Merge branch 'llvm:main' into llvmgh-101657

.ci/metrics/metrics.py

@@ -12,7 +12,7 @@
     "https://influx-prod-13-prod-us-east-0.grafana.net/api/v1/push/influx/write"
 )
 GITHUB_PROJECT = "llvm/llvm-project"
-WORKFLOWS_TO_TRACK = ["Check code formatting", "LLVM Premerge Checks"]
+WORKFLOWS_TO_TRACK = ["LLVM Premerge Checks"]
 SCRAPE_INTERVAL_SECONDS = 5 * 60
 
 

clang-tools-extra/clang-tidy/bugprone/NarrowingConversionsCheck.cpp

@@ -513,7 +513,9 @@ void NarrowingConversionsCheck::handleFloatingCast(const ASTContext &Context,
       return;
     }
     const BuiltinType *FromType = getBuiltinType(Rhs);
-    if (ToType->getKind() < FromType->getKind())
+    if (!llvm::APFloatBase::isRepresentableBy(
+            Context.getFloatTypeSemantics(FromType->desugar()),
+            Context.getFloatTypeSemantics(ToType->desugar())))
       diagNarrowType(SourceLoc, Lhs, Rhs);
   }
 }

clang-tools-extra/clang-tidy/modernize/UseAutoCheck.cpp

@@ -342,7 +342,7 @@ static void ignoreTypeLocClasses(
     Loc = Loc.getNextTypeLoc();
 }
 
-static bool isMutliLevelPointerToTypeLocClasses(
+static bool isMultiLevelPointerToTypeLocClasses(
     TypeLoc Loc,
     std::initializer_list<TypeLoc::TypeLocClass> const &LocClasses) {
   ignoreTypeLocClasses(Loc, {TypeLoc::Paren, TypeLoc::Qualified});
@@ -424,7 +424,7 @@ void UseAutoCheck::replaceExpr(
 
   auto Diag = diag(Range.getBegin(), Message);
 
-  bool ShouldReplenishVariableName = isMutliLevelPointerToTypeLocClasses(
+  bool ShouldReplenishVariableName = isMultiLevelPointerToTypeLocClasses(
       TSI->getTypeLoc(), {TypeLoc::FunctionProto, TypeLoc::ConstantArray});
 
   // Space after 'auto' to handle cases where the '*' in the pointer type is

clang-tools-extra/docs/ReleaseNotes.rst

@@ -210,6 +210,11 @@ Changes in existing checks
   <clang-tidy/checks/bugprone/forwarding-reference-overload>` check by fixing
   a crash when determining if an ``enable_if[_t]`` was found.
 
+- Improve :doc:`bugprone-narrowing-conversions
+  <clang-tidy/checks/bugprone/narrowing-conversions>` to avoid incorrect check
+  results when floating point type is not ``float``, ``double`` and
+  ``long double``.
+
 - Improved :doc:`bugprone-optional-value-conversion
   <clang-tidy/checks/bugprone/optional-value-conversion>` to support detecting
   conversion directly by ``std::make_unique`` and ``std::make_shared``.
@@ -346,8 +351,8 @@ Changes in existing checks
   <clang-tidy/checks/performance/move-const-arg>` check to fix a crash when
   an argument type is declared but not defined.
 
-- Improved :doc:`performance-unnecessary-copy-initialization`
-  <clang-tidy/checks/performance/unnecessary-copy-initialization> check
+- Improved :doc:`performance-unnecessary-copy-initialization
+  <clang-tidy/checks/performance/unnecessary-copy-initialization>` check
   to consider static member functions the same way as free functions.
 
 - Improved :doc:`readability-container-contains

clang-tools-extra/test/clang-tidy/checkers/bugprone/narrowing-conversions-narrowingfloatingpoint-option.cpp

@@ -36,6 +36,15 @@ void narrow_double_to_float_not_ok(double d) {
   f = narrow_double_to_float_return();
 }
 
+float narrow_float16_to_float_return(_Float16 f) {
+  return f;
+}
+
+_Float16 narrow_float_to_float16_return(float f) {
+  return f;
+  // CHECK-MESSAGES: :[[@LINE-1]]:10: warning: narrowing conversion from 'float' to '_Float16' [bugprone-narrowing-conversions]
+}
+
 void narrow_fp_constants() {
   float f;
   f = 0.5; // [dcl.init.list] 7.2 : in-range fp constant to narrower float is not a narrowing.

clang/docs/ReleaseNotes.rst

@@ -1068,11 +1068,23 @@ X86 Support
 Arm and AArch64 Support
 ^^^^^^^^^^^^^^^^^^^^^^^
 
+- Implementation of SVE2.1 and SME2.1 in accordance with the Arm C Language
+  Extensions (ACLE) is now available.
+
 - In the ARM Target, the frame pointer (FP) of a leaf function can be retained
   by using the ``-fno-omit-frame-pointer`` option. If you want to eliminate the FP
   in leaf functions after enabling ``-fno-omit-frame-pointer``, you can do so by adding
   the ``-momit-leaf-frame-pointer`` option.
 
+- SME keyword attributes which apply to function types are now represented in the
+  mangling of the type. This means that ``void foo(void (*f)() __arm_streaming);``
+  now has a different mangling from ``void foo(void (*f)());``.
+
+- The ``__arm_agnostic`` keyword attribute was added to let users describe
+  a function that preserves SME state enabled by PSTATE.ZA without having to share
+  this state with its callers and without making the assumption that this state
+  exists.
+
 - Support has been added for the following processors (-mcpu identifiers in parenthesis):
 
   For AArch64:

clang/include/clang/AST/DeclBase.h

@@ -1334,7 +1334,7 @@ class DeclListNode {
 
     reference operator*() const {
       assert(Ptr && "dereferencing end() iterator");
-      if (DeclListNode *CurNode = Ptr.dyn_cast<DeclListNode*>())
+      if (DeclListNode *CurNode = dyn_cast<DeclListNode *>(Ptr))
         return CurNode->D;
       return cast<NamedDecl *>(Ptr);
     }
@@ -1344,7 +1344,7 @@ class DeclListNode {
     inline iterator &operator++() { // ++It
       assert(!Ptr.isNull() && "Advancing empty iterator");
 
-      if (DeclListNode *CurNode = Ptr.dyn_cast<DeclListNode*>())
+      if (DeclListNode *CurNode = dyn_cast<DeclListNode *>(Ptr))
         Ptr = CurNode->Rest;
       else
         Ptr = nullptr;

clang/include/clang/Basic/Builtins.td

@@ -3347,10 +3347,12 @@ def VFork : LibBuiltin<"unistd.h"> {
 }
 
 // POSIX pthread.h
-// FIXME: This should be a GNULibBuiltin, but it's currently missing the prototype.
 
-def PthreadCreate : CustomEntry {
-  let Entry = "LIBBUILTIN(pthread_create, \"\",  \"fC<2,3>\", PTHREAD_H, ALL_GNU_LANGUAGES)";
+def PthreadCreate : GNULibBuiltin<"pthread.h"> {
+  let Spellings = ["pthread_create"];
+  let Attributes = [FunctionWithoutBuiltinPrefix, Callback<[2, 3]>];
+  // Note that we don't have an expressable prototype so we leave it empty.
+  let Prototype = "";
 }
 
 def SigSetJmp : LibBuiltin<"setjmp.h"> {

clang/include/clang/Basic/BuiltinsBase.td

@@ -17,6 +17,11 @@ class IndexedAttribute<string baseMangling, int I> : Attribute<baseMangling> {
   int Index = I;
 }
 
+class MultiIndexAttribute<string baseMangling, list<int> Is>
+    : Attribute<baseMangling> {
+  list<int> Indices = Is;
+}
+
 // Standard Attributes
 // -------------------
 def NoReturn : Attribute<"r">;
@@ -77,6 +82,10 @@ def Constexpr : Attribute<"E">;
 // Builtin is immediate and must be constant evaluated. Implies Constexpr, and will only be supported in C++20 mode.
 def Consteval : Attribute<"EG">;
 
+// Callback behavior: the first index argument is called with the arguments
+// indicated by the remaining indices.
+class Callback<list<int> ArgIndices> : MultiIndexAttribute<"C", ArgIndices>;
+
 // Builtin kinds
 // =============
 
@@ -92,10 +101,6 @@ class Builtin {
   bit EnableOpenCLLong = 0;
 }
 
-class CustomEntry {
-  string Entry;
-}
-
 class AtomicBuiltin : Builtin;
 
 class LibBuiltin<string header, string languages = "ALL_LANGUAGES"> : Builtin {

clang/include/clang/Driver/Options.td

@@ -4108,7 +4108,7 @@ defm ms_tls_guards : BoolFOption<"ms-tls-guards",
           "Do not emit code to perform on-demand initialization of thread-local variables">,
   PosFlag<SetTrue>>;
 def ftime_report : Flag<["-"], "ftime-report">, Group<f_Group>,
-  Visibility<[ClangOption, CC1Option]>,
+  Visibility<[ClangOption, CC1Option, FlangOption, FC1Option]>,
   MarshallingInfoFlag<CodeGenOpts<"TimePasses">>;
 def ftime_report_EQ: Joined<["-"], "ftime-report=">, Group<f_Group>,
   Visibility<[ClangOption, CC1Option]>, Values<"per-pass,per-pass-run">,
@@ -4364,7 +4364,7 @@ defm split_machine_functions: BoolFOption<"split-machine-functions",
   CodeGenOpts<"SplitMachineFunctions">, DefaultFalse,
   PosFlag<SetTrue, [], [ClangOption, CC1Option], "Enable">,
   NegFlag<SetFalse, [], [ClangOption], "Disable">,
-  BothFlags<[], [ClangOption], " late function splitting using profile information (x86 ELF)">>;
+  BothFlags<[], [ClangOption], " late function splitting using profile information (x86 and aarch64 ELF)">>;
 
 defm strict_return : BoolFOption<"strict-return",
   CodeGenOpts<"StrictReturn">, DefaultTrue,

clang/lib/AST/ByteCode/Compiler.cpp

@@ -90,6 +90,10 @@ bool InitLink::emit(Compiler<Emitter> *Ctx, const Expr *E) const {
     if (!Ctx->emitConstUint32(Offset, E))
       return false;
     return Ctx->emitArrayElemPtrPopUint32(E);
+  case K_RVO:
+    return Ctx->emitRVOPtr(E);
+  case K_InitList:
+    return true;
   default:
     llvm_unreachable("Unhandled InitLink kind");
   }
@@ -1717,6 +1721,8 @@ bool Compiler<Emitter>::VisitArraySubscriptExpr(const ArraySubscriptExpr *E) {
 template <class Emitter>
 bool Compiler<Emitter>::visitInitList(ArrayRef<const Expr *> Inits,
                                       const Expr *ArrayFiller, const Expr *E) {
+  InitLinkScope<Emitter> ILS(this, InitLink::InitList());
+
   QualType QT = E->getType();
   if (const auto *AT = QT->getAs<AtomicType>())
     QT = AT->getValueType();
@@ -1754,6 +1760,7 @@ bool Compiler<Emitter>::visitInitList(ArrayRef<const Expr *> Inits,
     auto initPrimitiveField = [=](const Record::Field *FieldToInit,
                                   const Expr *Init, PrimType T) -> bool {
       InitStackScope<Emitter> ISS(this, isa<CXXDefaultInitExpr>(Init));
+      InitLinkScope<Emitter> ILS(this, InitLink::Field(FieldToInit->Offset));
       if (!this->visit(Init))
         return false;
 
@@ -1766,6 +1773,7 @@ bool Compiler<Emitter>::visitInitList(ArrayRef<const Expr *> Inits,
                                   const Expr *Init) -> bool {
       InitStackScope<Emitter> ISS(this, isa<CXXDefaultInitExpr>(Init));
       InitLinkScope<Emitter> ILS(this, InitLink::Field(FieldToInit->Offset));
+
       // Non-primitive case. Get a pointer to the field-to-initialize
       // on the stack and recurse into visitInitializer().
       if (!this->emitGetPtrField(FieldToInit->Offset, Init))
@@ -3812,6 +3820,7 @@ template <class Emitter> bool Compiler<Emitter>::visit(const Expr *E) {
 
     if (!this->emitGetPtrLocal(*LocalIndex, E))
       return false;
+    InitLinkScope<Emitter> ILS(this, InitLink::Temp(*LocalIndex));
     return this->visitInitializer(E);
   }
 
@@ -4848,18 +4857,49 @@ bool Compiler<Emitter>::VisitCXXThisExpr(const CXXThisExpr *E) {
   // instance pointer of the current function frame, but e.g. to the declaration
   // currently being initialized. Here we emit the necessary instruction(s) for
   // this scenario.
-  if (!InitStackActive || !E->isImplicit())
+  if (!InitStackActive)
     return this->emitThis(E);
 
-  if (InitStackActive && !InitStack.empty()) {
+  if (!InitStack.empty()) {
+    // If our init stack is, for example:
+    // 0 Stack: 3 (decl)
+    // 1 Stack: 6 (init list)
+    // 2 Stack: 1 (field)
+    // 3 Stack: 6 (init list)
+    // 4 Stack: 1 (field)
+    //
+    // We want to find the LAST element in it that's an init list,
+    // which is marked with the K_InitList marker. The index right
+    // before that points to an init list. We need to find the
+    // elements before the K_InitList element that point to a base
+    // (e.g. a decl or This), optionally followed by field, elem, etc.
+    // In the example above, we want to emit elements [0..2].
     unsigned StartIndex = 0;
+    unsigned EndIndex = 0;
+    // Find the init list.
     for (StartIndex = InitStack.size() - 1; StartIndex > 0; --StartIndex) {
+      if (InitStack[StartIndex].Kind == InitLink::K_InitList ||
+          InitStack[StartIndex].Kind == InitLink::K_This) {
+        EndIndex = StartIndex;
+        --StartIndex;
+        break;
+      }
+    }
+
+    // Walk backwards to find the base.
+    for (; StartIndex > 0; --StartIndex) {
+      if (InitStack[StartIndex].Kind == InitLink::K_InitList)
+        continue;
+
       if (InitStack[StartIndex].Kind != InitLink::K_Field &&
           InitStack[StartIndex].Kind != InitLink::K_Elem)
         break;
     }
 
-    for (unsigned I = StartIndex, N = InitStack.size(); I != N; ++I) {
+    // Emit the instructions.
+    for (unsigned I = StartIndex; I != EndIndex; ++I) {
+      if (InitStack[I].Kind == InitLink::K_InitList)
+        continue;
       if (!InitStack[I].template emit<Emitter>(this, E))
         return false;
     }
@@ -4960,6 +5000,7 @@ bool Compiler<Emitter>::visitReturnStmt(const ReturnStmt *RS) {
       if (!this->visit(RE))
         return false;
     } else {
+      InitLinkScope<Emitter> ILS(this, InitLink::RVO());
       // RVO - construct the value in the return location.
       if (!this->emitRVOPtr(RE))
         return false;

clang/lib/AST/ByteCode/Compiler.h

@@ -51,9 +51,13 @@ struct InitLink {
     K_Temp = 2,
     K_Decl = 3,
     K_Elem = 5,
+    K_RVO = 6,
+    K_InitList = 7
   };
 
   static InitLink This() { return InitLink{K_This}; }
+  static InitLink InitList() { return InitLink{K_InitList}; }
+  static InitLink RVO() { return InitLink{K_RVO}; }
   static InitLink Field(unsigned Offset) {
     InitLink IL{K_Field};
     IL.Offset = Offset;

clang/lib/AST/ByteCode/EvalEmitter.cpp

@@ -71,6 +71,18 @@ EvaluationResult EvalEmitter::interpretDecl(const VarDecl *VD,
   if (!this->visitDeclAndReturn(VD, S.inConstantContext()))
     EvalResult.setInvalid();
 
+  // Mark global variables as invalid if any diagnostic was produced.
+  if (S.hasPriorDiagnostic() && Context::shouldBeGloballyIndexed(VD)) {
+    if (auto GlobalIndex = P.getGlobal(VD)) {
+      Block *GlobalBlock = P.getGlobal(*GlobalIndex);
+      GlobalInlineDescriptor &GD =
+          *reinterpret_cast<GlobalInlineDescriptor *>(GlobalBlock->rawData());
+      GD.InitState = GlobalInitState::InitializerFailed;
+      if (GlobalBlock->isInitialized())
+        GlobalBlock->invokeDtor();
+    }
+  }
+
   S.EvaluatingDecl = nullptr;
   updateGlobalTemporaries();
   return std::move(this->EvalResult);

clang/lib/Analysis/ThreadSafetyCommon.cpp

@@ -336,7 +336,7 @@ til::SExpr *SExprBuilder::translateDeclRefExpr(const DeclRefExpr *DRE,
               : (cast<ObjCMethodDecl>(D)->getCanonicalDecl() == Canonical)) {
         // Substitute call arguments for references to function parameters
         if (const Expr *const *FunArgs =
-                Ctx->FunArgs.dyn_cast<const Expr *const *>()) {
+                dyn_cast<const Expr *const *>(Ctx->FunArgs)) {
           assert(I < Ctx->NumArgs);
           return translate(FunArgs[I], Ctx->Prev);
         }

clang/lib/Analysis/UnsafeBufferUsage.cpp

@@ -453,8 +453,11 @@ AST_MATCHER(ArraySubscriptExpr, isSafeArraySubscript) {
     return false;
   }
 
-  if (const auto *IdxLit = dyn_cast<IntegerLiteral>(Node.getIdx())) {
-    const APInt ArrIdx = IdxLit->getValue();
+  Expr::EvalResult EVResult;
+  if (Node.getIdx()->EvaluateAsInt(EVResult, Finder->getASTContext())) {
+    llvm::APSInt ArrIdx = EVResult.Val.getInt();
+    // FIXME: ArrIdx.isNegative() we could immediately emit an error as that's a
+    // bug
     if (ArrIdx.isNonNegative() && ArrIdx.getLimitedValue() < limit)
       return true;
   }

clang/lib/Basic/SourceManager.cpp

@@ -1222,7 +1222,7 @@ unsigned SourceManager::getPresumedColumnNumber(SourceLocation Loc,
   return PLoc.getColumn();
 }
 
-// Check if mutli-byte word x has bytes between m and n, included. This may also
+// Check if multi-byte word x has bytes between m and n, included. This may also
 // catch bytes equal to n + 1.
 // The returned value holds a 0x80 at each byte position that holds a match.
 // see http://graphics.stanford.edu/~seander/bithacks.html#HasBetweenInWord

clang/lib/CodeGen/CGCall.cpp

@@ -4507,7 +4507,7 @@ void CodeGenFunction::EmitCallArgs(
   // First, if a prototype was provided, use those argument types.
   bool IsVariadic = false;
   if (Prototype.P) {
-    const auto *MD = Prototype.P.dyn_cast<const ObjCMethodDecl *>();
+    const auto *MD = dyn_cast<const ObjCMethodDecl *>(Prototype.P);
     if (MD) {
       IsVariadic = MD->isVariadic();
       ExplicitCC = getCallingConventionForDecl(

clang/lib/CodeGen/CGOpenMPRuntime.cpp

@@ -5880,10 +5880,13 @@ void CGOpenMPRuntime::emitUsesAllocatorsFini(CodeGenFunction &CGF,
 
 void CGOpenMPRuntime::computeMinAndMaxThreadsAndTeams(
     const OMPExecutableDirective &D, CodeGenFunction &CGF,
-    int32_t &MinThreadsVal, int32_t &MaxThreadsVal, int32_t &MinTeamsVal,
-    int32_t &MaxTeamsVal) {
+    llvm::OpenMPIRBuilder::TargetKernelDefaultAttrs &Attrs) {
+  assert(Attrs.MaxTeams.size() == 1 && Attrs.MaxThreads.size() == 1 &&
+         "invalid default attrs structure");
+  int32_t &MaxTeamsVal = Attrs.MaxTeams.front();
+  int32_t &MaxThreadsVal = Attrs.MaxThreads.front();
 
-  getNumTeamsExprForTargetDirective(CGF, D, MinTeamsVal, MaxTeamsVal);
+  getNumTeamsExprForTargetDirective(CGF, D, Attrs.MinTeams, MaxTeamsVal);
   getNumThreadsExprForTargetDirective(CGF, D, MaxThreadsVal,
                                       /*UpperBoundOnly=*/true);
 
@@ -5901,12 +5904,12 @@ void CGOpenMPRuntime::computeMinAndMaxThreadsAndTeams(
       else
         continue;
 
-      MinThreadsVal = std::max(MinThreadsVal, AttrMinThreadsVal);
+      Attrs.MinThreads = std::max(Attrs.MinThreads, AttrMinThreadsVal);
       if (AttrMaxThreadsVal > 0)
         MaxThreadsVal = MaxThreadsVal > 0
                             ? std::min(MaxThreadsVal, AttrMaxThreadsVal)
                             : AttrMaxThreadsVal;
-      MinTeamsVal = std::max(MinTeamsVal, AttrMinBlocksVal);
+      Attrs.MinTeams = std::max(Attrs.MinTeams, AttrMinBlocksVal);
       if (AttrMaxBlocksVal > 0)
         MaxTeamsVal = MaxTeamsVal > 0 ? std::min(MaxTeamsVal, AttrMaxBlocksVal)
                                       : AttrMaxBlocksVal;