[OPENMP] Pass scalar firstprivate vars by value.

For better performance and to unify code with offloading part we pass
scalar firstprivate values by value, instead of by reference. It will
remove some extra copying operations.

llvm-svn: 269751

3 files changed, 43 insertions, 23 deletions

diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index d0b1de81006..87e2ed61af2 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -4709,7 +4709,8 @@ void CGOpenMPRuntime::emitTargetOutlinedFunctionHelper(
   CGOpenMPTargetRegionInfo CGInfo(CS, CodeGen, EntryFnName);
   CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CGInfo);
 
-  OutlinedFn = CGF.GenerateOpenMPCapturedStmtFunction(CS);
+  OutlinedFn =
+      CGF.GenerateOpenMPCapturedStmtFunction(CS, /*CastValToPtr=*/true);
 
   // If this target outline function is not an offload entry, we don't need to
   // register it.
@@ -5553,8 +5554,7 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
           CurMapTypes.push_back(MappableExprsHandler::OMP_MAP_BYCOPY);
           if (!RI->getType()->isAnyPointerType()) {
             // If the field is not a pointer, we need to save the actual value
-            // and
-            // load it as a void pointer.
+            // and load it as a void pointer.
             auto DstAddr = CGF.CreateMemTemp(
                 Ctx.getUIntPtrType(),
                 Twine(CI->getCapturedVar()->getName()) + ".casted");
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 5cf0cae161b..d28f2cceb08 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -172,7 +172,8 @@ static Address castValueFromUintptr(CodeGenFunction &CGF, QualType DstType,
 }
 
 llvm::Function *
-CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
+CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S,
+                                                    bool CastValToPtr) {
   assert(
       CapturedStmtInfo &&
       "CapturedStmtInfo should be set when generating the captured function");
@@ -196,9 +197,11 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
     // uintptr. This is necessary given that the runtime library is only able to
     // deal with pointers. We can pass in the same way the VLA type sizes to the
     // outlined function.
-    if ((I->capturesVariableByCopy() && !ArgType->isAnyPointerType()) ||
-        I->capturesVariableArrayType())
-      ArgType = Ctx.getUIntPtrType();
+    if (CastValToPtr) {
+      if ((I->capturesVariableByCopy() && !ArgType->isAnyPointerType()) ||
+          I->capturesVariableArrayType())
+        ArgType = Ctx.getUIntPtrType();
+    }
 
     if (I->capturesVariable() || I->capturesVariableByCopy()) {
       CapVar = I->getCapturedVar();
@@ -252,9 +255,12 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
                        AlignmentSource::Decl);
     if (FD->hasCapturedVLAType()) {
       LValue CastedArgLVal =
-          MakeAddrLValue(castValueFromUintptr(*this, FD->getType(),
-                                              Args[Cnt]->getName(), ArgLVal),
-                         FD->getType(), AlignmentSource::Decl);
+          CastValToPtr
+              ? MakeAddrLValue(castValueFromUintptr(*this, FD->getType(),
+                                                    Args[Cnt]->getName(),
+                                                    ArgLVal),
+                               FD->getType(), AlignmentSource::Decl)
+              : ArgLVal;
       auto *ExprArg =
           EmitLoadOfLValue(CastedArgLVal, SourceLocation()).getScalarVal();
       auto VAT = FD->getCapturedVLAType();
@@ -274,10 +280,16 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
              "Not expecting a captured pointer.");
       auto *Var = I->getCapturedVar();
       QualType VarTy = Var->getType();
-      setAddrOfLocalVar(I->getCapturedVar(),
-                        castValueFromUintptr(*this, FD->getType(),
-                                             Args[Cnt]->getName(), ArgLVal,
-                                             VarTy->isReferenceType()));
+      if (!CastValToPtr && VarTy->isReferenceType()) {
+        Address Temp = CreateMemTemp(VarTy);
+        Builder.CreateStore(ArgLVal.getPointer(), Temp);
+        ArgLVal = MakeAddrLValue(Temp, VarTy);
+      }
+      setAddrOfLocalVar(Var, CastValToPtr ? castValueFromUintptr(
+                                                *this, FD->getType(),
+                                                Args[Cnt]->getName(), ArgLVal,
+                                                VarTy->isReferenceType())
+                                          : ArgLVal.getAddress());
     } else {
       // If 'this' is captured, load it into CXXThisValue.
       assert(I->capturesThis());
@@ -564,18 +576,25 @@ bool CodeGenFunction::EmitOMPFirstprivateClause(const OMPExecutableDirective &D,
     auto InitsRef = C->inits().begin();
     for (auto IInit : C->private_copies()) {
       auto *OrigVD = cast<VarDecl>(cast<DeclRefExpr>(*IRef)->getDecl());
+      bool ThisFirstprivateIsLastprivate =
+          Lastprivates.count(OrigVD->getCanonicalDecl()) > 0;
+      auto *FD = CapturedStmtInfo->lookup(OrigVD);
+      if (!ThisFirstprivateIsLastprivate && FD &&
+          !FD->getType()->isReferenceType()) {
+        EmittedAsFirstprivate.insert(OrigVD->getCanonicalDecl());
+        ++IRef;
+        ++InitsRef;
+        continue;
+      }
       FirstprivateIsLastprivate =
-          FirstprivateIsLastprivate ||
-          (Lastprivates.count(OrigVD->getCanonicalDecl()) > 0);
+          FirstprivateIsLastprivate || ThisFirstprivateIsLastprivate;
       if (EmittedAsFirstprivate.insert(OrigVD->getCanonicalDecl()).second) {
         auto *VD = cast<VarDecl>(cast<DeclRefExpr>(IInit)->getDecl());
         auto *VDInit = cast<VarDecl>(cast<DeclRefExpr>(*InitsRef)->getDecl());
         bool IsRegistered;
-        DeclRefExpr DRE(
-            const_cast<VarDecl *>(OrigVD),
-            /*RefersToEnclosingVariableOrCapture=*/CapturedStmtInfo->lookup(
-                OrigVD) != nullptr,
-            (*IRef)->getType(), VK_LValue, (*IRef)->getExprLoc());
+        DeclRefExpr DRE(const_cast<VarDecl *>(OrigVD),
+                        /*RefersToEnclosingVariableOrCapture=*/FD != nullptr,
+                        (*IRef)->getType(), VK_LValue, (*IRef)->getExprLoc());
         Address OriginalAddr = EmitLValue(&DRE).getAddress();
         QualType Type = VD->getType();
         if (Type->isArrayType()) {
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index 3a3e54ce620..0100ac3ecda 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -191,6 +191,8 @@ public:
           CXXThisFieldDecl = *Field;
         else if (I->capturesVariable())
           CaptureFields[I->getCapturedVar()] = *Field;
+        else if (I->capturesVariableByCopy())
+          CaptureFields[I->getCapturedVar()] = *Field;
       }
     }
 
@@ -2229,8 +2231,7 @@ public:
   llvm::Function *GenerateCapturedStmtFunction(const CapturedStmt &S);
   Address GenerateCapturedStmtArgument(const CapturedStmt &S);
   llvm::Function *GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S,
-                                                     QualType ReturnQTy);
-  llvm::Function *GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S);
+                                                     bool CastValToPtr = false);
   void GenerateOpenMPCapturedVars(const CapturedStmt &S,
                                   SmallVectorImpl<llvm::Value *> &CapturedVars);
   void emitOMPSimpleStore(LValue LVal, RValue RVal, QualType RValTy,