Merge llvm/main into amd-debug

Author: SC llvm team

clang/docs/LanguageExtensions.rst

@@ -848,6 +848,14 @@ of different sizes and signs is forbidden in binary and ternary builtins.
                                                 semantics, see `LangRef
                                                 <http://llvm.org/docs/LangRef.html#llvm-min-intrinsics-comparation>`_
                                                 for the comparison.
+ T __builtin_elementwise_maximumnum(T x, T y)   return x or y, whichever is larger. Follows IEEE 754-2019              floating point types
+                                                semantics, see `LangRef
+                                                <http://llvm.org/docs/LangRef.html#llvm-min-intrinsics-comparation>`_
+                                                for the comparison.
+ T __builtin_elementwise_minimumnum(T x, T y)   return x or y, whichever is smaller. Follows IEEE 754-2019             floating point types
+                                                semantics, see `LangRef
+                                                <http://llvm.org/docs/LangRef.html#llvm-min-intrinsics-comparation>`_
+                                                for the comparison.
 ============================================== ====================================================================== =========================================
 
 

clang/docs/ReleaseNotes.rst

@@ -344,6 +344,7 @@ Non-comprehensive list of changes in this release
 - Added `__builtin_elementwise_exp10`.
 - For AMDPGU targets, added `__builtin_v_cvt_off_f32_i4` that maps to the `v_cvt_off_f32_i4` instruction.
 - Added `__builtin_elementwise_minnum` and `__builtin_elementwise_maxnum`.
+- Added `__builtin_elementwise_minnumnum` and `__builtin_elementwise_maxnumnum`.
 - No longer crashing on invalid Objective-C categories and extensions when
   dumping the AST as JSON. (#GH137320)
 - Clang itself now uses split stacks instead of threads for allocating more

clang/include/clang/Basic/Builtins.td

@@ -1334,6 +1334,18 @@ def ElementwiseMinimum : Builtin {
   let Prototype = "void(...)";
 }
 
+def ElementwiseMaximumNum : Builtin {
+  let Spellings = ["__builtin_elementwise_maximumnum"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
+def ElementwiseMinimumNum : Builtin {
+  let Spellings = ["__builtin_elementwise_minimumnum"];
+  let Attributes = [NoThrow, Const, CustomTypeChecking];
+  let Prototype = "void(...)";
+}
+
 def ElementwiseCeil : Builtin {
   let Spellings = ["__builtin_elementwise_ceil"];
   let Attributes = [NoThrow, Const, CustomTypeChecking];

clang/include/clang/Basic/CodeGenOptions.def

@@ -313,9 +313,10 @@ CODEGENOPT(SpeculativeLoadHardening, 1, 0, Benign) ///< Enable speculative load
 CODEGENOPT(FineGrainedBitfieldAccesses, 1, 0, Benign) ///< Enable fine-grained bitfield accesses.
 CODEGENOPT(StrictEnums       , 1, 0, Benign) ///< Optimize based on strict enum definition.
 CODEGENOPT(StrictVTablePointers, 1, 0, Benign) ///< Optimize based on the strict vtable pointers
-CODEGENOPT(TimePasses        , 1, 0, Benign) ///< Set when -ftime-report or -ftime-report= or -ftime-report-json is enabled.
+CODEGENOPT(TimePasses        , 1, 0, Benign) ///< Set when -ftime-report, -ftime-report=, -ftime-report-json, or -stats-file-timers is enabled.
 CODEGENOPT(TimePassesPerRun  , 1, 0, Benign) ///< Set when -ftime-report=per-pass-run is enabled.
 CODEGENOPT(TimePassesJson    , 1, 0, Benign) ///< Set when -ftime-report-json is enabled.
+CODEGENOPT(TimePassesStatsFile     , 1, 0, Benign) ///< Set when -stats-file-timers is enabled.
 CODEGENOPT(TimeTrace         , 1, 0, Benign) ///< Set when -ftime-trace is enabled.
 VALUE_CODEGENOPT(TimeTraceGranularity, 32, 500, Benign) ///< Minimum time granularity (in microseconds),
                                                         ///< traced by time profiler

clang/include/clang/Driver/Options.td

@@ -8278,6 +8278,9 @@ def stats_file : Joined<["-"], "stats-file=">,
 def stats_file_append : Flag<["-"], "stats-file-append">,
   HelpText<"If stats should be appended to stats-file instead of overwriting it">,
   MarshallingInfoFlag<FrontendOpts<"AppendStats">>;
+def stats_file_timers : Flag<["-"], "stats-file-timers">,
+                        HelpText<"If stats should include timers.">,
+                        MarshallingInfoFlag<CodeGenOpts<"TimePassesStatsFile">>;
 def fdump_record_layouts_simple : Flag<["-"], "fdump-record-layouts-simple">,
   HelpText<"Dump record layout information in a simple form used for testing">,
   MarshallingInfoFlag<LangOpts<"DumpRecordLayoutsSimple">>;

clang/lib/AST/ByteCode/Compiler.cpp

@@ -25,34 +25,6 @@ using APSInt = llvm::APSInt;
 namespace clang {
 namespace interp {
 
-static bool refersToUnion(const Expr *E) {
-  for (;;) {
-    if (const auto *ME = dyn_cast<MemberExpr>(E)) {
-      if (const auto *FD = dyn_cast<FieldDecl>(ME->getMemberDecl());
-          FD && FD->getParent()->isUnion())
-        return true;
-      E = ME->getBase();
-      continue;
-    }
-
-    if (const auto *ASE = dyn_cast<ArraySubscriptExpr>(E)) {
-      E = ASE->getBase()->IgnoreImplicit();
-      continue;
-    }
-
-    if (const auto *ICE = dyn_cast<ImplicitCastExpr>(E);
-        ICE && (ICE->getCastKind() == CK_NoOp ||
-                ICE->getCastKind() == CK_DerivedToBase ||
-                ICE->getCastKind() == CK_UncheckedDerivedToBase)) {
-      E = ICE->getSubExpr();
-      continue;
-    }
-
-    break;
-  }
-  return false;
-}
-
 static std::optional<bool> getBoolValue(const Expr *E) {
   if (const auto *CE = dyn_cast_if_present<ConstantExpr>(E);
       CE && CE->hasAPValueResult() &&
@@ -5401,6 +5373,53 @@ bool Compiler<Emitter>::maybeEmitDeferredVarInit(const VarDecl *VD) {
   return true;
 }
 
+static bool hasTrivialDefaultCtorParent(const FieldDecl *FD) {
+  assert(FD);
+  assert(FD->getParent()->isUnion());
+  const auto *CXXRD = dyn_cast<CXXRecordDecl>(FD->getParent());
+  return !CXXRD || CXXRD->hasTrivialDefaultConstructor();
+}
+
+template <class Emitter> bool Compiler<Emitter>::refersToUnion(const Expr *E) {
+  for (;;) {
+    if (const auto *ME = dyn_cast<MemberExpr>(E)) {
+      if (const auto *FD = dyn_cast<FieldDecl>(ME->getMemberDecl());
+          FD && FD->getParent()->isUnion() && hasTrivialDefaultCtorParent(FD))
+        return true;
+      E = ME->getBase();
+      continue;
+    }
+
+    if (const auto *ASE = dyn_cast<ArraySubscriptExpr>(E)) {
+      E = ASE->getBase()->IgnoreImplicit();
+      continue;
+    }
+
+    if (const auto *ICE = dyn_cast<ImplicitCastExpr>(E);
+        ICE && (ICE->getCastKind() == CK_NoOp ||
+                ICE->getCastKind() == CK_DerivedToBase ||
+                ICE->getCastKind() == CK_UncheckedDerivedToBase)) {
+      E = ICE->getSubExpr();
+      continue;
+    }
+
+    if (const auto *This = dyn_cast<CXXThisExpr>(E)) {
+      const auto *ThisRecord =
+          This->getType()->getPointeeType()->getAsRecordDecl();
+      if (!ThisRecord->isUnion())
+        return false;
+      // Otherwise, always activate if we're in the ctor.
+      if (const auto *Ctor =
+              dyn_cast_if_present<CXXConstructorDecl>(CompilingFunction))
+        return Ctor->getParent() == ThisRecord;
+      return false;
+    }
+
+    break;
+  }
+  return false;
+}
+
 template <class Emitter>
 bool Compiler<Emitter>::visitDeclStmt(const DeclStmt *DS,
                                       bool EvaluateConditionDecl) {
@@ -5933,16 +5952,15 @@ bool Compiler<Emitter>::compileConstructor(const CXXConstructorDecl *Ctor) {
       return false;
 
     if (OptPrimType T = this->classify(InitExpr)) {
+      if (Activate && !this->emitActivateThisField(FieldOffset, InitExpr))
+        return false;
+
       if (!this->visit(InitExpr))
         return false;
 
       bool BitField = F->isBitField();
-      if (BitField && Activate)
-        return this->emitInitThisBitFieldActivate(*T, F, FieldOffset, InitExpr);
       if (BitField)
         return this->emitInitThisBitField(*T, F, FieldOffset, InitExpr);
-      if (Activate)
-        return this->emitInitThisFieldActivate(*T, FieldOffset, InitExpr);
       return this->emitInitThisField(*T, FieldOffset, InitExpr);
     }
     // Non-primitive case. Get a pointer to the field-to-initialize

clang/lib/AST/ByteCode/Compiler.h

@@ -401,6 +401,8 @@ class Compiler : public ConstStmtVisitor<Compiler<Emitter>, bool>,
   bool checkLiteralType(const Expr *E);
   bool maybeEmitDeferredVarInit(const VarDecl *VD);
 
+  bool refersToUnion(const Expr *E);
+
 protected:
   /// Variable to storage mapping.
   llvm::DenseMap<const ValueDecl *, Scope::Local> Locals;

clang/lib/AST/ByteCode/Interp.h

@@ -1983,6 +1983,16 @@ static inline bool Activate(InterpState &S, CodePtr OpPC) {
   return true;
 }
 
+static inline bool ActivateThisField(InterpState &S, CodePtr OpPC, uint32_t I) {
+  if (S.checkingPotentialConstantExpression())
+    return false;
+
+  const Pointer &Ptr = S.Current->getThis();
+  assert(Ptr.atField(I).canBeInitialized());
+  Ptr.atField(I).activate();
+  return true;
+}
+
 template <PrimType Name, class T = typename PrimConv<Name>::T>
 bool StoreActivate(InterpState &S, CodePtr OpPC) {
   const T &Value = S.Stk.pop<T>();

clang/lib/AST/ByteCode/Opcodes.td

@@ -510,6 +510,7 @@ def StoreBitFieldActivate : StoreBitFieldOpcode {}
 def StoreBitFieldActivatePop : StoreBitFieldOpcode {}
 
 def Activate : Opcode {}
+def ActivateThisField : Opcode { let Args = [ArgUint32]; }
 
 // [Pointer, Value] -> []
 def Init : StoreOpcode {}

clang/lib/Basic/TargetInfo.cpp

@@ -172,7 +172,7 @@ TargetInfo::TargetInfo(const llvm::Triple &T) : Triple(T) {
   ComplexLongDoubleUsesFP2Ret = false;
 
   // Set the C++ ABI based on the triple.
-  TheCXXABI.set(Triple.isKnownWindowsMSVCEnvironment()
+  TheCXXABI.set(Triple.isKnownWindowsMSVCEnvironment() || Triple.isUEFI()
                     ? TargetCXXABI::Microsoft
                     : TargetCXXABI::GenericItanium);