Implement semantic analysis and an AST representation for the named

return value optimization. Sema marks return statements with their
NRVO candidates (which may or may not end up using the NRVO), then, at
the end of a function body, computes and marks those variables that
can be allocated into the return slot.

I've checked this locally with some debugging statements (not
committed), but there won't be any tests until CodeGen comes along.



git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@103865 91177308-0d34-0410-b5e6-96231b3b80d8

include/clang/AST/Decl.h

@@ -519,6 +519,10 @@ private:
   /// or an Objective-C @catch statement.
   bool ExceptionVar : 1;
   
+  /// \brief Whether this local variable could be allocated in the return
+  /// slot of its function, enabling the named return value optimization (NRVO).
+  bool NRVOVariable : 1;
+  
   friend class StmtIteratorBase;
 protected:
   VarDecl(Kind DK, DeclContext *DC, SourceLocation L, IdentifierInfo *Id,
@@ -526,7 +530,7 @@ protected:
           StorageClass SCAsWritten)
     : DeclaratorDecl(DK, DC, L, Id, T, TInfo), Init(),
       ThreadSpecified(false), HasCXXDirectInit(false),
-      DeclaredInCondition(false), ExceptionVar(false) {
+      DeclaredInCondition(false), ExceptionVar(false), NRVOVariable(false) {
     SClass = SC;
     SClassAsWritten = SCAsWritten;
   }
@@ -869,6 +873,19 @@ public:
   }
   void setExceptionVariable(bool EV) { ExceptionVar = EV; }
   
+  /// \brief Determine whether this local variable can be used with the named
+  /// return value optimization (NRVO).
+  ///
+  /// The named return value optimization (NRVO) works by marking certain
+  /// non-volatile local variables of class type as NRVO objects. These
+  /// locals can be allocated within the return slot of their containing
+  /// function, in which case there is no need to copy the object to the
+  /// return slot when returning from the function. Within the function body,
+  /// each return that returns the NRVO object will have this variable as its
+  /// NRVO candidate.
+  bool isNRVOVariable() const { return NRVOVariable; }
+  void setNRVOVariable(bool NRVO) { NRVOVariable = NRVO; }
+  
   /// \brief If this variable is an instantiated static data member of a
   /// class template specialization, returns the templated static data member
   /// from which it was instantiated.

include/clang/AST/Stmt.h

@@ -1096,9 +1096,15 @@ public:
 class ReturnStmt : public Stmt {
   Stmt *RetExpr;
   SourceLocation RetLoc;
+  const VarDecl *NRVOCandidate;
+  
 public:
-  ReturnStmt(SourceLocation RL, Expr *E = 0) : Stmt(ReturnStmtClass),
-    RetExpr((Stmt*) E), RetLoc(RL) {}
+  ReturnStmt(SourceLocation RL)
+    : Stmt(ReturnStmtClass), RetExpr(0), NRVOCandidate(0) { }
+
+  ReturnStmt(SourceLocation RL, Expr *E, const VarDecl *NRVOCandidate)
+    : Stmt(ReturnStmtClass), RetExpr((Stmt*) E), RetLoc(RL),
+      NRVOCandidate(NRVOCandidate) {}
 
   /// \brief Build an empty return expression.
   explicit ReturnStmt(EmptyShell Empty) : Stmt(ReturnStmtClass, Empty) { }
@@ -1110,6 +1116,14 @@ public:
   SourceLocation getReturnLoc() const { return RetLoc; }
   void setReturnLoc(SourceLocation L) { RetLoc = L; }
 
+  /// \brief Retrieve the variable that might be used for the named return
+  /// value optimization.
+  ///
+  /// The optimization itself can only be performed if the variable is
+  /// also marked as an NRVO object.
+  const VarDecl *getNRVOCandidate() const { return NRVOCandidate; }
+  void setNRVOCandidate(const VarDecl *Var) { NRVOCandidate = Var; }
+  
   virtual SourceRange getSourceRange() const;
 
   static bool classof(const Stmt *T) {

lib/CodeGen/CGObjC.cpp

@@ -255,7 +255,8 @@ void CodeGenFunction::GenerateObjCGetter(ObjCImplementationDecl *IMP,
         if (PID->getGetterCXXConstructor()) {
           ReturnStmt *Stmt = 
             new (getContext()) ReturnStmt(SourceLocation(), 
-                                          PID->getGetterCXXConstructor());
+                                          PID->getGetterCXXConstructor(),
+                                          0);
           EmitReturnStmt(*Stmt);
         }
         else {

lib/Frontend/PCHReaderDecl.cpp

@@ -435,6 +435,7 @@ void PCHDeclReader::VisitVarDecl(VarDecl *VD) {
   VD->setCXXDirectInitializer(Record[Idx++]);
   VD->setDeclaredInCondition(Record[Idx++]);
   VD->setExceptionVariable(Record[Idx++]);
+  VD->setNRVOVariable(Record[Idx++]);
   VD->setPreviousDeclaration(
                          cast_or_null<VarDecl>(Reader.GetDecl(Record[Idx++])));
   if (Record[Idx++])

lib/Frontend/PCHReaderStmt.cpp

@@ -291,6 +291,7 @@ unsigned PCHStmtReader::VisitReturnStmt(ReturnStmt *S) {
   VisitStmt(S);
   S->setRetValue(cast_or_null<Expr>(StmtStack.back()));
   S->setReturnLoc(SourceLocation::getFromRawEncoding(Record[Idx++]));
+  S->setNRVOCandidate(cast_or_null<VarDecl>(Reader.GetDecl(Record[Idx++])));
   return 1;
 }
 

lib/Frontend/PCHWriterDecl.cpp

@@ -415,6 +415,7 @@ void PCHDeclWriter::VisitVarDecl(VarDecl *D) {
   Record.push_back(D->hasCXXDirectInitializer());
   Record.push_back(D->isDeclaredInCondition());
   Record.push_back(D->isExceptionVariable());
+  Record.push_back(D->isNRVOVariable());
   Writer.AddDeclRef(D->getPreviousDeclaration(), Record);
   Record.push_back(D->getInit() ? 1 : 0);
   if (D->getInit())
@@ -683,6 +684,7 @@ void PCHWriter::WriteDeclsBlockAbbrevs() {
   Abv->Add(BitCodeAbbrevOp(0));                       // hasCXXDirectInitializer
   Abv->Add(BitCodeAbbrevOp(0));                       // isDeclaredInCondition
   Abv->Add(BitCodeAbbrevOp(0));                       // isExceptionVariable
+  Abv->Add(BitCodeAbbrevOp(0));                       // isNRVOVariable
   Abv->Add(BitCodeAbbrevOp(0));                       // PrevDecl
   Abv->Add(BitCodeAbbrevOp(0));                       // HasInit
   // ParmVarDecl

lib/Frontend/PCHWriterStmt.cpp

@@ -275,6 +275,7 @@ void PCHStmtWriter::VisitReturnStmt(ReturnStmt *S) {
   VisitStmt(S);
   Writer.WriteSubStmt(S->getRetValue());
   Writer.AddSourceLocation(S->getReturnLoc(), Record);
+  Writer.AddDeclRef(S->getNRVOCandidate(), Record);
   Code = pch::STMT_RETURN;
 }
 

lib/Sema/Sema.cpp

@@ -33,6 +33,7 @@ void FunctionScopeInfo::Clear(unsigned NumErrors) {
   NeedsScopeChecking = false;
   LabelMap.clear();
   SwitchStack.clear();
+  Returns.clear();
   NumErrorsAtStartOfFunction = NumErrors;
 }
 

lib/Sema/Sema.h

@@ -134,6 +134,11 @@ struct FunctionScopeInfo {
   /// block.
   llvm::SmallVector<SwitchStmt*, 8> SwitchStack;
 
+  /// \brief The list of return statements that occur within the function or
+  /// block, if there is any chance of applying the named return value
+  /// optimization.
+  llvm::SmallVector<ReturnStmt *, 4> Returns;
+  
   FunctionScopeInfo(unsigned NumErrors)
     : IsBlockInfo(false), NeedsScopeChecking(false),
       NumErrorsAtStartOfFunction(NumErrors) { }

lib/Sema/SemaDecl.cpp

@@ -4548,6 +4548,38 @@ Sema::DeclPtrTy Sema::ActOnStartOfFunctionDef(Scope *FnBodyScope, DeclPtrTy D) {
   return DeclPtrTy::make(FD);
 }
 
+/// \brief Given the set of return statements within a function body,
+/// compute the variables that are subject to the named return value 
+/// optimization.
+///
+/// Each of the variables that is subject to the named return value 
+/// optimization will be marked as NRVO variables in the AST, and any
+/// return statement that has a marked NRVO variable as its NRVO candidate can
+/// use the named return value optimization.
+///
+/// This function applies a very simplistic algorithm for NRVO: if every return
+/// statement in the function has the same NRVO candidate, that candidate is
+/// the NRVO variable.
+///
+/// FIXME: Employ a smarter algorithm that accounts for multiple return 
+/// statements and the lifetimes of the NRVO candidates. We should be able to
+/// find a maximal set of NRVO variables.
+static void ComputeNRVO(Stmt *Body, ReturnStmt **Returns, unsigned NumReturns) {
+  const VarDecl *NRVOCandidate = 0;
+  for (unsigned I = 0; I != NumReturns; ++I) {
+    if (!Returns[I]->getNRVOCandidate())
+      return;
+    
+    if (!NRVOCandidate)
+      NRVOCandidate = Returns[I]->getNRVOCandidate();
+    else if (NRVOCandidate != Returns[I]->getNRVOCandidate())
+      return;
+  }
+  
+  if (NRVOCandidate)
+    const_cast<VarDecl*>(NRVOCandidate)->setNRVOVariable(true);
+}
+
 Sema::DeclPtrTy Sema::ActOnFinishFunctionBody(DeclPtrTy D, StmtArg BodyArg) {
   return ActOnFinishFunctionBody(D, move(BodyArg), false);
 }
@@ -4581,6 +4613,9 @@ Sema::DeclPtrTy Sema::ActOnFinishFunctionBody(DeclPtrTy D, StmtArg BodyArg,
       // If this is a constructor, we need a vtable.
       if (CXXConstructorDecl *Constructor = dyn_cast<CXXConstructorDecl>(FD))
         MarkVTableUsed(FD->getLocation(), Constructor->getParent());
+      
+      ComputeNRVO(Body, FunctionScopes.back()->Returns.data(),
+                  FunctionScopes.back()->Returns.size());
     }
     
     assert(FD == getCurFunctionDecl() && "Function parsing confused");

lib/Sema/SemaStmt.cpp

@@ -1058,29 +1058,42 @@ Sema::ActOnBreakStmt(SourceLocation BreakLoc, Scope *CurScope) {
   return Owned(new (Context) BreakStmt(BreakLoc));
 }
 
-/// IsReturnCopyElidable - Whether returning @p RetExpr from a function that
-/// returns a @p RetType fulfills the criteria for copy elision (C++0x 12.8p34).
-static bool IsReturnCopyElidable(ASTContext &Ctx, QualType RetType,
-                                 Expr *RetExpr) {
+/// \brief Determine whether a return statement is a candidate for the named
+/// return value optimization (C++0x 12.8p34, bullet 1).
+///
+/// \param Ctx The context in which the return expression and type occur.
+///
+/// \param RetType The return type of the function or block.
+///
+/// \param RetExpr The expression being returned from the function or block.
+///
+/// \returns The NRVO candidate variable, if the return statement may use the
+/// NRVO, or NULL if there is no such candidate.
+static const VarDecl *getNRVOCandidate(ASTContext &Ctx, QualType RetType,
+                                       Expr *RetExpr) {
   QualType ExprType = RetExpr->getType();
   // - in a return statement in a function with ...
   // ... a class return type ...
   if (!RetType->isRecordType())
-    return false;
+    return 0;
   // ... the same cv-unqualified type as the function return type ...
   if (!Ctx.hasSameUnqualifiedType(RetType, ExprType))
-    return false;
+    return 0;
   // ... the expression is the name of a non-volatile automatic object ...
   // We ignore parentheses here.
   // FIXME: Is this compliant? (Everyone else does it)
   const DeclRefExpr *DR = dyn_cast<DeclRefExpr>(RetExpr->IgnoreParens());
   if (!DR)
-    return false;
+    return 0;
   const VarDecl *VD = dyn_cast<VarDecl>(DR->getDecl());
   if (!VD)
-    return false;
-  return VD->getKind() == Decl::Var && VD->hasLocalStorage() && 
-    !VD->getType()->isReferenceType() && !VD->getType().isVolatileQualified();
+    return 0;
+  
+  if (VD->getKind() == Decl::Var && VD->hasLocalStorage() && 
+      !VD->getType()->isReferenceType() && !VD->getType().isVolatileQualified())
+    return VD;
+  
+  return 0;
 }
 
 /// ActOnBlockReturnStmt - Utility routine to figure out block's return type.
@@ -1117,46 +1130,58 @@ Sema::ActOnBlockReturnStmt(SourceLocation ReturnLoc, Expr *RetValExp) {
   // Otherwise, verify that this result type matches the previous one.  We are
   // pickier with blocks than for normal functions because we don't have GCC
   // compatibility to worry about here.
+  ReturnStmt *Result = 0;
   if (CurBlock->ReturnType->isVoidType()) {
     if (RetValExp) {
       Diag(ReturnLoc, diag::err_return_block_has_expr);
       RetValExp->Destroy(Context);
       RetValExp = 0;
     }
-    return Owned(new (Context) ReturnStmt(ReturnLoc, RetValExp));
-  }
-
-  if (!RetValExp)
+    Result = new (Context) ReturnStmt(ReturnLoc, RetValExp, 0);
+  } else if (!RetValExp) {
     return StmtError(Diag(ReturnLoc, diag::err_block_return_missing_expr));
+  } else {
+    const VarDecl *NRVOCandidate = 0;
+    
+    if (!FnRetType->isDependentType() && !RetValExp->isTypeDependent()) {
+      // we have a non-void block with an expression, continue checking
 
-  if (!FnRetType->isDependentType() && !RetValExp->isTypeDependent()) {
-    // we have a non-void block with an expression, continue checking
+      // C99 6.8.6.4p3(136): The return statement is not an assignment. The
+      // overlap restriction of subclause 6.5.16.1 does not apply to the case of
+      // function return.
 
-    // C99 6.8.6.4p3(136): The return statement is not an assignment. The
-    // overlap restriction of subclause 6.5.16.1 does not apply to the case of
-    // function return.
+      // In C++ the return statement is handled via a copy initialization.
+      // the C version of which boils down to CheckSingleAssignmentConstraints.
+      NRVOCandidate = getNRVOCandidate(Context, FnRetType, RetValExp);
+      OwningExprResult Res = PerformCopyInitialization(
+                               InitializedEntity::InitializeResult(ReturnLoc, 
+                                                                   FnRetType,
+                                                            NRVOCandidate != 0),
+                               SourceLocation(),
+                               Owned(RetValExp));
+      if (Res.isInvalid()) {
+        // FIXME: Cleanup temporaries here, anyway?
+        return StmtError();
+      }
+      
+      if (RetValExp)
+        RetValExp = MaybeCreateCXXExprWithTemporaries(RetValExp);
 
-    // In C++ the return statement is handled via a copy initialization.
-    // the C version of which boils down to CheckSingleAssignmentConstraints.
-    OwningExprResult Res = PerformCopyInitialization(
-                             InitializedEntity::InitializeResult(ReturnLoc, 
-                                                                 FnRetType,
-                                               IsReturnCopyElidable(Context, 
-                                                                    FnRetType, 
-                                                                    RetValExp)),
-                             SourceLocation(),
-                             Owned(RetValExp));
-    if (Res.isInvalid()) {
-      // FIXME: Cleanup temporaries here, anyway?
-      return StmtError();
+      RetValExp = Res.takeAs<Expr>();
+      if (RetValExp) 
+        CheckReturnStackAddr(RetValExp, FnRetType, ReturnLoc);
     }
     
-    RetValExp = Res.takeAs<Expr>();
-    if (RetValExp) 
-      CheckReturnStackAddr(RetValExp, FnRetType, ReturnLoc);
+    Result = new (Context) ReturnStmt(ReturnLoc, RetValExp, NRVOCandidate);
   }
 
-  return Owned(new (Context) ReturnStmt(ReturnLoc, RetValExp));
+  // If we need to check for the named return value optimization, save the 
+  // return statement in our scope for later processing.
+  if (getLangOptions().CPlusPlus && FnRetType->isRecordType() &&
+      !CurContext->isDependentContext())
+    FunctionScopes.back()->Returns.push_back(Result);
+  
+  return Owned(Result);
 }
 
 Action::OwningStmtResult
@@ -1177,6 +1202,7 @@ Sema::ActOnReturnStmt(SourceLocation ReturnLoc, ExprArg rex) {
   else // If we don't have a function/method context, bail.
     return StmtError();
 
+  ReturnStmt *Result = 0;
   if (FnRetType->isVoidType()) {
     if (RetValExp && !RetValExp->isTypeDependent()) {
       // C99 6.8.6.4p1 (ext_ since GCC warns)
@@ -1195,10 +1221,9 @@ Sema::ActOnReturnStmt(SourceLocation ReturnLoc, ExprArg rex) {
 
       RetValExp = MaybeCreateCXXExprWithTemporaries(RetValExp);
     }
-    return Owned(new (Context) ReturnStmt(ReturnLoc, RetValExp));
-  }
-
-  if (!RetValExp && !FnRetType->isDependentType()) {
+    
+    Result = new (Context) ReturnStmt(ReturnLoc, RetValExp, 0);
+  } else if (!RetValExp && !FnRetType->isDependentType()) {
     unsigned DiagID = diag::warn_return_missing_expr;  // C90 6.6.6.4p4
     // C99 6.8.6.4p1 (ext_ since GCC warns)
     if (getLangOptions().C99) DiagID = diag::ext_return_missing_expr;
@@ -1207,43 +1232,47 @@ Sema::ActOnReturnStmt(SourceLocation ReturnLoc, ExprArg rex) {
       Diag(ReturnLoc, DiagID) << FD->getIdentifier() << 0/*fn*/;
     else
       Diag(ReturnLoc, DiagID) << getCurMethodDecl()->getDeclName() << 1/*meth*/;
-    return Owned(new (Context) ReturnStmt(ReturnLoc, (Expr*)0));
-  }
+    Result = new (Context) ReturnStmt(ReturnLoc);
+  } else {
+    const VarDecl *NRVOCandidate = 0;
+    if (!FnRetType->isDependentType() && !RetValExp->isTypeDependent()) {
+      // we have a non-void function with an expression, continue checking
 
-  if (!FnRetType->isDependentType() && !RetValExp->isTypeDependent()) {
-    // we have a non-void function with an expression, continue checking
+      // C99 6.8.6.4p3(136): The return statement is not an assignment. The
+      // overlap restriction of subclause 6.5.16.1 does not apply to the case of
+      // function return.
 
-    // C99 6.8.6.4p3(136): The return statement is not an assignment. The
-    // overlap restriction of subclause 6.5.16.1 does not apply to the case of
-    // function return.
+      // In C++ the return statement is handled via a copy initialization.
+      // the C version of which boils down to CheckSingleAssignmentConstraints.
+      NRVOCandidate = getNRVOCandidate(Context, FnRetType, RetValExp);
+      OwningExprResult Res = PerformCopyInitialization(
+                               InitializedEntity::InitializeResult(ReturnLoc, 
+                                                                   FnRetType,
+                                                            NRVOCandidate != 0),
+                               SourceLocation(),
+                               Owned(RetValExp));
+      if (Res.isInvalid()) {
+        // FIXME: Cleanup temporaries here, anyway?
+        return StmtError();
+      }
 
-    // C++0x [class.copy]p34:
-    //
-    
-    
-    // In C++ the return statement is handled via a copy initialization.
-    // the C version of which boils down to CheckSingleAssignmentConstraints.
-    OwningExprResult Res = PerformCopyInitialization(
-                             InitializedEntity::InitializeResult(ReturnLoc, 
-                                                                 FnRetType,
-                                               IsReturnCopyElidable(Context, 
-                                                                    FnRetType, 
-                                                                    RetValExp)),
-                             SourceLocation(),
-                             Owned(RetValExp));
-    if (Res.isInvalid()) {
-      // FIXME: Cleanup temporaries here, anyway?
-      return StmtError();
+      RetValExp = Res.takeAs<Expr>();
+      if (RetValExp) 
+        CheckReturnStackAddr(RetValExp, FnRetType, ReturnLoc);
     }
-
-    RetValExp = Res.takeAs<Expr>();
-    if (RetValExp) 
-      CheckReturnStackAddr(RetValExp, FnRetType, ReturnLoc);
+    
+    if (RetValExp)
+      RetValExp = MaybeCreateCXXExprWithTemporaries(RetValExp);
+    Result = new (Context) ReturnStmt(ReturnLoc, RetValExp, NRVOCandidate);
   }
-
-  if (RetValExp)
-    RetValExp = MaybeCreateCXXExprWithTemporaries(RetValExp);
-  return Owned(new (Context) ReturnStmt(ReturnLoc, RetValExp));
+  
+  // If we need to check for the named return value optimization, save the 
+  // return statement in our scope for later processing.
+  if (getLangOptions().CPlusPlus && FnRetType->isRecordType() &&
+      !CurContext->isDependentContext())
+    FunctionScopes.back()->Returns.push_back(Result);
+  
+  return Owned(Result);
 }
 
 /// CheckAsmLValue - GNU C has an extremely ugly extension whereby they silently