s
+// RUN: env "PATH=%T/testroot-gcc/bin:%PATH%" %T/testroot-clang/bin/x86_64-w64-mingw32-clang --target=x86_64-w64-mingw32 -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s -no-canonical-prefixes 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CLANG %s
+// CHECK_TESTROOT_GCC2: "{{[^"]+}}/testroot-gcc{{/|\\\\}}x86_64-w64-mingw32{{/|\\\\}}include"
 // CHECK_TESTROOT_CLANG: "{{[^"]+}}/testroot-clang{{/|\\\\}}x86_64-w64-mingw32{{/|\\\\}}include"
 
 
@@ -82,7 +84,7 @@
 // that indicates that we did choose the right base, even if this particular directory
 // actually doesn't exist here.
 
-// RUN: env "PATH=%T/testroot-gcc/bin:%PATH%" %T/testroot-clang-native/bin/clang -target x86_64-w64-mingw32 -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CLANG_NATIVE %s
+// RUN: env "PATH=%T/testroot-gcc/bin:%PATH%" %T/testroot-clang-native/bin/clang -no-canonical-prefixes --target=x86_64-w64-mingw32 -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CLANG_NATIVE %s
 // CHECK_TESTROOT_CLANG_NATIVE: "{{[^"]+}}/testroot-clang-native{{/|\\\\}}x86_64-w64-mingw32{{/|\\\\}}include"
 
 
@@ -93,12 +95,12 @@
 // that defaults to x86_64 mingw, but it's easier to test this in cross setups
 // with symlinks, like the other tests here.)
 
-// RUN: env "PATH=%T/testroot-gcc/bin:%PATH%" %T/testroot-clang/bin/x86_64-w64-mingw32-clang --target=x86_64-w64-mingw32 -m32 -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CLANG_I686 %s
+// RUN: env "PATH=%T/testroot-gcc/bin:%PATH%" %T/testroot-clang/bin/x86_64-w64-mingw32-clang -no-canonical-prefixes --target=x86_64-w64-mingw32 -m32 -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CLANG_I686 %s
 // CHECK_TESTROOT_CLANG_I686: "{{[^"]+}}/testroot-clang{{/|\\\\}}i686-w64-mingw32{{/|\\\\}}include"
 
 
 // If the user calls clang with a custom literal triple, make sure this maps
 // to sysroots with the matching spelling.
 
-// RUN: %T/testroot-custom-triple/bin/clang --target=x86_64-w64-mingw32foo -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CUSTOM_TRIPLE %s
+// RUN: %T/testroot-custom-triple/bin/clang -no-canonical-prefixes --target=x86_64-w64-mingw32foo -rtlib=compiler-rt -stdlib=libstdc++ --sysroot="" -c -### %s 2>&1 | FileCheck -check-prefix=CHECK_TESTROOT_CUSTOM_TRIPLE %s
 // CHECK_TESTROOT_CUSTOM_TRIPLE: "{{[^"]+}}/testroot-custom-triple{{/|\\\\}}x86_64-w64-mingw32foo{{/|\\\\}}include"
diff --git a/clang/test/Driver/no-canonical-prefixes.c b/clang/test/Driver/no-canonical-prefixes.c
@@ -26,7 +26,7 @@
 // RUN:     | FileCheck --check-prefix=NON-CANONICAL %s
 //
 // FIXME: This should really be '.real'.
-// CANONICAL: InstalledDir: {{.*}}.fake
+// CANONICAL: InstalledDir: {{.*}}bin
 // CANONICAL: {{[/|\\]*}}clang{{.*}}" -cc1
 //
 // NON-CANONICAL: InstalledDir: .{{$}}

diff --git a/clang/test/Driver/program-path-priority.c b/clang/test/Driver/program-path-priority.c
@@ -36,7 +36,7 @@
 // RUN: touch %t/notreal-none-elf-gcc && chmod +x %t/notreal-none-elf-gcc
 // RUN: env "PATH=" %t/clang -### -target notreal-none-elf %s 2>&1 | \
 // RUN:   FileCheck --check-prefix=PROG_PATH_NOTREAL_GCC %s
-// PROG_PATH_NOTREAL_GCC: notreal-none-elf-gcc"
+// PROG_PATH_NOTREAL_GCC: notreal-none-unknown-elf
 
 /// <triple>-gcc on the PATH is found
 // RUN: mkdir -p %t/env
@@ -57,7 +57,7 @@
 // RUN: touch %t/gcc && chmod +x %t/gcc
 // RUN: env "PATH=" %t/clang -### -target notreal-none-elf %s 2>&1 | \
 // RUN:   FileCheck --check-prefix=NOTREAL_GCC_PREFERRED %s
-// NOTREAL_GCC_PREFERRED: notreal-none-elf-gcc"
+// NOTREAL_GCC_PREFERRED: notreal-none-unknown-elf"
 // NOTREAL_GCC_PREFERRED-NOT: /gcc"
 
 /// <triple>-gcc on the PATH is preferred to gcc in program path
@@ -125,6 +125,9 @@
 /// Only if there is nothing in the prefix will we search other paths
 /// -f in case $DEFAULT_TRIPLE == %target_triple
 // RUN: rm -f %t/prefix/$DEFAULT_TRIPLE-gcc %t/prefix/%target_triple-gcc %t/prefix/gcc
-// RUN: env "PATH=" %t/clang -### -target notreal-none-elf %s -B %t/prefix 2>&1 | \
-// RUN:   FileCheck --check-prefix=EMPTY_PREFIX_DIR %s
-// EMPTY_PREFIX_DIR: notreal-none-elf-gcc"
+// RUN: env "PATH=" %t/clang -### -canonical-prefixes --target=notreal-none-elf %s -B %t/prefix 2>&1 | \
+// RUN:   FileCheck --check-prefix=EMPTY_PREFIX_DIR1 %s
+// EMPTY_PREFIX_DIR1: gcc"
+// RUN: env "PATH=" %t/clang -### -no-canonical-prefixes --target=notreal-none-elf %s -B %t/prefix 2>&1 | \
+// RUN:   FileCheck --check-prefix=EMPTY_PREFIX_DIR2 %s
+// EMPTY_PREFIX_DIR2: notreal-none-elf-gcc"
diff --git a/clang/test/Driver/rocm-detect.hip b/clang/test/Driver/rocm-detect.hip
@@ -102,7 +102,7 @@
 // RUN: rm -rf %t/rocm-spack
 // RUN: cp -r %S/Inputs/rocm-spack %t
 // RUN: ln -fs %clang %t/rocm-spack/llvm-amdgpu-4.0.0-ieagcs7inf7runpyfvepqkurasoglq4z/bin/clang
-// RUN: %t/rocm-spack/llvm-amdgpu-4.0.0-ieagcs7inf7runpyfvepqkurasoglq4z/bin/clang -### -v \
+// RUN: %t/rocm-spack/llvm-amdgpu-4.0.0-ieagcs7inf7runpyfvepqkurasoglq4z/bin/clang -### -no-canonical-prefixes -v \
 // RUN:   -resource-dir=%t/rocm-spack/llvm-amdgpu-4.0.0-ieagcs7inf7runpyfvepqkurasoglq4z/lib/clang \
 // RUN:   -target x86_64-linux-gnu --cuda-gpu-arch=gfx900 --print-rocm-search-dirs %s 2>&1 \
 // RUN:   | FileCheck -check-prefixes=SPACK %s
@@ -111,7 +111,7 @@
 // ROCm release. --hip-path and --rocm-device-lib-path can be used to specify them.
 
 // RUN: cp -r %t/rocm-spack/hip-* %t/rocm-spack/hip-4.0.0-abcd
-// RUN: %t/rocm-spack/llvm-amdgpu-4.0.0-ieagcs7inf7runpyfvepqkurasoglq4z/bin/clang -### -v \
+// RUN: %t/rocm-spack/llvm-amdgpu-4.0.0-ieagcs7inf7runpyfvepqkurasoglq4z/bin/clang -### -no-canonical-prefixes -v \
 // RUN:   -target x86_64-linux-gnu --cuda-gpu-arch=gfx900 \
 // RUN:   --hip-path=%t/rocm-spack/hip-4.0.0-abcd \
 // RUN:    %s 2>&1 | FileCheck -check-prefixes=SPACK-SET %s

diff --git a/clang/test/OpenMP/interop_codegen.cpp b/clang/test/OpenMP/interop_codegen.cpp
@@ -0,0 +1,35 @@
+// expected-no-diagnostics
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=nvptx64-nvidia-cuda -emit-llvm %s -o - | FileCheck %s
+// RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s
+
+#ifndef HEADER
+#define HEADER
+
+typedef void *omp_interop_t;
+#define omp_interop_none 0
+#define omp_ipr_fr_id -1
+typedef long omp_intptr_t;
+#define NULL 0
+
+extern omp_intptr_t omp_get_interop_int(const omp_interop_t, int, int *);
+
+int main() {
+  omp_interop_t obj = omp_interop_none;
+  omp_interop_t i1 = omp_interop_none;
+  omp_interop_t i2 = omp_interop_none;
+  omp_interop_t i3 = omp_interop_none;
+  omp_interop_t i4 = omp_interop_none;
+  omp_interop_t i5 = omp_interop_none;
+
+  #pragma omp interop init(targetsync: i1) init(targetsync: obj)
+  int id = (int )omp_get_interop_int(obj, omp_ipr_fr_id, NULL);
+  int id1 = (int )omp_get_interop_int(i1, omp_ipr_fr_id, NULL);
+
+
+}
+#endif
+
+// CHECK-LABEL: define {{.+}}main{{.+}} 
+// CHECK: call {{.+}}__tgt_interop_init({{.+}}i1{{.*}})
+// CHECK: call {{.+}}__tgt_interop_init({{.+}}obj{{.*}})
diff --git a/clang/test/OpenMP/scan_ast_print.cpp b/clang/test/OpenMP/scan_ast_print.cpp
@@ -17,6 +17,10 @@ T tmain(T argc) {
   static T a;
 #pragma omp for reduction(inscan, +: a)
   for (int i = 0; i < 10; ++i) {
+#pragma omp scan inclusive(a)
+  }
+#pragma omp parallel for reduction(inscan, +:a)
+  for (int i = 0; i < 10; ++i) {
 #pragma omp scan inclusive(a)
   }
   return a + argc;
@@ -25,15 +29,29 @@ T tmain(T argc) {
 // CHECK-NEXT: #pragma omp for reduction(inscan, +: a)
 // CHECK-NEXT: for (int i = 0; i < 10; ++i) {
 // CHECK-NEXT: #pragma omp scan inclusive(a){{$}}
+
+// CHECK: #pragma omp parallel for reduction(inscan, +: a)
+// CHECK-NEXT: for (int i = 0; i < 10; ++i) {
+// CHECK-NEXT: #pragma omp scan inclusive(a){{$}}
+
 // CHECK:      static int a;
 // CHECK-NEXT: #pragma omp for reduction(inscan, +: a)
 // CHECK-NEXT: for (int i = 0; i < 10; ++i) {
 // CHECK-NEXT: #pragma omp scan inclusive(a)
+
+// CHECK: #pragma omp parallel for reduction(inscan, +: a)
+// CHECK-NEXT: for (int i = 0; i < 10; ++i) {
+// CHECK-NEXT: #pragma omp scan inclusive(a)
+
 // CHECK:      static char a;
 // CHECK-NEXT: #pragma omp for reduction(inscan, +: a)
 // CHECK-NEXT: for (int i = 0; i < 10; ++i) {
 // CHECK-NEXT: #pragma omp scan inclusive(a)
 
+// CHECK: #pragma omp parallel for reduction(inscan, +: a)
+// CHECK-NEXT: for (int i = 0; i < 10; ++i) {
+// CHECK-NEXT: #pragma omp scan inclusive(a)
+
 int main(int argc, char **argv) {
   static int a;
 // CHECK: static int a;

diff --git a/clang/test/Preprocessor/riscv-target-features.c b/clang/test/Preprocessor/riscv-target-features.c
@@ -74,6 +74,7 @@
 // CHECK-NOT: __riscv_xventanacondops {{.*$}}
 // CHECK-NOT: __riscv_za128rs {{.*$}}
 // CHECK-NOT: __riscv_za64rs {{.*$}}
+// CHECK-NOT: __riscv_zacas {{.*$}}
 // CHECK-NOT: __riscv_zawrs {{.*$}}
 // CHECK-NOT: __riscv_zba {{.*$}}
 // CHECK-NOT: __riscv_zbb {{.*$}}
@@ -166,7 +167,6 @@
 // CHECK-NOT: __riscv_ssqosid{{.*$}}
 // CHECK-NOT: __riscv_supm{{.*$}}
 // CHECK-NOT: __riscv_zaamo {{.*$}}
-// CHECK-NOT: __riscv_zacas {{.*$}}
 // CHECK-NOT: __riscv_zalasr {{.*$}}
 // CHECK-NOT: __riscv_zalrsc {{.*$}}
 // CHECK-NOT: __riscv_zcmop {{.*$}}
@@ -660,6 +660,14 @@
 // RUN:   -o - | FileCheck --check-prefix=CHECK-ZA64RS-EXT %s
 // CHECK-ZA64RS-EXT: __riscv_za64rs 1000000{{$}}
 
+// RUN: %clang --target=riscv32 \
+// RUN:   -march=rv32i_zacas1p0 -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-ZACAS-EXT %s
+// RUN: %clang --target=riscv64 \
+// RUN:   -march=rv64i_zacas1p0 -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-ZACAS-EXT %s
+// CHECK-ZACAS-EXT: __riscv_zacas 1000000{{$}}
+
 // RUN: %clang --target=riscv32-unknown-linux-gnu \
 // RUN:   -march=rv32izawrs -E -dM %s \
 // RUN:   -o - | FileCheck --check-prefix=CHECK-ZAWRS-EXT %s
@@ -1485,14 +1493,6 @@
 // RUN:   -o - | FileCheck --check-prefix=CHECK-ZAAMO-EXT %s
 // CHECK-ZAAMO-EXT: __riscv_zaamo 2000{{$}}
 
-// RUN: %clang --target=riscv32 -menable-experimental-extensions \
-// RUN:   -march=rv32i_zacas1p0 -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-ZACAS-EXT %s
-// RUN: %clang --target=riscv64 -menable-experimental-extensions \
-// RUN:   -march=rv64i_zacas1p0 -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-ZACAS-EXT %s
-// CHECK-ZACAS-EXT: __riscv_zacas 1000000{{$}}
-
 // RUN: %clang --target=riscv32 -menable-experimental-extensions \
 // RUN:   -march=rv32i_zalasr0p1 -E -dM %s \
 // RUN:   -o - | FileCheck --check-prefix=CHECK-ZALASR-EXT %s

diff --git a/clang/test/Sema/builtin-popcountg.c b/clang/test/Sema/builtin-popcountg.c
@@ -1,14 +1,23 @@
-// RUN: %clang_cc1 -triple=x86_64-pc-linux-gnu -fsyntax-only -verify -Wpedantic %s
+// RUN: %clang_cc1 -std=c23 -triple=x86_64-pc-linux-gnu -fsyntax-only -verify -Wpedantic %s
 
 typedef int int2 __attribute__((ext_vector_type(2)));
 
-void test_builtin_popcountg(int i, double d, int2 i2) {
+void test_builtin_popcountg(short s, int i, __int128 i128, _BitInt(128) bi128,
+                            double d, int2 i2) {
   __builtin_popcountg();
   // expected-error@-1 {{too few arguments to function call, expected 1, have 0}}
   __builtin_popcountg(i, i);
   // expected-error@-1 {{too many arguments to function call, expected 1, have 2}}
+  __builtin_popcountg(s);
+  // expected-error@-1 {{1st argument must be a type of unsigned integer (was 'short')}}
+  __builtin_popcountg(i);
+  // expected-error@-1 {{1st argument must be a type of unsigned integer (was 'int')}}
+  __builtin_popcountg(i128);
+  // expected-error@-1 {{1st argument must be a type of unsigned integer (was '__int128')}}
+  __builtin_popcountg(bi128);
+  // expected-error@-1 {{1st argument must be a type of unsigned integer (was '_BitInt(128)')}}
   __builtin_popcountg(d);
-  // expected-error@-1 {{1st argument must be a type of integer (was 'double')}}
+  // expected-error@-1 {{1st argument must be a type of unsigned integer (was 'double')}}
   __builtin_popcountg(i2);
-  // expected-error@-1 {{1st argument must be a type of integer (was 'int2' (vector of 2 'int' values))}}
+  // expected-error@-1 {{1st argument must be a type of unsigned integer (was 'int2' (vector of 2 'int' values))}}
 }
diff --git a/clang/test/SemaCXX/warn-bool-conversion.cpp b/clang/test/SemaCXX/warn-bool-conversion.cpp
@@ -81,6 +81,18 @@ struct S2 {
 
 bool f5();
 bool f6(int);
+#if __cplusplus >= 201103L
+auto f7 = []{};
+auto f8 = [](){};
+
+void foo() {
+  bool b;
+  b = f7; // expected-warning {{address of lambda function pointer conversion operator will always evaluate to 'true'}}
+  b = f8; // expected-warning {{address of lambda function pointer conversion operator will always evaluate to 'true'}}
+  bool is_true = [](){ return true; };
+  // expected-warning@-1{{address of lambda function pointer conversion operator will always evaluate to 'true'}}
+}
+#endif
 
 void bar() {
   bool b;

diff --git a/clang/test/SemaOpenACC/no-branch-in-out.c b/clang/test/SemaOpenACC/no-branch-in-out.c
@@ -113,3 +113,200 @@ void Return() {
     }
   }
 }
+
+void Goto() {
+  int j;
+#pragma acc parallel // expected-note{{invalid branch out of OpenACC Compute Construct}}
+  while(j) {
+    if (j <3)
+      goto LABEL; // expected-error{{cannot jump from this goto statement to its label}}
+  }
+
+LABEL:
+  {}
+
+  goto LABEL_IN; // expected-error{{cannot jump from this goto statement to its label}}
+
+#pragma acc parallel // expected-note{{invalid branch into OpenACC Compute Construct}}
+  for(int i = 0; i < 5; ++i) {
+LABEL_IN:
+    {}
+  }
+
+#pragma acc parallel
+  for(int i = 0; i < 5; ++i) {
+LABEL_NOT_CALLED:
+    {}
+  }
+
+#pragma acc parallel
+  {
+    goto ANOTHER_LOOP; // expected-error{{cannot jump from this goto statement to its label}}
+
+  }
+#pragma acc parallel// expected-note{{invalid branch into OpenACC Compute Construct}}
+
+  {
+ANOTHER_LOOP:
+    {}
+  }
+
+#pragma acc parallel
+  {
+  while (j) {
+    --j;
+    if (j < 3)
+      goto LABEL2;
+
+    if (j > 4)
+      break;
+  }
+LABEL2:
+  {}
+  }
+
+#pragma acc parallel
+  do {
+    if (j < 3)
+      goto LABEL3;
+
+    if (j > 4)
+      break; // expected-error{{invalid branch out of OpenACC Compute Construct}}
+
+LABEL3:
+  {}
+  } while (j);
+
+LABEL4:
+  {}
+#pragma acc parallel// expected-note{{invalid branch out of OpenACC Compute Construct}}
+  {
+    goto LABEL4;// expected-error{{cannot jump from this goto statement to its label}}
+  }
+
+#pragma acc parallel// expected-note{{invalid branch into OpenACC Compute Construct}}
+
+  {
+LABEL5:
+    {}
+  }
+
+  {
+    goto LABEL5;// expected-error{{cannot jump from this goto statement to its label}}
+  }
+
+#pragma acc parallel
+  {
+LABEL6:
+    {}
+    goto LABEL6;
+
+  }
+
+#pragma acc parallel
+  goto LABEL7; // expected-error{{cannot jump from this goto statement to its label}}
+#pragma acc parallel// expected-note{{invalid branch into OpenACC Compute Construct}}
+  {
+LABEL7:{}
+  }
+
+#pragma acc parallel
+  LABEL8:{}
+#pragma acc parallel// expected-note{{invalid branch out of OpenACC Compute Construct}}
+  {
+    goto LABEL8;// expected-error{{cannot jump from this goto statement to its label}}
+  }
+
+
+#pragma acc parallel// expected-note{{invalid branch into OpenACC Compute Construct}}
+  {
+LABEL9:{}
+  }
+
+  ({goto LABEL9;});// expected-error{{cannot jump from this goto statement to its label}}
+
+#pragma acc parallel// expected-note{{invalid branch out of OpenACC Compute Construct}}
+  {
+  ({goto LABEL10;});// expected-error{{cannot jump from this goto statement to its label}}
+  }
+
+LABEL10:{}
+
+  ({goto LABEL11;});// expected-error{{cannot jump from this goto statement to its label}}
+#pragma acc parallel// expected-note{{invalid branch into OpenACC Compute Construct}}
+  {
+LABEL11:{}
+  }
+
+LABEL12:{}
+#pragma acc parallel// expected-note{{invalid branch out of OpenACC Compute Construct}}
+  {
+  ({goto LABEL12;});// expected-error{{cannot jump from this goto statement to its label}}
+  }
+
+#pragma acc parallel
+  {
+  ({goto LABEL13;});
+LABEL13:{}
+  }
+
+#pragma acc parallel
+  {
+  LABEL14:{}
+  ({goto LABEL14;});
+  }
+}
+
+void IndirectGoto1() {
+  void* ptr;
+#pragma acc parallel
+  {
+LABEL1:{}
+    ptr = &&LABEL1;
+
+    goto *ptr;
+
+  }
+}
+
+void IndirectGoto2() {
+  void* ptr;
+LABEL2:{} // #GOTOLBL2
+    ptr = &&LABEL2;
+#pragma acc parallel // #GOTOPAR2
+  {
+// expected-error@+3{{cannot jump from this indirect goto statement to one of its possible targets}}
+// expected-note@#GOTOLBL2{{possible target of indirect goto statement}}
+// expected-note@#GOTOPAR2{{invalid branch out of OpenACC Compute Construct}}
+    goto *ptr;
+  }
+}
+
+void IndirectGoto3() {
+  void* ptr;
+#pragma acc parallel // #GOTOPAR3
+  {
+LABEL3:{} // #GOTOLBL3
+    ptr = &&LABEL3;
+  }
+// expected-error@+3{{cannot jump from this indirect goto statement to one of its possible targets}}
+// expected-note@#GOTOLBL3{{possible target of indirect goto statement}}
+// expected-note@#GOTOPAR3{{invalid branch into OpenACC Compute Construct}}
+  goto *ptr;
+}
+
+void IndirectGoto4() {
+  void* ptr;
+#pragma acc parallel // #GOTOPAR4
+  {
+LABEL4:{}
+    ptr = &&LABEL4;
+// expected-error@+3{{cannot jump from this indirect goto statement to one of its possible targets}}
+// expected-note@#GOTOLBL5{{possible target of indirect goto statement}}
+// expected-note@#GOTOPAR4{{invalid branch out of OpenACC Compute Construct}}
+    goto *ptr;
+  }
+LABEL5:// #GOTOLBL5
+
+  ptr=&&LABEL5;
+}
diff --git a/clang/tools/clang-installapi/ClangInstallAPI.cpp b/clang/tools/clang-installapi/ClangInstallAPI.cpp
@@ -96,7 +96,6 @@ static bool run(ArrayRef<const char *> Args, const char *ProgName) {
   auto DriverArgs = llvm::ArrayRef(Args).slice(1);
   clang::driver::Driver Driver(ProgName, llvm::sys::getDefaultTargetTriple(),
                                *Diag, "clang installapi tool");
-  Driver.setInstalledDir(llvm::sys::path::parent_path(ProgName));
   auto TargetAndMode =
       clang::driver::ToolChain::getTargetAndModeFromProgramName(ProgName);
   Driver.setTargetAndMode(TargetAndMode);

diff --git a/clang/tools/driver/driver.cpp b/clang/tools/driver/driver.cpp
@@ -323,28 +323,6 @@ static void FixupDiagPrefixExeName(TextDiagnosticPrinter *DiagClient,
   DiagClient->setPrefix(std::string(ExeBasename));
 }
 
-static void SetInstallDir(SmallVectorImpl<const char *> &argv,
-                          Driver &TheDriver, bool CanonicalPrefixes) {
-  // Attempt to find the original path used to invoke the driver, to determine
-  // the installed path. We do this manually, because we want to support that
-  // path being a symlink.
-  SmallString<128> InstalledPath(argv[0]);
-
-  // Do a PATH lookup, if there are no directory components.
-  if (llvm::sys::path::filename(InstalledPath) == InstalledPath)
-    if (llvm::ErrorOr<std::string> Tmp = llvm::sys::findProgramByName(
-            llvm::sys::path::filename(InstalledPath.str())))
-      InstalledPath = *Tmp;
-
-  // FIXME: We don't actually canonicalize this, we just make it absolute.
-  if (CanonicalPrefixes)
-    llvm::sys::fs::make_absolute(InstalledPath);
-
-  StringRef InstalledPathParent(llvm::sys::path::parent_path(InstalledPath));
-  if (llvm::sys::fs::exists(InstalledPathParent))
-    TheDriver.setInstalledDir(InstalledPathParent);
-}
-
 static int ExecuteCC1Tool(SmallVectorImpl<const char *> &ArgV,
                           const llvm::ToolContext &ToolContext) {
   // If we call the cc1 tool from the clangDriver library (through
@@ -484,7 +462,6 @@ int clang_main(int Argc, char **Argv, const llvm::ToolContext &ToolContext) {
   ProcessWarningOptions(Diags, *DiagOpts, /*ReportDiags=*/false);
 
   Driver TheDriver(Path, llvm::sys::getDefaultTargetTriple(), Diags);
-  SetInstallDir(Args, TheDriver, CanonicalPrefixes);
   auto TargetAndMode = ToolChain::getTargetAndModeFromProgramName(ProgName);
   TheDriver.setTargetAndMode(TargetAndMode);
   // If -canonical-prefixes is set, GetExecutablePath will have resolved Path

diff --git a/compiler-rt/lib/scudo/standalone/allocator_common.h b/compiler-rt/lib/scudo/standalone/allocator_common.h
@@ -40,6 +40,7 @@ template <class SizeClassAllocator> struct TransferBatch {
     B->Count = static_cast<u16>(B->Count - N);
   }
   void clear() { Count = 0; }
+  bool empty() { return Count == 0; }
   void add(CompactPtrT P) {
     DCHECK_LT(Count, MaxNumCached);
     Batch[Count++] = P;
@@ -48,6 +49,12 @@ template <class SizeClassAllocator> struct TransferBatch {
     memcpy(Array, Batch, sizeof(Batch[0]) * Count);
     clear();
   }
+
+  void moveNToArray(CompactPtrT *Array, u16 N) {
+    DCHECK_LE(N, Count);
+    memcpy(Array, Batch + Count - N, sizeof(Batch[0]) * N);
+    Count = static_cast<u16>(Count - N);
+  }
   u16 getCount() const { return Count; }
   bool isEmpty() const { return Count == 0U; }
   CompactPtrT get(u16 I) const {

diff --git a/compiler-rt/lib/scudo/standalone/primary32.h b/compiler-rt/lib/scudo/standalone/primary32.h
@@ -191,38 +191,21 @@ template <typename Config> class SizeClassAllocator32 {
     return BlockSize > PageSize;
   }
 
-  // Note that the `MaxBlockCount` will be used when we support arbitrary blocks
-  // count. Now it's the same as the number of blocks stored in the
-  // `TransferBatch`.
   u16 popBlocks(CacheT *C, uptr ClassId, CompactPtrT *ToArray,
-                UNUSED const u16 MaxBlockCount) {
-    TransferBatchT *B = popBatch(C, ClassId);
-    if (!B)
-      return 0;
-
-    const u16 Count = B->getCount();
-    DCHECK_GT(Count, 0U);
-    B->moveToArray(ToArray);
-
-    if (ClassId != SizeClassMap::BatchClassId)
-      C->deallocate(SizeClassMap::BatchClassId, B);
-
-    return Count;
-  }
-
-  TransferBatchT *popBatch(CacheT *C, uptr ClassId) {
+                const u16 MaxBlockCount) {
     DCHECK_LT(ClassId, NumClasses);
     SizeClassInfo *Sci = getSizeClassInfo(ClassId);
     ScopedLock L(Sci->Mutex);
-    TransferBatchT *B = popBatchImpl(C, ClassId, Sci);
-    if (UNLIKELY(!B)) {
+
+    u16 PopCount = popBlocksImpl(C, ClassId, Sci, ToArray, MaxBlockCount);
+    if (UNLIKELY(PopCount == 0)) {
       if (UNLIKELY(!populateFreeList(C, ClassId, Sci)))
-        return nullptr;
-      B = popBatchImpl(C, ClassId, Sci);
-      // if `populateFreeList` succeeded, we are supposed to get free blocks.
-      DCHECK_NE(B, nullptr);
+        return 0U;
+      PopCount = popBlocksImpl(C, ClassId, Sci, ToArray, MaxBlockCount);
+      DCHECK_NE(PopCount, 0U);
     }
-    return B;
+
+    return PopCount;
   }
 
   // Push the array of free blocks to the designated batch group.
@@ -510,7 +493,7 @@ template <typename Config> class SizeClassAllocator32 {
     // by TransferBatch is also free for use. We don't need to recycle the
     // TransferBatch. Note that the correctness is maintained by the invariant,
     //
-    //   The unit of each popBatch() request is entire TransferBatch. Return
+    //   Each popBlocks() request returns the entire TransferBatch. Returning
     //   part of the blocks in a TransferBatch is invalid.
     //
     // This ensures that TransferBatch won't leak the address itself while it's
@@ -634,7 +617,7 @@ template <typename Config> class SizeClassAllocator32 {
       BG->Batches.push_front(TB);
       BG->PushedBlocks = 0;
       BG->BytesInBGAtLastCheckpoint = 0;
-      BG->MaxCachedPerBatch = CacheT::getMaxCached(getSizeByClassId(ClassId));
+      BG->MaxCachedPerBatch = TransferBatchT::MaxNumCached;
 
       return BG;
     };
@@ -726,14 +709,11 @@ template <typename Config> class SizeClassAllocator32 {
     InsertBlocks(Cur, Array + Size - Count, Count);
   }
 
-  // Pop one TransferBatch from a BatchGroup. The BatchGroup with the smallest
-  // group id will be considered first.
-  //
-  // The region mutex needs to be held while calling this method.
-  TransferBatchT *popBatchImpl(CacheT *C, uptr ClassId, SizeClassInfo *Sci)
+  u16 popBlocksImpl(CacheT *C, uptr ClassId, SizeClassInfo *Sci,
+                    CompactPtrT *ToArray, const u16 MaxBlockCount)
       REQUIRES(Sci->Mutex) {
     if (Sci->FreeListInfo.BlockList.empty())
-      return nullptr;
+      return 0U;
 
     SinglyLinkedList<TransferBatchT> &Batches =
         Sci->FreeListInfo.BlockList.front()->Batches;
@@ -746,33 +726,57 @@ template <typename Config> class SizeClassAllocator32 {
       // Block used by `BatchGroup` is from BatchClassId. Turn the block into
       // `TransferBatch` with single block.
       TransferBatchT *TB = reinterpret_cast<TransferBatchT *>(BG);
-      TB->clear();
-      TB->add(
-          compactPtr(SizeClassMap::BatchClassId, reinterpret_cast<uptr>(TB)));
+      ToArray[0] =
+          compactPtr(SizeClassMap::BatchClassId, reinterpret_cast<uptr>(TB));
       Sci->FreeListInfo.PoppedBlocks += 1;
-      return TB;
+      return 1U;
     }
 
+    // So far, instead of always filling the blocks to `MaxBlockCount`, we only
+    // examine single `TransferBatch` to minimize the time spent on the primary
+    // allocator. Besides, the sizes of `TransferBatch` and
+    // `CacheT::getMaxCached()` may also impact the time spent on accessing the
+    // primary allocator.
+    // TODO(chiahungduan): Evaluate if we want to always prepare `MaxBlockCount`
+    // blocks and/or adjust the size of `TransferBatch` according to
+    // `CacheT::getMaxCached()`.
     TransferBatchT *B = Batches.front();
-    Batches.pop_front();
     DCHECK_NE(B, nullptr);
     DCHECK_GT(B->getCount(), 0U);
 
-    if (Batches.empty()) {
-      BatchGroupT *BG = Sci->FreeListInfo.BlockList.front();
-      Sci->FreeListInfo.BlockList.pop_front();
-
-      // We don't keep BatchGroup with zero blocks to avoid empty-checking while
-      // allocating. Note that block used by constructing BatchGroup is recorded
-      // as free blocks in the last element of BatchGroup::Batches. Which means,
-      // once we pop the last TransferBatch, the block is implicitly
-      // deallocated.
+    // BachClassId should always take all blocks in the TransferBatch. Read the
+    // comment in `pushBatchClassBlocks()` for more details.
+    const u16 PopCount = ClassId == SizeClassMap::BatchClassId
+                             ? B->getCount()
+                             : Min(MaxBlockCount, B->getCount());
+    B->moveNToArray(ToArray, PopCount);
+
+    // TODO(chiahungduan): The deallocation of unused BatchClassId blocks can be
+    // done without holding `Mutex`.
+    if (B->empty()) {
+      Batches.pop_front();
+      // `TransferBatch` of BatchClassId is self-contained, no need to
+      // deallocate. Read the comment in `pushBatchClassBlocks()` for more
+      // details.
       if (ClassId != SizeClassMap::BatchClassId)
-        C->deallocate(SizeClassMap::BatchClassId, BG);
+        C->deallocate(SizeClassMap::BatchClassId, B);
+
+      if (Batches.empty()) {
+        BatchGroupT *BG = Sci->FreeListInfo.BlockList.front();
+        Sci->FreeListInfo.BlockList.pop_front();
+
+        // We don't keep BatchGroup with zero blocks to avoid empty-checking
+        // while allocating. Note that block used for constructing BatchGroup is
+        // recorded as free blocks in the last element of BatchGroup::Batches.
+        // Which means, once we pop the last TransferBatch, the block is
+        // implicitly deallocated.
+        if (ClassId != SizeClassMap::BatchClassId)
+          C->deallocate(SizeClassMap::BatchClassId, BG);
+      }
     }
 
-    Sci->FreeListInfo.PoppedBlocks += B->getCount();
-    return B;
+    Sci->FreeListInfo.PoppedBlocks += PopCount;
+    return PopCount;
   }
 
   NOINLINE bool populateFreeList(CacheT *C, uptr ClassId, SizeClassInfo *Sci)

diff --git a/compiler-rt/lib/scudo/standalone/primary64.h b/compiler-rt/lib/scudo/standalone/primary64.h
@@ -12,6 +12,7 @@
 #include "allocator_common.h"
 #include "bytemap.h"
 #include "common.h"
+#include "condition_variable.h"
 #include "list.h"
 #include "local_cache.h"
 #include "mem_map.h"
@@ -22,8 +23,6 @@
 #include "string_utils.h"
 #include "thread_annotations.h"
 
-#include "condition_variable.h"
-
 namespace scudo {
 
 // SizeClassAllocator64 is an allocator tuned for 64-bit address space.
@@ -221,41 +220,24 @@ template <typename Config> class SizeClassAllocator64 {
     DCHECK_EQ(BlocksInUse, BatchClassUsedInFreeLists);
   }
 
-  // Note that the `MaxBlockCount` will be used when we support arbitrary blocks
-  // count. Now it's the same as the number of blocks stored in the
-  // `TransferBatch`.
   u16 popBlocks(CacheT *C, uptr ClassId, CompactPtrT *ToArray,
-                UNUSED const u16 MaxBlockCount) {
-    TransferBatchT *B = popBatch(C, ClassId);
-    if (!B)
-      return 0;
-
-    const u16 Count = B->getCount();
-    DCHECK_GT(Count, 0U);
-    B->moveToArray(ToArray);
-
-    if (ClassId != SizeClassMap::BatchClassId)
-      C->deallocate(SizeClassMap::BatchClassId, B);
-
-    return Count;
-  }
-
-  TransferBatchT *popBatch(CacheT *C, uptr ClassId) {
+                const u16 MaxBlockCount) {
     DCHECK_LT(ClassId, NumClasses);
     RegionInfo *Region = getRegionInfo(ClassId);
+    u16 PopCount = 0;
 
     {
       ScopedLock L(Region->FLLock);
-      TransferBatchT *B = popBatchImpl(C, ClassId, Region);
-      if (LIKELY(B))
-        return B;
+      PopCount = popBlocksImpl(C, ClassId, Region, ToArray, MaxBlockCount);
+      if (PopCount != 0U)
+        return PopCount;
     }
 
     bool ReportRegionExhausted = false;
-    TransferBatchT *B = nullptr;
 
     if (conditionVariableEnabled()) {
-      B = popBatchWithCV(C, ClassId, Region, ReportRegionExhausted);
+      PopCount = popBlocksWithCV(C, ClassId, Region, ToArray, MaxBlockCount,
+                                 ReportRegionExhausted);
     } else {
       while (true) {
         // When two threads compete for `Region->MMLock`, we only want one of
@@ -264,13 +246,15 @@ template <typename Config> class SizeClassAllocator64 {
         ScopedLock ML(Region->MMLock);
         {
           ScopedLock FL(Region->FLLock);
-          if ((B = popBatchImpl(C, ClassId, Region)))
-            break;
+          PopCount = popBlocksImpl(C, ClassId, Region, ToArray, MaxBlockCount);
+          if (PopCount != 0U)
+            return PopCount;
         }
 
         const bool RegionIsExhausted = Region->Exhausted;
         if (!RegionIsExhausted)
-          B = populateFreeListAndPopBatch(C, ClassId, Region);
+          PopCount = populateFreeListAndPopBlocks(C, ClassId, Region, ToArray,
+                                                  MaxBlockCount);
         ReportRegionExhausted = !RegionIsExhausted && Region->Exhausted;
         break;
       }
@@ -286,7 +270,7 @@ template <typename Config> class SizeClassAllocator64 {
         reportOutOfBatchClass();
     }
 
-    return B;
+    return PopCount;
   }
 
   // Push the array of free blocks to the designated batch group.
@@ -640,7 +624,7 @@ template <typename Config> class SizeClassAllocator64 {
     // by TransferBatch is also free for use. We don't need to recycle the
     // TransferBatch. Note that the correctness is maintained by the invariant,
     //
-    //   The unit of each popBatch() request is entire TransferBatch. Return
+    //   Each popBlocks() request returns the entire TransferBatch. Returning
     //   part of the blocks in a TransferBatch is invalid.
     //
     // This ensures that TransferBatch won't leak the address itself while it's
@@ -763,7 +747,7 @@ template <typename Config> class SizeClassAllocator64 {
       BG->Batches.push_front(TB);
       BG->PushedBlocks = 0;
       BG->BytesInBGAtLastCheckpoint = 0;
-      BG->MaxCachedPerBatch = CacheT::getMaxCached(getSizeByClassId(ClassId));
+      BG->MaxCachedPerBatch = TransferBatchT::MaxNumCached;
 
       return BG;
     };
@@ -855,9 +839,10 @@ template <typename Config> class SizeClassAllocator64 {
     InsertBlocks(Cur, Array + Size - Count, Count);
   }
 
-  TransferBatchT *popBatchWithCV(CacheT *C, uptr ClassId, RegionInfo *Region,
-                                 bool &ReportRegionExhausted) {
-    TransferBatchT *B = nullptr;
+  u16 popBlocksWithCV(CacheT *C, uptr ClassId, RegionInfo *Region,
+                      CompactPtrT *ToArray, const u16 MaxBlockCount,
+                      bool &ReportRegionExhausted) {
+    u16 PopCount = 0;
 
     while (true) {
       // We only expect one thread doing the freelist refillment and other
@@ -878,7 +863,8 @@ template <typename Config> class SizeClassAllocator64 {
 
         const bool RegionIsExhausted = Region->Exhausted;
         if (!RegionIsExhausted)
-          B = populateFreeListAndPopBatch(C, ClassId, Region);
+          PopCount = populateFreeListAndPopBlocks(C, ClassId, Region, ToArray,
+                                                  MaxBlockCount);
         ReportRegionExhausted = !RegionIsExhausted && Region->Exhausted;
 
         {
@@ -905,7 +891,8 @@ template <typename Config> class SizeClassAllocator64 {
       // blocks were used up right after the refillment. Therefore, we have to
       // check if someone is still populating the freelist.
       ScopedLock FL(Region->FLLock);
-      if (LIKELY(B = popBatchImpl(C, ClassId, Region)))
+      PopCount = popBlocksImpl(C, ClassId, Region, ToArray, MaxBlockCount);
+      if (PopCount != 0U)
         break;
 
       if (!Region->isPopulatingFreeList)
@@ -918,21 +905,19 @@ template <typename Config> class SizeClassAllocator64 {
       // `pushBatchClassBlocks()` and `mergeGroupsToReleaseBack()`.
       Region->FLLockCV.wait(Region->FLLock);
 
-      if (LIKELY(B = popBatchImpl(C, ClassId, Region)))
+      PopCount = popBlocksImpl(C, ClassId, Region, ToArray, MaxBlockCount);
+      if (PopCount != 0U)
         break;
     }
 
-    return B;
+    return PopCount;
   }
 
-  // Pop one TransferBatch from a BatchGroup. The BatchGroup with the smallest
-  // group id will be considered first.
-  //
-  // The region mutex needs to be held while calling this method.
-  TransferBatchT *popBatchImpl(CacheT *C, uptr ClassId, RegionInfo *Region)
+  u16 popBlocksImpl(CacheT *C, uptr ClassId, RegionInfo *Region,
+                    CompactPtrT *ToArray, const u16 MaxBlockCount)
       REQUIRES(Region->FLLock) {
     if (Region->FreeListInfo.BlockList.empty())
-      return nullptr;
+      return 0U;
 
     SinglyLinkedList<TransferBatchT> &Batches =
         Region->FreeListInfo.BlockList.front()->Batches;
@@ -945,39 +930,64 @@ template <typename Config> class SizeClassAllocator64 {
       // Block used by `BatchGroup` is from BatchClassId. Turn the block into
       // `TransferBatch` with single block.
       TransferBatchT *TB = reinterpret_cast<TransferBatchT *>(BG);
-      TB->clear();
-      TB->add(
-          compactPtr(SizeClassMap::BatchClassId, reinterpret_cast<uptr>(TB)));
+      ToArray[0] =
+          compactPtr(SizeClassMap::BatchClassId, reinterpret_cast<uptr>(TB));
       Region->FreeListInfo.PoppedBlocks += 1;
-      return TB;
+      return 1U;
     }
 
+    // So far, instead of always filling blocks to `MaxBlockCount`, we only
+    // examine single `TransferBatch` to minimize the time spent in the primary
+    // allocator. Besides, the sizes of `TransferBatch` and
+    // `CacheT::getMaxCached()` may also impact the time spent on accessing the
+    // primary allocator.
+    // TODO(chiahungduan): Evaluate if we want to always prepare `MaxBlockCount`
+    // blocks and/or adjust the size of `TransferBatch` according to
+    // `CacheT::getMaxCached()`.
     TransferBatchT *B = Batches.front();
-    Batches.pop_front();
     DCHECK_NE(B, nullptr);
     DCHECK_GT(B->getCount(), 0U);
 
-    if (Batches.empty()) {
-      BatchGroupT *BG = Region->FreeListInfo.BlockList.front();
-      Region->FreeListInfo.BlockList.pop_front();
-
-      // We don't keep BatchGroup with zero blocks to avoid empty-checking while
-      // allocating. Note that block used by constructing BatchGroup is recorded
-      // as free blocks in the last element of BatchGroup::Batches. Which means,
-      // once we pop the last TransferBatch, the block is implicitly
-      // deallocated.
+    // BachClassId should always take all blocks in the TransferBatch. Read the
+    // comment in `pushBatchClassBlocks()` for more details.
+    const u16 PopCount = ClassId == SizeClassMap::BatchClassId
+                             ? B->getCount()
+                             : Min(MaxBlockCount, B->getCount());
+    B->moveNToArray(ToArray, PopCount);
+
+    // TODO(chiahungduan): The deallocation of unused BatchClassId blocks can be
+    // done without holding `FLLock`.
+    if (B->empty()) {
+      Batches.pop_front();
+      // `TransferBatch` of BatchClassId is self-contained, no need to
+      // deallocate. Read the comment in `pushBatchClassBlocks()` for more
+      // details.
       if (ClassId != SizeClassMap::BatchClassId)
-        C->deallocate(SizeClassMap::BatchClassId, BG);
+        C->deallocate(SizeClassMap::BatchClassId, B);
+
+      if (Batches.empty()) {
+        BatchGroupT *BG = Region->FreeListInfo.BlockList.front();
+        Region->FreeListInfo.BlockList.pop_front();
+
+        // We don't keep BatchGroup with zero blocks to avoid empty-checking
+        // while allocating. Note that block used for constructing BatchGroup is
+        // recorded as free blocks in the last element of BatchGroup::Batches.
+        // Which means, once we pop the last TransferBatch, the block is
+        // implicitly deallocated.
+        if (ClassId != SizeClassMap::BatchClassId)
+          C->deallocate(SizeClassMap::BatchClassId, BG);
+      }
     }
 
-    Region->FreeListInfo.PoppedBlocks += B->getCount();
+    Region->FreeListInfo.PoppedBlocks += PopCount;
 
-    return B;
+    return PopCount;
   }
 
-  // Refill the freelist and return one batch.
-  NOINLINE TransferBatchT *populateFreeListAndPopBatch(CacheT *C, uptr ClassId,
-                                                       RegionInfo *Region)
+  NOINLINE u16 populateFreeListAndPopBlocks(CacheT *C, uptr ClassId,
+                                            RegionInfo *Region,
+                                            CompactPtrT *ToArray,
+                                            const u16 MaxBlockCount)
       REQUIRES(Region->MMLock) EXCLUDES(Region->FLLock) {
     const uptr Size = getSizeByClassId(ClassId);
     const u16 MaxCount = CacheT::getMaxCached(Size);
@@ -994,15 +1004,15 @@ template <typename Config> class SizeClassAllocator64 {
       const uptr RegionBase = RegionBeg - getRegionBaseByClassId(ClassId);
       if (UNLIKELY(RegionBase + MappedUser + MapSize > RegionSize)) {
         Region->Exhausted = true;
-        return nullptr;
+        return 0U;
       }
 
       if (UNLIKELY(!Region->MemMapInfo.MemMap.remap(
               RegionBeg + MappedUser, MapSize, "scudo:primary",
               MAP_ALLOWNOMEM | MAP_RESIZABLE |
                   (useMemoryTagging<Config>(Options.load()) ? MAP_MEMTAG
                                                             : 0)))) {
-        return nullptr;
+        return 0U;
       }
       Region->MemMapInfo.MappedUser += MapSize;
       C->getStats().add(StatMapped, MapSize);
@@ -1049,8 +1059,9 @@ template <typename Config> class SizeClassAllocator64 {
       pushBatchClassBlocks(Region, ShuffleArray, NumberOfBlocks);
     }
 
-    TransferBatchT *B = popBatchImpl(C, ClassId, Region);
-    DCHECK_NE(B, nullptr);
+    const u16 PopCount =
+        popBlocksImpl(C, ClassId, Region, ToArray, MaxBlockCount);
+    DCHECK_NE(PopCount, 0U);
 
     // Note that `PushedBlocks` and `PoppedBlocks` are supposed to only record
     // the requests from `PushBlocks` and `PopBatch` which are external
@@ -1062,7 +1073,7 @@ template <typename Config> class SizeClassAllocator64 {
     C->getStats().add(StatFree, AllocatedUser);
     Region->MemMapInfo.AllocatedUser += AllocatedUser;
 
-    return B;
+    return PopCount;
   }
 
   void getStats(ScopedString *Str, uptr ClassId, RegionInfo *Region)
@@ -1186,7 +1197,7 @@ template <typename Config> class SizeClassAllocator64 {
     }
 
     // Note that we have extracted the `GroupsToRelease` from region freelist.
-    // It's safe to let pushBlocks()/popBatches() access the remaining region
+    // It's safe to let pushBlocks()/popBlocks() access the remaining region
     // freelist. In the steps 3 and 4, we will temporarily release the FLLock
     // and lock it again before step 5.
 

diff --git a/compiler-rt/lib/scudo/standalone/tests/primary_test.cpp b/compiler-rt/lib/scudo/standalone/tests/primary_test.cpp
@@ -237,37 +237,33 @@ struct SmallRegionsConfig {
 // For the 32-bit one, it requires actually exhausting memory, so we skip it.
 TEST(ScudoPrimaryTest, Primary64OOM) {
   using Primary = scudo::SizeClassAllocator64<SmallRegionsConfig>;
-  using TransferBatch = Primary::TransferBatchT;
   Primary Allocator;
   Allocator.init(/*ReleaseToOsInterval=*/-1);
   typename Primary::CacheT Cache;
   scudo::GlobalStats Stats;
   Stats.init();
   Cache.init(&Stats, &Allocator);
   bool AllocationFailed = false;
-  std::vector<TransferBatch *> Batches;
+  std::vector<void *> Blocks;
   const scudo::uptr ClassId = Primary::SizeClassMap::LargestClassId;
   const scudo::uptr Size = Primary::getSizeByClassId(ClassId);
-  typename Primary::CacheT::CompactPtrT Blocks[TransferBatch::MaxNumCached];
+  const scudo::u16 MaxCachedBlockCount = Primary::CacheT::getMaxCached(Size);
 
   for (scudo::uptr I = 0; I < 10000U; I++) {
-    TransferBatch *B = Allocator.popBatch(&Cache, ClassId);
-    if (!B) {
-      AllocationFailed = true;
-      break;
+    for (scudo::uptr J = 0; J < MaxCachedBlockCount; ++J) {
+      void *Ptr = Cache.allocate(ClassId);
+      if (Ptr == nullptr) {
+        AllocationFailed = true;
+        break;
+      }
+      memset(Ptr, 'B', Size);
+      Blocks.push_back(Ptr);
     }
-    for (scudo::u16 J = 0; J < B->getCount(); J++)
-      memset(Allocator.decompactPtr(ClassId, B->get(J)), 'B', Size);
-    Batches.push_back(B);
-  }
-  while (!Batches.empty()) {
-    TransferBatch *B = Batches.back();
-    Batches.pop_back();
-    const scudo::u16 Count = B->getCount();
-    B->moveToArray(Blocks);
-    Allocator.pushBlocks(&Cache, ClassId, Blocks, Count);
-    Cache.deallocate(Primary::SizeClassMap::BatchClassId, B);
   }
+
+  for (auto *Ptr : Blocks)
+    Cache.deallocate(ClassId, Ptr);
+
   Cache.destroy(nullptr);
   Allocator.releaseToOS(scudo::ReleaseToOS::Force);
   scudo::ScopedString Str;
@@ -342,7 +338,7 @@ SCUDO_TYPED_TEST(ScudoPrimaryTest, PrimaryThreaded) {
           V.push_back(std::make_pair(ClassId, P));
       }
 
-      // Try to interleave pushBlocks(), popBatch() and releaseToOS().
+      // Try to interleave pushBlocks(), popBlocks() and releaseToOS().
       Allocator->releaseToOS(scudo::ReleaseToOS::Force);
 
       while (!V.empty()) {

diff --git a/flang/include/flang/Lower/LoweringOptions.def b/flang/include/flang/Lower/LoweringOptions.def
@@ -24,8 +24,8 @@ LOWERINGOPT(Name, Bits, Default)
 /// If true, lower transpose without a runtime call.
 ENUM_LOWERINGOPT(OptimizeTranspose, unsigned, 1, 1)
 
-/// If true, enable polymorphic type lowering feature. Off by default.
-ENUM_LOWERINGOPT(PolymorphicTypeImpl, unsigned, 1, 0)
+/// If true, enable polymorphic type lowering feature. On by default.
+ENUM_LOWERINGOPT(PolymorphicTypeImpl, unsigned, 1, 1)
 
 /// If true, lower to High level FIR before lowering to FIR. On by default.
 ENUM_LOWERINGOPT(LowerToHighLevelFIR, unsigned, 1, 1)

diff --git a/flang/include/flang/Runtime/reduction.h b/flang/include/flang/Runtime/reduction.h
@@ -364,9 +364,12 @@ double RTDECL(Norm2_8)(
 #if LDBL_MANT_DIG == 64
 long double RTDECL(Norm2_10)(
     const Descriptor &, const char *source, int line, int dim = 0);
-#elif LDBL_MANT_DIG == 113
+#endif
+#if LDBL_MANT_DIG == 113 || HAS_FLOAT128
 long double RTDECL(Norm2_16)(
     const Descriptor &, const char *source, int line, int dim = 0);
+void RTDECL(Norm2DimReal16)(
+    Descriptor &, const Descriptor &, int dim, const char *source, int line);
 #endif
 void RTDECL(Norm2Dim)(
     Descriptor &, const Descriptor &, int dim, const char *source, int line);

diff --git a/flang/lib/Optimizer/Builder/Runtime/Reduction.cpp b/flang/lib/Optimizer/Builder/Runtime/Reduction.cpp
@@ -149,6 +149,22 @@ struct ForcedNorm2Real16 {
   }
 };
 
+/// Placeholder for real*16 version of Norm2Dim Intrinsic
+struct ForcedNorm2DimReal16 {
+  static constexpr const char *name = ExpandAndQuoteKey(RTNAME(Norm2DimReal16));
+  static constexpr fir::runtime::FuncTypeBuilderFunc getTypeModel() {
+    return [](mlir::MLIRContext *ctx) {
+      auto boxTy =
+          fir::runtime::getModel<const Fortran::runtime::Descriptor &>()(ctx);
+      auto strTy = fir::ReferenceType::get(mlir::IntegerType::get(ctx, 8));
+      auto intTy = mlir::IntegerType::get(ctx, 8 * sizeof(int));
+      return mlir::FunctionType::get(
+          ctx, {fir::ReferenceType::get(boxTy), boxTy, intTy, strTy, intTy},
+          mlir::NoneType::get(ctx));
+    };
+  }
+};
+
 /// Placeholder for real*10 version of Product Intrinsic
 struct ForcedProductReal10 {
   static constexpr const char *name = ExpandAndQuoteKey(RTNAME(ProductReal10));
@@ -876,7 +892,14 @@ mlir::Value fir::runtime::genMinval(fir::FirOpBuilder &builder,
 void fir::runtime::genNorm2Dim(fir::FirOpBuilder &builder, mlir::Location loc,
                                mlir::Value resultBox, mlir::Value arrayBox,
                                mlir::Value dim) {
-  auto func = fir::runtime::getRuntimeFunc<mkRTKey(Norm2Dim)>(loc, builder);
+  mlir::func::FuncOp func;
+  auto ty = arrayBox.getType();
+  auto arrTy = fir::dyn_cast_ptrOrBoxEleTy(ty);
+  auto eleTy = arrTy.cast<fir::SequenceType>().getEleTy();
+  if (eleTy.isF128())
+    func = fir::runtime::getRuntimeFunc<ForcedNorm2DimReal16>(loc, builder);
+  else
+    func = fir::runtime::getRuntimeFunc<mkRTKey(Norm2Dim)>(loc, builder);
   auto fTy = func.getFunctionType();
   auto sourceFile = fir::factory::locationToFilename(builder, loc);
   auto sourceLine =

diff --git a/flang/runtime/Float128Math/CMakeLists.txt b/flang/runtime/Float128Math/CMakeLists.txt
@@ -69,6 +69,7 @@ set(sources
   log.cpp
   log10.cpp
   lround.cpp
+  norm2.cpp
   pow.cpp
   round.cpp
   sin.cpp

diff --git a/flang/runtime/Float128Math/math-entries.h b/flang/runtime/Float128Math/math-entries.h
@@ -54,6 +54,7 @@ namespace Fortran::runtime {
   };
 
 // Define fallback callers.
+DEFINE_FALLBACK(Abs)
 DEFINE_FALLBACK(Acos)
 DEFINE_FALLBACK(Acosh)
 DEFINE_FALLBACK(Asin)
@@ -99,6 +100,7 @@ DEFINE_FALLBACK(Yn)
 // Use STD math functions. They provide IEEE-754 128-bit float
 // support either via 'long double' or __float128.
 // The Bessel's functions are not present in STD namespace.
+DEFINE_SIMPLE_ALIAS(Abs, std::abs)
 DEFINE_SIMPLE_ALIAS(Acos, std::acos)
 DEFINE_SIMPLE_ALIAS(Acosh, std::acosh)
 DEFINE_SIMPLE_ALIAS(Asin, std::asin)
@@ -155,6 +157,7 @@ DEFINE_SIMPLE_ALIAS(Yn, ynl)
 #elif HAS_QUADMATHLIB
 // Define wrapper callers for libquadmath.
 #include "quadmath.h"
+DEFINE_SIMPLE_ALIAS(Abs, fabsq)
 DEFINE_SIMPLE_ALIAS(Acos, acosq)
 DEFINE_SIMPLE_ALIAS(Acosh, acoshq)
 DEFINE_SIMPLE_ALIAS(Asin, asinq)
@@ -191,6 +194,19 @@ DEFINE_SIMPLE_ALIAS(Y0, y0q)
 DEFINE_SIMPLE_ALIAS(Y1, y1q)
 DEFINE_SIMPLE_ALIAS(Yn, ynq)
 #endif
+
+extern "C" {
+// Declarations of the entry points that might be referenced
+// within the Float128Math library itself.
+// Note that not all of these entry points are actually
+// defined in this library. Some of them are used just
+// as template parameters to call the corresponding callee directly.
+CppTypeFor<TypeCategory::Real, 16> RTDECL(AbsF128)(
+    CppTypeFor<TypeCategory::Real, 16> x);
+CppTypeFor<TypeCategory::Real, 16> RTDECL(SqrtF128)(
+    CppTypeFor<TypeCategory::Real, 16> x);
+} // extern "C"
+
 } // namespace Fortran::runtime
 
 #endif // FORTRAN_RUNTIME_FLOAT128MATH_MATH_ENTRIES_H_
diff --git a/flang/runtime/Float128Math/norm2.cpp b/flang/runtime/Float128Math/norm2.cpp
@@ -0,0 +1,59 @@
+//===-- runtime/Float128Math/norm2.cpp ------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "math-entries.h"
+#include "reduction-templates.h"
+#include <cmath>
+
+#if LDBL_MANT_DIG == 113 || HAS_FLOAT128
+
+namespace {
+using namespace Fortran::runtime;
+
+using AccumType = Norm2AccumType<16>;
+
+struct ABSTy {
+  static AccumType compute(AccumType x) {
+    return Sqrt<RTNAME(AbsF128)>::invoke(x);
+  }
+};
+
+struct SQRTTy {
+  static AccumType compute(AccumType x) {
+    return Sqrt<RTNAME(SqrtF128)>::invoke(x);
+  }
+};
+
+using Float128Norm2Accumulator = Norm2Accumulator<16, ABSTy, SQRTTy>;
+} // namespace
+
+namespace Fortran::runtime {
+extern "C" {
+
+CppTypeFor<TypeCategory::Real, 16> RTDEF(Norm2_16)(
+    const Descriptor &x, const char *source, int line, int dim) {
+  auto accumulator{::Float128Norm2Accumulator(x)};
+  return GetTotalReduction<TypeCategory::Real, 16>(
+      x, source, line, dim, nullptr, accumulator, "NORM2");
+}
+
+void RTDEF(Norm2DimReal16)(Descriptor &result, const Descriptor &x, int dim,
+    const char *source, int line) {
+  Terminator terminator{source, line};
+  auto type{x.type().GetCategoryAndKind()};
+  RUNTIME_CHECK(terminator, type);
+  RUNTIME_CHECK(
+      terminator, type->first == TypeCategory::Real && type->second == 16);
+  DoMaxMinNorm2<TypeCategory::Real, 16, ::Float128Norm2Accumulator>(
+      result, x, dim, nullptr, "NORM2", terminator);
+}
+
+} // extern "C"
+} // namespace Fortran::runtime
+
+#endif
diff --git a/flang/runtime/extrema.cpp b/flang/runtime/extrema.cpp
@@ -528,35 +528,6 @@ inline RT_API_ATTRS CppTypeFor<CAT, KIND> TotalNumericMaxOrMin(
       NumericExtremumAccumulator<CAT, KIND, IS_MAXVAL>{x}, intrinsic);
 }
 
-template <TypeCategory CAT, int KIND, typename ACCUMULATOR>
-static RT_API_ATTRS void DoMaxMinNorm2(Descriptor &result, const Descriptor &x,
-    int dim, const Descriptor *mask, const char *intrinsic,
-    Terminator &terminator) {
-  using Type = CppTypeFor<CAT, KIND>;
-  ACCUMULATOR accumulator{x};
-  if (dim == 0 || x.rank() == 1) {
-    // Total reduction
-
-    // Element size of the destination descriptor is the same
-    // as the element size of the source.
-    result.Establish(x.type(), x.ElementBytes(), nullptr, 0, nullptr,
-        CFI_attribute_allocatable);
-    if (int stat{result.Allocate()}) {
-      terminator.Crash(
-          "%s: could not allocate memory for result; STAT=%d", intrinsic, stat);
-    }
-    DoTotalReduction<Type>(x, dim, mask, accumulator, intrinsic, terminator);
-    accumulator.GetResult(result.OffsetElement<Type>());
-  } else {
-    // Partial reduction
-
-    // Element size of the destination descriptor is the same
-    // as the element size of the source.
-    PartialReduction<ACCUMULATOR, CAT, KIND>(result, x, x.ElementBytes(), dim,
-        mask, terminator, intrinsic, accumulator);
-  }
-}
-
 template <TypeCategory CAT, bool IS_MAXVAL> struct MaxOrMinHelper {
   template <int KIND> struct Functor {
     RT_API_ATTRS void operator()(Descriptor &result, const Descriptor &x,
@@ -802,66 +773,11 @@ RT_EXT_API_GROUP_END
 
 // NORM2
 
-RT_VAR_GROUP_BEGIN
-
-// Use at least double precision for accumulators.
-// Don't use __float128, it doesn't work with abs() or sqrt() yet.
-static constexpr RT_CONST_VAR_ATTRS int largestLDKind {
-#if LDBL_MANT_DIG == 113
-  16
-#elif LDBL_MANT_DIG == 64
-  10
-#else
-  8
-#endif
-};
-
-RT_VAR_GROUP_END
-
-template <int KIND> class Norm2Accumulator {
-public:
-  using Type = CppTypeFor<TypeCategory::Real, KIND>;
-  using AccumType =
-      CppTypeFor<TypeCategory::Real, std::clamp(KIND, 8, largestLDKind)>;
-  explicit RT_API_ATTRS Norm2Accumulator(const Descriptor &array)
-      : array_{array} {}
-  RT_API_ATTRS void Reinitialize() { max_ = sum_ = 0; }
-  template <typename A>
-  RT_API_ATTRS void GetResult(A *p, int /*zeroBasedDim*/ = -1) const {
-    // m * sqrt(1 + sum((others(:)/m)**2))
-    *p = static_cast<Type>(max_ * std::sqrt(1 + sum_));
-  }
-  RT_API_ATTRS bool Accumulate(Type x) {
-    auto absX{std::abs(static_cast<AccumType>(x))};
-    if (!max_) {
-      max_ = absX;
-    } else if (absX > max_) {
-      auto t{max_ / absX}; // < 1.0
-      auto tsq{t * t};
-      sum_ *= tsq; // scale sum to reflect change to the max
-      sum_ += tsq; // include a term for the previous max
-      max_ = absX;
-    } else { // absX <= max_
-      auto t{absX / max_};
-      sum_ += t * t;
-    }
-    return true;
-  }
-  template <typename A>
-  RT_API_ATTRS bool AccumulateAt(const SubscriptValue at[]) {
-    return Accumulate(*array_.Element<A>(at));
-  }
-
-private:
-  const Descriptor &array_;
-  AccumType max_{0}; // value (m) with largest magnitude
-  AccumType sum_{0}; // sum((others(:)/m)**2)
-};
-
 template <int KIND> struct Norm2Helper {
   RT_API_ATTRS void operator()(Descriptor &result, const Descriptor &x, int dim,
       const Descriptor *mask, Terminator &terminator) const {
-    DoMaxMinNorm2<TypeCategory::Real, KIND, Norm2Accumulator<KIND>>(
+    DoMaxMinNorm2<TypeCategory::Real, KIND,
+        typename Norm2AccumulatorGetter<KIND>::Type>(
         result, x, dim, mask, "NORM2", terminator);
   }
 };
@@ -872,26 +788,27 @@ RT_EXT_API_GROUP_BEGIN
 // TODO: REAL(2 & 3)
 CppTypeFor<TypeCategory::Real, 4> RTDEF(Norm2_4)(
     const Descriptor &x, const char *source, int line, int dim) {
-  return GetTotalReduction<TypeCategory::Real, 4>(
-      x, source, line, dim, nullptr, Norm2Accumulator<4>{x}, "NORM2");
+  return GetTotalReduction<TypeCategory::Real, 4>(x, source, line, dim, nullptr,
+      Norm2AccumulatorGetter<4>::create(x), "NORM2");
 }
 CppTypeFor<TypeCategory::Real, 8> RTDEF(Norm2_8)(
     const Descriptor &x, const char *source, int line, int dim) {
-  return GetTotalReduction<TypeCategory::Real, 8>(
-      x, source, line, dim, nullptr, Norm2Accumulator<8>{x}, "NORM2");
+  return GetTotalReduction<TypeCategory::Real, 8>(x, source, line, dim, nullptr,
+      Norm2AccumulatorGetter<8>::create(x), "NORM2");
 }
 #if LDBL_MANT_DIG == 64
 CppTypeFor<TypeCategory::Real, 10> RTDEF(Norm2_10)(
     const Descriptor &x, const char *source, int line, int dim) {
-  return GetTotalReduction<TypeCategory::Real, 10>(
-      x, source, line, dim, nullptr, Norm2Accumulator<10>{x}, "NORM2");
+  return GetTotalReduction<TypeCategory::Real, 10>(x, source, line, dim,
+      nullptr, Norm2AccumulatorGetter<10>::create(x), "NORM2");
 }
 #endif
 #if LDBL_MANT_DIG == 113
+// The __float128 implementation resides in FortranFloat128Math library.
 CppTypeFor<TypeCategory::Real, 16> RTDEF(Norm2_16)(
     const Descriptor &x, const char *source, int line, int dim) {
-  return GetTotalReduction<TypeCategory::Real, 16>(
-      x, source, line, dim, nullptr, Norm2Accumulator<16>{x}, "NORM2");
+  return GetTotalReduction<TypeCategory::Real, 16>(x, source, line, dim,
+      nullptr, Norm2AccumulatorGetter<16>::create(x), "NORM2");
 }
 #endif
 
@@ -901,7 +818,7 @@ void RTDEF(Norm2Dim)(Descriptor &result, const Descriptor &x, int dim,
   auto type{x.type().GetCategoryAndKind()};
   RUNTIME_CHECK(terminator, type);
   if (type->first == TypeCategory::Real) {
-    ApplyFloatingPointKind<Norm2Helper, void>(
+    ApplyFloatingPointKind<Norm2Helper, void, true>(
         type->second, terminator, result, x, dim, nullptr, terminator);
   } else {
     terminator.Crash("NORM2: bad type code %d", x.type().raw());

diff --git a/flang/runtime/reduction-templates.h b/flang/runtime/reduction-templates.h
@@ -25,6 +25,7 @@
 #include "tools.h"
 #include "flang/Runtime/cpp-type.h"
 #include "flang/Runtime/descriptor.h"
+#include <algorithm>
 
 namespace Fortran::runtime {
 
@@ -332,5 +333,119 @@ template <typename ACCUMULATOR> struct PartialLocationHelper {
   };
 };
 
+// NORM2 templates
+
+RT_VAR_GROUP_BEGIN
+
+// Use at least double precision for accumulators.
+// Don't use __float128, it doesn't work with abs() or sqrt() yet.
+static constexpr RT_CONST_VAR_ATTRS int Norm2LargestLDKind {
+#if LDBL_MANT_DIG == 113 || HAS_FLOAT128
+  16
+#elif LDBL_MANT_DIG == 64
+  10
+#else
+  8
+#endif
+};
+
+RT_VAR_GROUP_END
+
+template <TypeCategory CAT, int KIND, typename ACCUMULATOR>
+inline RT_API_ATTRS void DoMaxMinNorm2(Descriptor &result, const Descriptor &x,
+    int dim, const Descriptor *mask, const char *intrinsic,
+    Terminator &terminator) {
+  using Type = CppTypeFor<CAT, KIND>;
+  ACCUMULATOR accumulator{x};
+  if (dim == 0 || x.rank() == 1) {
+    // Total reduction
+
+    // Element size of the destination descriptor is the same
+    // as the element size of the source.
+    result.Establish(x.type(), x.ElementBytes(), nullptr, 0, nullptr,
+        CFI_attribute_allocatable);
+    if (int stat{result.Allocate()}) {
+      terminator.Crash(
+          "%s: could not allocate memory for result; STAT=%d", intrinsic, stat);
+    }
+    DoTotalReduction<Type>(x, dim, mask, accumulator, intrinsic, terminator);
+    accumulator.GetResult(result.OffsetElement<Type>());
+  } else {
+    // Partial reduction
+
+    // Element size of the destination descriptor is the same
+    // as the element size of the source.
+    PartialReduction<ACCUMULATOR, CAT, KIND>(result, x, x.ElementBytes(), dim,
+        mask, terminator, intrinsic, accumulator);
+  }
+}
+
+// The data type used by Norm2Accumulator.
+template <int KIND>
+using Norm2AccumType =
+    CppTypeFor<TypeCategory::Real, std::clamp(KIND, 8, Norm2LargestLDKind)>;
+
+template <int KIND, typename ABS, typename SQRT> class Norm2Accumulator {
+public:
+  using Type = CppTypeFor<TypeCategory::Real, KIND>;
+  using AccumType = Norm2AccumType<KIND>;
+  explicit RT_API_ATTRS Norm2Accumulator(const Descriptor &array)
+      : array_{array} {}
+  RT_API_ATTRS void Reinitialize() { max_ = sum_ = 0; }
+  template <typename A>
+  RT_API_ATTRS void GetResult(A *p, int /*zeroBasedDim*/ = -1) const {
+    // m * sqrt(1 + sum((others(:)/m)**2))
+    *p = static_cast<Type>(max_ * SQRT::compute(1 + sum_));
+  }
+  RT_API_ATTRS bool Accumulate(Type x) {
+    auto absX{ABS::compute(static_cast<AccumType>(x))};
+    if (!max_) {
+      max_ = absX;
+    } else if (absX > max_) {
+      auto t{max_ / absX}; // < 1.0
+      auto tsq{t * t};
+      sum_ *= tsq; // scale sum to reflect change to the max
+      sum_ += tsq; // include a term for the previous max
+      max_ = absX;
+    } else { // absX <= max_
+      auto t{absX / max_};
+      sum_ += t * t;
+    }
+    return true;
+  }
+  template <typename A>
+  RT_API_ATTRS bool AccumulateAt(const SubscriptValue at[]) {
+    return Accumulate(*array_.Element<A>(at));
+  }
+
+private:
+  const Descriptor &array_;
+  AccumType max_{0}; // value (m) with largest magnitude
+  AccumType sum_{0}; // sum((others(:)/m)**2)
+};
+
+// Helper class for creating Norm2Accumulator instance
+// based on the given KIND. This helper returns and instance
+// that uses std::abs and std::sqrt for the computations.
+template <int KIND> class Norm2AccumulatorGetter {
+  using AccumType = Norm2AccumType<KIND>;
+
+public:
+  struct ABSTy {
+    static constexpr RT_API_ATTRS AccumType compute(AccumType &&x) {
+      return std::abs(std::forward<AccumType>(x));
+    }
+  };
+  struct SQRTTy {
+    static constexpr RT_API_ATTRS AccumType compute(AccumType &&x) {
+      return std::sqrt(std::forward<AccumType>(x));
+    }
+  };
+
+  using Type = Norm2Accumulator<KIND, ABSTy, SQRTTy>;
+
+  static RT_API_ATTRS Type create(const Descriptor &x) { return Type(x); }
+};
+
 } // namespace Fortran::runtime
 #endif // FORTRAN_RUNTIME_REDUCTION_TEMPLATES_H_
diff --git a/flang/runtime/tools.h b/flang/runtime/tools.h
@@ -266,7 +266,8 @@ inline RT_API_ATTRS RESULT ApplyIntegerKind(
   }
 }
 
-template <template <int KIND> class FUNC, typename RESULT, typename... A>
+template <template <int KIND> class FUNC, typename RESULT,
+    bool NEEDSMATH = false, typename... A>
 inline RT_API_ATTRS RESULT ApplyFloatingPointKind(
     int kind, Terminator &terminator, A &&...x) {
   switch (kind) {
@@ -287,7 +288,13 @@ inline RT_API_ATTRS RESULT ApplyFloatingPointKind(
     break;
   case 16:
     if constexpr (HasCppTypeFor<TypeCategory::Real, 16>) {
-      return FUNC<16>{}(std::forward<A>(x)...);
+      // If FUNC implemenation relies on FP math functions,
+      // then we should not be here. The compiler should have
+      // generated a call to an entry in FortranFloat128Math
+      // library.
+      if constexpr (!NEEDSMATH) {
+        return FUNC<16>{}(std::forward<A>(x)...);
+      }
     }
     break;
   }

diff --git a/flang/test/Driver/flang-experimental-polymorphism-flag.f90 b/flang/test/Driver/flang-experimental-polymorphism-flag.f90
@@ -1,10 +1,10 @@
 ! Test -flang-experimental-hlfir flag
 ! RUN: %flang_fc1 -flang-experimental-polymorphism -emit-fir -o - %s | FileCheck %s
-! RUN: not %flang_fc1 -emit-fir -o - %s 2>&1 | FileCheck %s --check-prefix NO-POLYMORPHISM
+! RUN: %flang_fc1 -emit-fir -o - %s 2>&1 | FileCheck %s --check-prefix NO-POLYMORPHISM
 
 ! CHECK: func.func @_QPtest(%{{.*}}: !fir.class<none> {fir.bindc_name = "poly"})
 subroutine test(poly)
   class(*) :: poly
 end subroutine test
 
-! NO-POLYMORPHISM: not yet implemented: support for polymorphic types
+! NO-POLYMORPHISM: func.func @_QPtest
diff --git a/flang/test/Lower/Intrinsics/norm2.f90 b/flang/test/Lower/Intrinsics/norm2.f90
@@ -76,3 +76,19 @@ subroutine norm2_test_dim_3(a,r)
   ! CHECK-DAG:  %[[addr:.*]] = fir.box_addr %[[box]] : (!fir.box<!fir.heap<!fir.array<?x?xf32>>>) -> !fir.heap<!fir.array<?x?xf32>>
   ! CHECK-DAG:  fir.freemem %[[addr]]
 end subroutine norm2_test_dim_3
+
+! CHECK-LABEL: func @_QPnorm2_test_real16(
+! CHECK-SAME: %[[arg0:.*]]: !fir.box<!fir.array<?x?x?xf128>>{{.*}}, %[[arg1:.*]]: !fir.box<!fir.array<?x?xf128>>{{.*}})
+subroutine norm2_test_real16(a,r)
+  real(16) :: a(:,:,:)
+  real(16) :: r(:,:)
+  ! CHECK-DAG:  %[[dim:.*]] = arith.constant 3 : i32
+  ! CHECK-DAG:  %[[r:.*]] = fir.alloca !fir.box<!fir.heap<!fir.array<?x?xf128>>>
+  ! CHECK-DAG:  %[[res:.*]] = fir.convert %[[r]] : (!fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf128>>>>) -> !fir.ref<!fir.box<none>>
+  ! CHECK:  %[[arr:.*]] = fir.convert %[[arg0]] : (!fir.box<!fir.array<?x?x?xf128>>) -> !fir.box<none>
+  r = norm2(a,dim=3)
+  ! CHECK:  %{{.*}} = fir.call @_FortranANorm2DimReal16(%[[res]], %[[arr]], %[[dim]], %{{.*}}, %{{.*}}) {{.*}} : (!fir.ref<!fir.box<none>>, !fir.box<none>, i32, !fir.ref<i8>, i32) -> none
+  ! CHECK:  %[[box:.*]] = fir.load %[[r]] : !fir.ref<!fir.box<!fir.heap<!fir.array<?x?xf128>>>>
+  ! CHECK-DAG:  %[[addr:.*]] = fir.box_addr %[[box]] : (!fir.box<!fir.heap<!fir.array<?x?xf128>>>) -> !fir.heap<!fir.array<?x?xf128>>
+  ! CHECK-DAG:  fir.freemem %[[addr]]
+end subroutine norm2_test_real16
diff --git a/libc/.clang-tidy b/libc/.clang-tidy
@@ -26,5 +26,7 @@ CheckOptions:
     value:           UPPER_CASE
   - key:             readability-identifier-naming.ConstexprVariableCase
     value:           UPPER_CASE
+  - key:             readability-identifier-naming.ConstexprFunctionCase
+    value:           lower_case
   - key:             readability-identifier-naming.GetConfigPerFile
     value:           true
diff --git a/libc/cmake/modules/compiler_features/check_float128.cpp b/libc/cmake/modules/compiler_features/check_float128.cpp
@@ -1,4 +1,4 @@
-#include "src/__support/macros/properties/float.h"
+#include "src/__support/macros/properties/types.h"
 
 #ifndef LIBC_COMPILER_HAS_FLOAT128
 #error unsupported

diff --git a/libc/config/baremetal/api.td b/libc/config/baremetal/api.td
@@ -2,6 +2,42 @@ include "config/public_api.td"
 
 include "spec/stdc.td"
 
+def AssertMacro : MacroDef<"assert"> {
+  let Defn = [{
+    #undef assert
+
+    #ifdef NDEBUG
+    #define assert(e) (void)0
+    #else
+
+    #ifdef __cplusplus
+    extern "C"
+    #endif
+    _Noreturn void __assert_fail(const char *, const char *, unsigned, const char *) __NOEXCEPT;
+
+    #define assert(e)  \
+      ((e) ? (void)0 : __assert_fail(#e, __FILE__, __LINE__, __PRETTY_FUNCTION__))
+
+    #endif
+  }];
+}
+
+def StaticAssertMacro : MacroDef<"static_assert"> {
+  let Defn = [{
+    #ifndef __cplusplus
+    #undef static_assert
+    #define static_assert _Static_assert
+    #endif
+  }];
+}
+
+def AssertAPI : PublicAPI<"assert.h"> {
+  let Macros = [
+    AssertMacro,
+    StaticAssertMacro,
+  ];
+}
+
 def CTypeAPI : PublicAPI<"ctype.h"> {
 }
 

diff --git a/libc/config/baremetal/arm/entrypoints.txt b/libc/config/baremetal/arm/entrypoints.txt
@@ -1,4 +1,7 @@
 set(TARGET_LIBC_ENTRYPOINTS
+    # assert.h entrypoints
+    libc.src.assert.__assert_fail
+
     # ctype.h entrypoints
     libc.src.ctype.isalnum
     libc.src.ctype.isalpha

diff --git a/libc/config/baremetal/arm/headers.txt b/libc/config/baremetal/arm/headers.txt
@@ -1,4 +1,5 @@
 set(TARGET_PUBLIC_HEADERS
+    libc.include.assert
     libc.include.ctype
     libc.include.fenv
     libc.include.errno

diff --git a/libc/config/baremetal/riscv/entrypoints.txt b/libc/config/baremetal/riscv/entrypoints.txt
@@ -1,4 +1,7 @@
 set(TARGET_LIBC_ENTRYPOINTS
+    # assert.h entrypoints
+    libc.src.assert.__assert_fail
+
     # ctype.h entrypoints
     libc.src.ctype.isalnum
     libc.src.ctype.isalpha

diff --git a/libc/config/baremetal/riscv/headers.txt b/libc/config/baremetal/riscv/headers.txt
@@ -1,4 +1,5 @@
 set(TARGET_PUBLIC_HEADERS
+    libc.include.assert
     libc.include.ctype
     libc.include.fenv
     libc.include.errno

diff --git a/libc/docs/dev/code_style.rst b/libc/docs/dev/code_style.rst
@@ -47,7 +47,7 @@ We define two kinds of macros:
        e.g., ``LIBC_COMPILER_IS_CLANG``.
      * ``cpu_features.h`` - Target cpu feature availability.
        e.g., ``LIBC_TARGET_CPU_HAS_AVX2``.
-     * ``float.h`` - Floating point type properties and availability.
+     * ``types.h`` - Type properties and availability.
        e.g., ``LIBC_COMPILER_HAS_FLOAT128``.
      * ``os.h`` - Target os properties.
        e.g., ``LIBC_TARGET_OS_IS_LINUX``.

diff --git a/libc/include/__llvm-libc-common.h b/libc/include/__llvm-libc-common.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_LIBC___COMMON_H
-#define LLVM_LIBC___COMMON_H
+#ifndef LLVM_LIBC_COMMON_H
+#define LLVM_LIBC_COMMON_H
 
 #ifdef __cplusplus
 
@@ -51,4 +51,4 @@
 
 #endif // __cplusplus
 
-#endif // LLVM_LIBC___COMMON_H
+#endif // LLVM_LIBC_COMMON_H
diff --git a/libc/include/llvm-libc-macros/containerof-macro.h b/libc/include/llvm-libc-macros/containerof-macro.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_CONTAINEROF_MACRO_H
-#define __LLVM_LIBC_MACROS_CONTAINEROF_MACRO_H
+#ifndef LLVM_LIBC_MACROS_CONTAINEROF_MACRO_H
+#define LLVM_LIBC_MACROS_CONTAINEROF_MACRO_H
 
 #include <llvm-libc-macros/offsetof-macro.h>
 
@@ -17,4 +17,4 @@
     (type *)(void *)((const char *)__ptr - offsetof(type, member));            \
   })
 
-#endif // __LLVM_LIBC_MACROS_CONTAINEROF_MACRO_H
+#endif // LLVM_LIBC_MACROS_CONTAINEROF_MACRO_H
diff --git a/libc/include/llvm-libc-macros/fcntl-macros.h b/libc/include/llvm-libc-macros/fcntl-macros.h
@@ -1,8 +1,8 @@
-#ifndef __LLVM_LIBC_MACROS_FCNTL_MACROS_H
-#define __LLVM_LIBC_MACROS_FCNTL_MACROS_H
+#ifndef LLVM_LIBC_MACROS_FCNTL_MACROS_H
+#define LLVM_LIBC_MACROS_FCNTL_MACROS_H
 
 #ifdef __linux__
 #include "linux/fcntl-macros.h"
 #endif
 
-#endif // __LLVM_LIBC_MACROS_FCNTL_MACROS_H
+#endif // LLVM_LIBC_MACROS_FCNTL_MACROS_H
diff --git a/libc/include/llvm-libc-macros/features-macros.h b/libc/include/llvm-libc-macros/features-macros.h
@@ -6,9 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_FEATURES_MACROS_H
-#define __LLVM_LIBC_MACROS_FEATURES_MACROS_H
+#ifndef LLVM_LIBC_MACROS_FEATURES_MACROS_H
+#define LLVM_LIBC_MACROS_FEATURES_MACROS_H
 
 #define __LLVM_LIBC__ 1
 
-#endif // __LLVM_LIBC_MACROS_FEATURES_MACROS_H
+#endif // LLVM_LIBC_MACROS_FEATURES_MACROS_H
diff --git a/libc/include/llvm-libc-macros/fenv-macros.h b/libc/include/llvm-libc-macros/fenv-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_FENV_MACROS_H
-#define __LLVM_LIBC_MACROS_FENV_MACROS_H
+#ifndef LLVM_LIBC_MACROS_FENV_MACROS_H
+#define LLVM_LIBC_MACROS_FENV_MACROS_H
 
 #define FE_DIVBYZERO 1
 #define FE_INEXACT 2
@@ -24,4 +24,4 @@
 
 #define FE_DFL_ENV ((fenv_t *)-1)
 
-#endif // __LLVM_LIBC_MACROS_FENV_MACROS_H
+#endif // LLVM_LIBC_MACROS_FENV_MACROS_H
diff --git a/libc/include/llvm-libc-macros/file-seek-macros.h b/libc/include/llvm-libc-macros/file-seek-macros.h
@@ -6,11 +6,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_FILE_SEEK_MACROS_H
-#define __LLVM_LIBC_MACROS_FILE_SEEK_MACROS_H
+#ifndef LLVM_LIBC_MACROS_FILE_SEEK_MACROS_H
+#define LLVM_LIBC_MACROS_FILE_SEEK_MACROS_H
 
 #define SEEK_SET 0
 #define SEEK_CUR 1
 #define SEEK_END 2
 
-#endif // __LLVM_LIBC_MACROS_FILE_SEEK_MACROS_H
+#endif // LLVM_LIBC_MACROS_FILE_SEEK_MACROS_H
diff --git a/libc/include/llvm-libc-macros/float-macros.h b/libc/include/llvm-libc-macros/float-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_FLOAT_MACROS_H
-#define __LLVM_LIBC_MACROS_FLOAT_MACROS_H
+#ifndef LLVM_LIBC_MACROS_FLOAT_MACROS_H
+#define LLVM_LIBC_MACROS_FLOAT_MACROS_H
 
 // Suppress `#include_next is a language extension` warnings.
 #ifdef __clang__
@@ -169,4 +169,4 @@
 
 // TODO: Add FLT16 and FLT128 constants.
 
-#endif // __LLVM_LIBC_MACROS_FLOAT_MACROS_H
+#endif // LLVM_LIBC_MACROS_FLOAT_MACROS_H
diff --git a/libc/include/llvm-libc-macros/generic-error-number-macros.h b/libc/include/llvm-libc-macros/generic-error-number-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_GENERIC_ERROR_NUMBER_MACROS_H
-#define __LLVM_LIBC_MACROS_GENERIC_ERROR_NUMBER_MACROS_H
+#ifndef LLVM_LIBC_MACROS_GENERIC_ERROR_NUMBER_MACROS_H
+#define LLVM_LIBC_MACROS_GENERIC_ERROR_NUMBER_MACROS_H
 
 #define EPERM 1
 #define ENOENT 2
@@ -45,4 +45,4 @@
 #define ERANGE 34
 #define EILSEQ 35
 
-#endif // __LLVM_LIBC_MACROS_GENERIC_ERROR_NUMBER_MACROS_H
+#endif // LLVM_LIBC_MACROS_GENERIC_ERROR_NUMBER_MACROS_H
diff --git a/libc/include/llvm-libc-macros/gpu/time-macros.h b/libc/include/llvm-libc-macros/gpu/time-macros.h
@@ -6,9 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_GPU_TIME_MACROS_H
-#define __LLVM_LIBC_MACROS_GPU_TIME_MACROS_H
+#ifndef LLVM_LIBC_MACROS_GPU_TIME_MACROS_H
+#define LLVM_LIBC_MACROS_GPU_TIME_MACROS_H
 
 #define CLOCKS_PER_SEC 1000000
 
-#endif // __LLVM_LIBC_MACROS_GPU_TIME_MACROS_H
+#endif // LLVM_LIBC_MACROS_GPU_TIME_MACROS_H
diff --git a/libc/include/llvm-libc-macros/inttypes-macros.h b/libc/include/llvm-libc-macros/inttypes-macros.h
@@ -5,8 +5,8 @@
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
-#ifndef __LLVM_LIBC_MACROS_INTTYPES_MACROS_H
-#define __LLVM_LIBC_MACROS_INTTYPES_MACROS_H
+#ifndef LLVM_LIBC_MACROS_INTTYPES_MACROS_H
+#define LLVM_LIBC_MACROS_INTTYPES_MACROS_H
 
 // fprintf/scanf format macros.
 // POSIX.1-2008, Technical Corrigendum 1, XBD/TC1-2008/0050 [211] is applied.
@@ -286,4 +286,4 @@
 #define SCNxMAX __UINTMAX_FMTx__
 #define SCNxPTR __UINTPTR_FMTx__
 
-#endif // __LLVM_LIBC_MACROS_INTTYPES_MACROS_H
+#endif // LLVM_LIBC_MACROS_INTTYPES_MACROS_H
diff --git a/libc/include/llvm-libc-macros/limits-macros.h b/libc/include/llvm-libc-macros/limits-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LIMITS_MACROS_H
-#define __LLVM_LIBC_MACROS_LIMITS_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LIMITS_MACROS_H
+#define LLVM_LIBC_MACROS_LIMITS_MACROS_H
 
 // Define all C23 macro constants of limits.h
 
@@ -225,4 +225,4 @@
 #define ULLONG_MIN 0ULL
 #endif // ULLONG_MIN
 
-#endif // __LLVM_LIBC_MACROS_LIMITS_MACROS_H
+#endif // LLVM_LIBC_MACROS_LIMITS_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/fcntl-macros.h b/libc/include/llvm-libc-macros/linux/fcntl-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_FCNTL_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_FCNTL_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_FCNTL_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_FCNTL_MACROS_H
 
 // File creation flags
 #define O_CLOEXEC 02000000
@@ -68,4 +68,4 @@
 #define F_GETFL 3
 #define F_SETFL 4
 
-#endif // __LLVM_LIBC_MACROS_LINUX_FCNTL_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_FCNTL_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sched-macros.h b/libc/include/llvm-libc-macros/linux/sched-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SCHED_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SCHED_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SCHED_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SCHED_MACROS_H
 
 // Definitions of SCHED_* macros must match was linux as at:
 // https://elixir.bootlin.com/linux/latest/source/include/uapi/linux/sched.h
@@ -26,4 +26,4 @@
 #define CPU_COUNT_S(setsize, set) __sched_getcpucount(setsize, set)
 #define CPU_COUNT(set) CPU_COUNT_S(sizeof(cpu_set_t), set)
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SCHED_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SCHED_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/signal-macros.h b/libc/include/llvm-libc-macros/linux/signal-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SIGNUM_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SIGNUM_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SIGNAL_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SIGNAL_MACROS_H
 
 #define SIGHUP 1
 #define SIGINT 2
@@ -101,4 +101,4 @@
 #define CLD_STOPPED 5   // child has stopped
 #define CLD_CONTINUED 6 // stopped child has continued
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SIGNUM_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SIGNAL_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sys-ioctl-macros.h b/libc/include/llvm-libc-macros/linux/sys-ioctl-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SYS_IOCTL_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SYS_IOCTL_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_IOCTL_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_IOCTL_MACROS_H
 
 // TODO (michaelrj): Finish defining these macros.
 // Just defining this macro for the moment since it's all that we need right
@@ -16,4 +16,4 @@
 // think is worth digging into right now.
 #define TIOCGETD 0x5424
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SYS_IOCTL_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_IOCTL_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sys-random-macros.h b/libc/include/llvm-libc-macros/linux/sys-random-macros.h
@@ -6,12 +6,12 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SYS_RANDOM_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SYS_RANDOM_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_RANDOM_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_RANDOM_MACROS_H
 
 // Getrandom flags
 #define GRND_RANDOM 0x0001
 #define GRND_NONBLOCK 0x0002
 #define GRND_INSECURE 0x0004
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SYS_RANDOM_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_RANDOM_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sys-resource-macros.h b/libc/include/llvm-libc-macros/linux/sys-resource-macros.h
@@ -6,6 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_RESOURCE_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_RESOURCE_MACROS_H
+
 #define RLIMIT_CPU 0
 #define RLIMIT_FSIZE 1
 #define RLIMIT_DATA 2
@@ -24,3 +27,5 @@
 #define RLIMIT_RTTIME 15
 
 #define RLIM_INFINITY (~0UL)
+
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_RESOURCE_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sys-socket-macros.h b/libc/include/llvm-libc-macros/linux/sys-socket-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SYS_SOCKET_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SYS_SOCKET_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_SOCKET_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_SOCKET_MACROS_H
 
 // IEEE Std 1003.1-2017 - basedefs/sys_socket.h.html
 // Macro values come from the Linux syscall interface.
@@ -25,4 +25,4 @@
 #define SOCK_SEQPACKET 5
 #define SOCK_PACKET 10
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SYS_SOCKET_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_SOCKET_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sys-stat-macros.h b/libc/include/llvm-libc-macros/linux/sys-stat-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SYS_STAT_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SYS_STAT_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_STAT_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_STAT_MACROS_H
 
 // Definitions from linux/stat.h
 #define S_IFMT 0170000
@@ -45,4 +45,4 @@
 #define S_IWOTH 00002
 #define S_IXOTH 00001
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SYS_STAT_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_STAT_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sys-time-macros.h b/libc/include/llvm-libc-macros/linux/sys-time-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SYS_TIME_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SYS_TIME_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_TIME_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_TIME_MACROS_H
 
 // Add two timevals and put the result in timeval_ptr_result. If the resulting
 // usec value is greater than 999,999 then the microseconds are turned into full
@@ -50,4 +50,4 @@
        ? ((timeval_ptr_a)->tv_usec CMP(timeval_ptr_b)->tv_usec)                \
        : ((timeval_ptr_a)->tv_sec CMP(timeval_ptr_b)->tv_sec))
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SYS_TIME_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_TIME_MACROS_H
diff --git a/libc/include/llvm-libc-macros/linux/sys-wait-macros.h b/libc/include/llvm-libc-macros/linux/sys-wait-macros.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef __LLVM_LIBC_MACROS_LINUX_SYS_WAIT_MACROS_H
-#define __LLVM_LIBC_MACROS_LINUX_SYS_WAIT_MACROS_H
+#ifndef LLVM_LIBC_MACROS_LINUX_SYS_WAIT_MACROS_H
+#define LLVM_LIBC_MACROS_LINUX_SYS_WAIT_MACROS_H
 
 // Wait flags
 #define WNOHANG 1    // Do not block
@@ -41,4 +41,4 @@
 #define P_PGID 2
 #define P_PIDFD 3
 
-#endif // __LLVM_LIBC_MACROS_LINUX_SYS_WAIT_MACROS_H
+#endif // LLVM_LIBC_MACROS_LINUX_SYS_WAIT_MACROS_H
-Original file line number
+Diff line change
@@ Expand Up / @@ -903,7 +903,7 @@ void CudaToolChain::addClangTargetOptions( @@
           return;
         addOpenMPDeviceRTL(getDriver(), DriverArgs, CC1Args, GpuArch.str(),
-                           getTriple());
+                           getTriple(), HostTC);
       }
     }
@@ Expand Down @@