Merged main:6932f47cfdf4734d68759586047aee240861058e into amd-gfx:781baf4fd9cc

Local branch amd-gfx 781baf4 Manually merged main:1e70122cbc187c08de91a3fb42843efb1221e0e9 into amd-gfx:e9f81d759667
Remote branch main 6932f47 [NFC][VPlan] Correct two typos in comments.

Author: SC llvm team

clang/docs/ReleaseNotes.rst

@@ -239,6 +239,8 @@ Improvements to Clang's diagnostics
 
 - Clang now diagnoses when the result of a [[nodiscard]] function is discarded after being cast in C. Fixes #GH104391.
 
+- Don't emit duplicated dangling diagnostics. (#GH93386).
+
 Improvements to Clang's time-trace
 ----------------------------------
 

clang/docs/StandardCPlusPlusModules.rst

@@ -1297,74 +1297,6 @@ A high-level overview of support for standards features, including modules, can
 be found on the `C++ Feature Status <https://clang.llvm.org/cxx_status.html>`_
 page.
 
-Missing VTables for classes attached to modules
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
-Now the compiler may miss emitting the definition of vtables
-for classes attached to modules, if the definition of the class
-doesn't contain any key function in that module units
-(The key function is the first non-pure virtual function that is
-not inline at the point of class definition.)
-
-(Note: technically, the key function is not a thing for modules.
-We use the concept here for convinient.)
-
-For example,
-
-.. code-block:: c++
-
-  // layer1.cppm
-  export module foo:layer1;
-  struct Fruit {
-      virtual ~Fruit() = default;
-      virtual void eval() = 0;
-  };
-  struct Banana : public Fruit {
-      Banana() {}
-      void eval() override;
-  };
-
-  // layer2.cppm
-  export module foo:layer2;
-  import :layer1;
-  export void layer2_fun() {
-      Banana *b = new Banana();
-      b->eval();
-  }
-  void Banana::eval() {
-  }
-
-For the above example, we can't find the definition for the vtable of
-class ``Banana`` in any object files.
-
-The expected behavior is, for dynamic classes attached to named modules,
-the vtable should always be emitted to the module units the class attaches
-to.
-
-To workaround the problem, users can add the key function manually in the
-corresponding module units. e.g.,
-
-.. code-block:: c++
-
-  // layer1.cppm
-  export module foo:layer1;
-  struct Fruit {
-      virtual ~Fruit() = default;
-      virtual void eval() = 0;
-  };
-  struct Banana : public Fruit {
-      // Hack a key function to hint the compiler to emit the virtual table.
-      virtual void anchor();
-
-      Banana() {}
-      void eval() override;
-  };
-
-  void Banana::anchor() {}
-
-This is tracked by
-`#70585 <https://github.com/llvm/llvm-project/issues/70585>`_.
-
 Including headers after import is not well-supported
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
@@ -1436,7 +1368,7 @@ non-module unit depending on the definition of some macros. However, this usage
 is forbidden by P1857R3 which is not yet implemented in Clang. This means that
 is possible to write invalid modules which will no longer be accepted once
 P1857R3 is implemented. This is tracked by
-`#56917 <https://github.com/llvm/llvm-project/issues/56917>`_.
+`#54047 <https://github.com/llvm/llvm-project/issues/54047>`_.
 
 Until then, it is recommended not to mix macros with module declarations.
 

clang/include/clang/Basic/Attr.td

@@ -2200,6 +2200,15 @@ def BTFTypeTag : TypeAttr {
   let LangOpts = [COnly];
 }
 
+def BPFFastCall : InheritableAttr,
+                  TargetSpecificAttr<TargetBPF> {
+  let Spellings = [Clang<"bpf_fastcall">];
+  let Subjects = SubjectList<[FunctionLike]>;
+  let Documentation = [BPFFastCallDocs];
+  let LangOpts = [COnly];
+  let SimpleHandler = 1;
+}
+
 def WebAssemblyExportName : InheritableAttr,
                             TargetSpecificAttr<TargetWebAssembly> {
   let Spellings = [Clang<"export_name">];

clang/include/clang/Basic/AttrDocs.td

@@ -2345,6 +2345,25 @@ section.
   }];
 }
 
+def BPFFastCallDocs : Documentation {
+  let Category = DocCatType;
+  let Content = [{
+Functions annotated with this attribute are likely to be inlined by BPF JIT.
+It is assumed that inlined implementation uses less caller saved registers,
+than a regular function.
+Specifically, the following registers are likely to be preserved:
+- ``R0`` if function return value is ``void``;
+- ``R2-R5` if function takes 1 argument;
+- ``R3-R5` if function takes 2 arguments;
+- ``R4-R5` if function takes 3 arguments;
+- ``R5`` if function takes 4 arguments;
+
+For such functions Clang generates code pattern that allows BPF JIT
+to recognize and remove unnecessary spills and fills of the preserved
+registers.
+  }];
+}
+
 def MipsInterruptDocs : Documentation {
   let Category = DocCatFunction;
   let Heading = "interrupt (MIPS)";

clang/include/clang/Driver/Options.td

@@ -6155,6 +6155,10 @@ def mv8plus : Flag<["-"], "mv8plus">, Group<m_sparc_Features_Group>,
   HelpText<"Enable V8+ mode, allowing use of 64-bit V9 instructions in 32-bit code">;
 def mno_v8plus : Flag<["-"], "mno-v8plus">, Group<m_sparc_Features_Group>,
   HelpText<"Disable V8+ mode">;
+def mfix_gr712rc : Flag<["-"], "mfix-gr712rc">, Group<m_sparc_Features_Group>,
+  HelpText<"Enable workarounds for GR712RC errata">;
+def mfix_ut700 : Flag<["-"], "mfix-ut700">, Group<m_sparc_Features_Group>,
+  HelpText<"Enable workarounds for UT700 errata">;
 foreach i = 1 ... 7 in
   def ffixed_g#i : Flag<["-"], "ffixed-g"#i>, Group<m_sparc_Features_Group>,
     HelpText<"Reserve the G"#i#" register (SPARC only)">;

clang/lib/AST/ByteCode/Compiler.cpp

@@ -1318,15 +1318,6 @@ bool Compiler<Emitter>::VisitArraySubscriptExpr(const ArraySubscriptExpr *E) {
 template <class Emitter>
 bool Compiler<Emitter>::visitInitList(ArrayRef<const Expr *> Inits,
                                       const Expr *ArrayFiller, const Expr *E) {
-
-  QualType QT = E->getType();
-
-  if (const auto *AT = QT->getAs<AtomicType>())
-    QT = AT->getValueType();
-
-  if (QT->isVoidType())
-    return this->emitInvalid(E);
-
   // Handle discarding first.
   if (DiscardResult) {
     for (const Expr *Init : Inits) {
@@ -1336,6 +1327,13 @@ bool Compiler<Emitter>::visitInitList(ArrayRef<const Expr *> Inits,
     return true;
   }
 
+  QualType QT = E->getType();
+  if (const auto *AT = QT->getAs<AtomicType>())
+    QT = AT->getValueType();
+
+  if (QT->isVoidType())
+    return this->emitInvalid(E);
+
   // Primitive values.
   if (std::optional<PrimType> T = classify(QT)) {
     assert(!DiscardResult);
@@ -3251,12 +3249,9 @@ template <class Emitter> bool Compiler<Emitter>::visit(const Expr *E) {
   if (E->getType().isNull())
     return false;
 
-  if (E->getType()->isVoidType())
-    return this->discard(E);
-
   // Create local variable to hold the return value.
-  if (!E->isGLValue() && !E->getType()->isAnyComplexType() &&
-      !classify(E->getType())) {
+  if (!E->getType()->isVoidType() && !E->isGLValue() &&
+      !E->getType()->isAnyComplexType() && !classify(E->getType())) {
     std::optional<unsigned> LocalIndex = allocateLocal(E);
     if (!LocalIndex)
       return false;
@@ -5150,7 +5145,7 @@ bool Compiler<Emitter>::VisitUnaryOperator(const UnaryOperator *E) {
     // We should already have a pointer when we get here.
     return this->delegate(SubExpr);
   case UO_Deref: // *x
-    if (DiscardResult)
+    if (DiscardResult || E->getType()->isVoidType())
       return this->discard(SubExpr);
     return this->visit(SubExpr);
   case UO_Not: // ~x

clang/lib/CodeGen/CGCall.cpp

@@ -2421,6 +2421,8 @@ void CodeGenModule::ConstructAttributeList(StringRef Name,
       FuncAttrs.addAttribute(llvm::Attribute::NoCfCheck);
     if (TargetDecl->hasAttr<LeafAttr>())
       FuncAttrs.addAttribute(llvm::Attribute::NoCallback);
+    if (TargetDecl->hasAttr<BPFFastCallAttr>())
+      FuncAttrs.addAttribute("bpf_fastcall");
 
     HasOptnone = TargetDecl->hasAttr<OptimizeNoneAttr>();
     if (auto *AllocSize = TargetDecl->getAttr<AllocSizeAttr>()) {

clang/lib/CodeGen/CodeGenPGO.cpp

@@ -1195,10 +1195,15 @@ void CodeGenPGO::emitCounterSetOrIncrement(CGBuilderTy &Builder, const Stmt *S,
 
   unsigned Counter = (*RegionCounterMap)[S];
 
-  llvm::Value *Args[] = {FuncNameVar,
-                         Builder.getInt64(FunctionHash),
-                         Builder.getInt32(NumRegionCounters),
-                         Builder.getInt32(Counter), StepV};
+  // Make sure that pointer to global is passed in with zero addrspace
+  // This is relevant during GPU profiling
+  auto *NormalizedFuncNameVarPtr =
+      llvm::ConstantExpr::getPointerBitCastOrAddrSpaceCast(
+          FuncNameVar, llvm::PointerType::get(CGM.getLLVMContext(), 0));
+
+  llvm::Value *Args[] = {
+      NormalizedFuncNameVarPtr, Builder.getInt64(FunctionHash),
+      Builder.getInt32(NumRegionCounters), Builder.getInt32(Counter), StepV};
 
   if (llvm::EnableSingleByteCoverage)
     Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::instrprof_cover),

clang/lib/Driver/ToolChains/Arch/Sparc.cpp

@@ -264,4 +264,17 @@ void sparc::getSparcTargetFeatures(const Driver &D, const ArgList &Args,
 
   if (Args.hasArg(options::OPT_ffixed_i5))
     Features.push_back("+reserve-i5");
+
+  if (Args.hasArg(options::OPT_mfix_gr712rc)) {
+    Features.push_back("+fix-tn0009");
+    Features.push_back("+fix-tn0011");
+    Features.push_back("+fix-tn0012");
+    Features.push_back("+fix-tn0013");
+  }
+
+  if (Args.hasArg(options::OPT_mfix_ut700)) {
+    Features.push_back("+fix-tn0009");
+    Features.push_back("+fix-tn0010");
+    Features.push_back("+fix-tn0013");
+  }
 }