merge main into amd-staging (#359)

Author: ronlieb

clang/include/clang/AST/Attr.h

@@ -239,6 +239,8 @@ class HLSLSemanticAttr : public HLSLAnnotationAttr {
   LLVM_PREFERRED_TYPE(bool)
   unsigned SemanticExplicitIndex : 1;
 
+  Decl *TargetDecl = nullptr;
+
 protected:
   HLSLSemanticAttr(ASTContext &Context, const AttributeCommonInfo &CommonInfo,
                    attr::Kind AK, bool IsLateParsed,
@@ -259,6 +261,11 @@ class HLSLSemanticAttr : public HLSLAnnotationAttr {
 
   unsigned getSemanticIndex() const { return SemanticIndex; }
 
+  bool isSemanticIndexExplicit() const { return SemanticExplicitIndex; }
+
+  void setTargetDecl(Decl *D) { TargetDecl = D; }
+  Decl *getTargetDecl() const { return TargetDecl; }
+
   // Implement isa/cast/dyncast/etc.
   static bool classof(const Attr *A) {
     return A->getKind() >= attr::FirstHLSLSemanticAttr &&

clang/include/clang/Basic/Attr.td

@@ -787,6 +787,8 @@ class HLSLSemanticAttr<bit Indexable> : HLSLAnnotationAttr {
   let Spellings = [];
   let Subjects = SubjectList<[ParmVar, Field, Function]>;
   let LangOpts = [HLSL];
+  let Args = [DeclArgument<Named, "Target">, IntArgument<"SemanticIndex">,
+              BoolArgument<"SemanticExplicitIndex">];
 }
 
 /// A target-specific attribute.  This class is meant to be used as a mixin

clang/include/clang/Basic/DiagnosticFrontendKinds.td

@@ -404,10 +404,6 @@ def warn_hlsl_langstd_minimal :
           "recommend using %1 instead">,
   InGroup<HLSLDXCCompat>;
 
-def err_hlsl_semantic_missing : Error<"semantic annotations must be present "
-                                      "for all input and outputs of an entry "
-                                      "function or patch constant function">;
-
 // ClangIR frontend errors
 def err_cir_to_cir_transform_failed : Error<
     "CIR-to-CIR transformation failed">, DefaultFatal;

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -13173,6 +13173,7 @@ def err_hlsl_duplicate_parameter_modifier : Error<"duplicate parameter modifier
 def err_hlsl_missing_semantic_annotation : Error<
   "semantic annotations must be present for all parameters of an entry "
   "function or patch constant function">;
+def note_hlsl_semantic_used_here : Note<"%0 used here">;
 def err_hlsl_unknown_semantic : Error<"unknown HLSL semantic %0">;
 def err_hlsl_semantic_output_not_supported
     : Error<"semantic %0 does not support output">;

clang/include/clang/CIR/Dialect/IR/CIROps.td

@@ -4052,6 +4052,43 @@ def CIR_ExpectOp : CIR_Op<"expect", [
   }];
 }
 
+//===----------------------------------------------------------------------===//
+// PrefetchOp
+//===----------------------------------------------------------------------===//
+
+def CIR_PrefetchOp : CIR_Op<"prefetch"> {
+  let summary = "Prefetch operation";
+  let description = [{
+    The `cir.prefetch` operation is a hint to the code generator to insert a
+    prefetch instruction if supported; otherwise, it is a noop. Prefetches
+    have no effect on the behavior of the program but can change its
+    performance characteristics.
+
+    ```mlir
+    cir.prefetch(%0 : !cir.ptr<!void>) locality(1) write
+    ```
+
+    $locality is a temporal locality specifier ranging from (0) - no locality,
+    to (3) - extremely local, keep in cache. If $locality is not present, the
+    default value is 3.
+    
+    $isWrite specifies whether the prefetch is for a 'read' or 'write'. If
+    $isWrite is not specified, it means that prefetch is prepared for 'read'.
+  }];
+
+  let arguments = (ins CIR_VoidPtrType:$addr,
+      DefaultValuedAttr<ConfinedAttr<I32Attr, [IntMinValue<0>, IntMaxValue<3>]>,
+                        "3">:$locality,
+      UnitAttr:$isWrite);
+
+  let assemblyFormat = [{
+        (`write` $isWrite^) : (`read`)?
+        `locality` `(` $locality `)`
+        $addr `:` qualified(type($addr))
+        attr-dict
+    }];
+}
+
 //===----------------------------------------------------------------------===//
 // PtrDiffOp
 //===----------------------------------------------------------------------===//

clang/include/clang/Sema/SemaHLSL.h

@@ -130,9 +130,6 @@ class SemaHLSL : public SemaBase {
   bool ActOnUninitializedVarDecl(VarDecl *D);
   void ActOnEndOfTranslationUnit(TranslationUnitDecl *TU);
   void CheckEntryPoint(FunctionDecl *FD);
-  bool isSemanticValid(FunctionDecl *FD, DeclaratorDecl *D);
-  void CheckSemanticAnnotation(FunctionDecl *EntryPoint, const Decl *Param,
-                               const HLSLAnnotationAttr *AnnotationAttr);
   bool CheckResourceBinOp(BinaryOperatorKind Opc, Expr *LHSExpr, Expr *RHSExpr,
                           SourceLocation Loc);
   void DiagnoseAttrStageMismatch(
@@ -179,17 +176,17 @@ class SemaHLSL : public SemaBase {
   bool handleResourceTypeAttr(QualType T, const ParsedAttr &AL);
 
   template <typename T>
-  T *createSemanticAttr(const ParsedAttr &AL,
+  T *createSemanticAttr(const AttributeCommonInfo &ACI, NamedDecl *TargetDecl,
                         std::optional<unsigned> Location) {
-    T *Attr = ::new (getASTContext()) T(getASTContext(), AL);
-    if (Attr->isSemanticIndexable())
-      Attr->setSemanticIndex(Location ? *Location : 0);
-    else if (Location.has_value()) {
+    T *Attr =
+        ::new (getASTContext()) T(getASTContext(), ACI, TargetDecl,
+                                  Location.value_or(0), Location.has_value());
+
+    if (!Attr->isSemanticIndexable() && Location.has_value()) {
       Diag(Attr->getLocation(), diag::err_hlsl_semantic_indexing_not_supported)
           << Attr->getAttrName()->getName();
       return nullptr;
     }
-
     return Attr;
   }
 
@@ -247,10 +244,25 @@ class SemaHLSL : public SemaBase {
 
   IdentifierInfo *RootSigOverrideIdent = nullptr;
 
+  struct SemanticInfo {
+    HLSLSemanticAttr *Semantic;
+    std::optional<uint32_t> Index;
+  };
+
 private:
   void collectResourceBindingsOnVarDecl(VarDecl *D);
   void collectResourceBindingsOnUserRecordDecl(const VarDecl *VD,
                                                const RecordType *RT);
+
+  void checkSemanticAnnotation(FunctionDecl *EntryPoint, const Decl *Param,
+                               const HLSLSemanticAttr *SemanticAttr);
+  HLSLSemanticAttr *createSemantic(const SemanticInfo &Semantic,
+                                   DeclaratorDecl *TargetDecl);
+  bool determineActiveSemanticOnScalar(FunctionDecl *FD, DeclaratorDecl *D,
+                                       SemanticInfo &ActiveSemantic);
+  bool determineActiveSemantic(FunctionDecl *FD, DeclaratorDecl *D,
+                               SemanticInfo &ActiveSemantic);
+
   void processExplicitBindingsOnDecl(VarDecl *D);
 
   void diagnoseAvailabilityViolations(TranslationUnitDecl *TU);

clang/lib/AST/ByteCode/Compiler.cpp

@@ -1842,7 +1842,6 @@ bool Compiler<Emitter>::visitInitList(ArrayRef<const Expr *> Inits,
                                   const Expr *Init, PrimType T,
                                   bool Activate = false) -> bool {
       InitStackScope<Emitter> ISS(this, isa<CXXDefaultInitExpr>(Init));
-      InitLinkScope<Emitter> ILS(this, InitLink::Field(FieldToInit->Offset));
       if (!this->visit(Init))
         return false;
 
@@ -5385,55 +5384,57 @@ 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)
+  if (!InitStackActive || InitStack.empty())
     return this->emitThis(E);
 
-  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;
-      }
+  // 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;
+  // 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;
-    }
+    if (InitStack[StartIndex].Kind != InitLink::K_Field &&
+        InitStack[StartIndex].Kind != InitLink::K_Elem)
+      break;
+  }
 
-    // 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;
-    }
-    return true;
+  if (StartIndex == 0 && EndIndex == 0)
+    EndIndex = InitStack.size() - 1;
+
+  assert(StartIndex < EndIndex);
+
+  // Emit the instructions.
+  for (unsigned I = StartIndex; I != (EndIndex + 1); ++I) {
+    if (InitStack[I].Kind == InitLink::K_InitList)
+      continue;
+    if (!InitStack[I].template emit<Emitter>(this, E))
+      return false;
   }
-  return this->emitThis(E);
+  return true;
 }
 
 template <class Emitter> bool Compiler<Emitter>::visitStmt(const Stmt *S) {
@@ -6295,6 +6296,10 @@ bool Compiler<Emitter>::compileConstructor(const CXXConstructorDecl *Ctor) {
       }
       assert(NestedField);
 
+      unsigned FirstLinkOffset =
+          R->getField(cast<FieldDecl>(IFD->chain()[0]))->Offset;
+      InitStackScope<Emitter> ISS(this, isa<CXXDefaultInitExpr>(InitExpr));
+      InitLinkScope<Emitter> ILS(this, InitLink::Field(FirstLinkOffset));
       if (!emitFieldInitializer(NestedField, NestedFieldOffset, InitExpr,
                                 IsUnion))
         return false;

clang/lib/CIR/CodeGen/CIRGenBuiltin.cpp

@@ -454,6 +454,27 @@ RValue CIRGenFunction::emitBuiltinExpr(const GlobalDecl &gd, unsigned builtinID,
     assert(!cir::MissingFeatures::coroSizeBuiltinCall());
     return getUndefRValue(e->getType());
   }
+  case Builtin::BI__builtin_prefetch: {
+    auto evaluateOperandAsInt = [&](const Expr *arg) {
+      Expr::EvalResult res;
+      [[maybe_unused]] bool evalSucceed =
+          arg->EvaluateAsInt(res, cgm.getASTContext());
+      assert(evalSucceed && "expression should be able to evaluate as int");
+      return res.Val.getInt().getZExtValue();
+    };
+
+    bool isWrite = false;
+    if (e->getNumArgs() > 1)
+      isWrite = evaluateOperandAsInt(e->getArg(1));
+
+    int locality = 3;
+    if (e->getNumArgs() > 2)
+      locality = evaluateOperandAsInt(e->getArg(2));
+
+    mlir::Value address = emitScalarExpr(e->getArg(0));
+    cir::PrefetchOp::create(builder, loc, address, locality, isWrite);
+    return RValue::get(nullptr);
+  }
   }
 
   // If this is an alias for a lib function (e.g. __builtin_sin), emit

clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp

@@ -1695,6 +1695,15 @@ static uint64_t getTypeSize(mlir::Type type, mlir::Operation &op) {
   return llvm::divideCeil(layout.getTypeSizeInBits(type), 8);
 }
 
+mlir::LogicalResult CIRToLLVMPrefetchOpLowering::matchAndRewrite(
+    cir::PrefetchOp op, OpAdaptor adaptor,
+    mlir::ConversionPatternRewriter &rewriter) const {
+  rewriter.replaceOpWithNewOp<mlir::LLVM::Prefetch>(
+      op, adaptor.getAddr(), adaptor.getIsWrite(), adaptor.getLocality(),
+      /*DataCache=*/1);
+  return mlir::success();
+}
+
 mlir::LogicalResult CIRToLLVMPtrDiffOpLowering::matchAndRewrite(
     cir::PtrDiffOp op, OpAdaptor adaptor,
     mlir::ConversionPatternRewriter &rewriter) const {