Fix for aggregate copying of variable length arrays.

[ABataev]

Summary

Patch fixes codegen for aggregate copying of VLAs. Currently method CodeGenFunction::EmitAggregateCopy() does not support copying of VLAs. Patch checks if the size of the type is 0, then checks if the type is actually a variable-length array. Then it calculates total length for this array and calculates total size of the array in bytes:

<total number of elements in array> * aligned_sizeof(ElementType) (if copy assignment is requested).

If simple copying is requested, size is calculated like:

<total number of elements in array> * aligned_sizeof(ElementType) - aligned_sizeof(ElementType) + sizeof(ElementType).

memcpy() is used with this calculated size of the VLA.

[rjmccall]

LGTM.

cfe/trunk/lib/CodeGen/CGExprAgg.cpp

void CodeGenFunction::EmitAggregateCopy(llvm::Value *DestPtr,
   if (isAssignment)
     TypeInfo = getContext().getTypeInfoDataSizeInChars(Ty);
   else
     TypeInfo = getContext().getTypeInfoInChars(Ty);
 
   if (alignment.isZero())
     alignment = TypeInfo.second;
 
-  // FIXME: Handle variable sized types.
+  llvm::Value *SizeVal = nullptr;
+  if (TypeInfo.first.isZero()) {
+    // But note that getTypeInfo returns 0 for a VLA.
+    if (auto *VAT = dyn_cast_or_null<VariableArrayType>(
+            getContext().getAsArrayType(Ty))) {
+      QualType BaseEltTy;
+      SizeVal = emitArrayLength(VAT, BaseEltTy, DestPtr);
+      TypeInfo = getContext().getTypeInfoDataSizeInChars(BaseEltTy);
+      std::pair<CharUnits, CharUnits> LastElementTypeInfo;
+      if (!isAssignment)
+        LastElementTypeInfo = getContext().getTypeInfoInChars(BaseEltTy);
+      assert(!TypeInfo.first.isZero());
+      SizeVal = Builder.CreateNUWMul(
+          SizeVal,
+          llvm::ConstantInt::get(SizeTy, TypeInfo.first.getQuantity()));
+      if (!isAssignment) {
+        SizeVal = Builder.CreateNUWSub(
+            SizeVal,
+            llvm::ConstantInt::get(SizeTy, TypeInfo.first.getQuantity()));
+        SizeVal = Builder.CreateNUWAdd(
+            SizeVal, llvm::ConstantInt::get(
+                         SizeTy, LastElementTypeInfo.first.getQuantity()));
+      }
+    }
+  }
+  if (!SizeVal) {
+    SizeVal = llvm::ConstantInt::get(SizeTy, TypeInfo.first.getQuantity());
+  }
 
   // FIXME: If we have a volatile struct, the optimizer can remove what might
   // appear to be `extra' memory ops:
   //
   // volatile struct { int i; } a, b;
   //
   // int main() {
   //   a = b;
void CodeGenFunction::EmitAggregateCopy(llvm::Value *DestPtr,
   SrcPtr = Builder.CreateBitCast(SrcPtr, SBP);
 
   // Don't do any of the memmove_collectable tests if GC isn't set.
   if (CGM.getLangOpts().getGC() == LangOptions::NonGC) {
     // fall through
   } else if (const RecordType *RecordTy = Ty->getAs<RecordType>()) {
     RecordDecl *Record = RecordTy->getDecl();
     if (Record->hasObjectMember()) {
-      CharUnits size = TypeInfo.first;
-      llvm::Type *SizeTy = ConvertType(getContext().getSizeType());
-      llvm::Value *SizeVal = llvm::ConstantInt::get(SizeTy, size.getQuantity());
       CGM.getObjCRuntime().EmitGCMemmoveCollectable(*this, DestPtr, SrcPtr, 
                                                     SizeVal);
       return;
     }
   } else if (Ty->isArrayType()) {
     QualType BaseType = getContext().getBaseElementType(Ty);
     if (const RecordType *RecordTy = BaseType->getAs<RecordType>()) {
       if (RecordTy->getDecl()->hasObjectMember()) {
-        CharUnits size = TypeInfo.first;
-        llvm::Type *SizeTy = ConvertType(getContext().getSizeType());
-        llvm::Value *SizeVal = 
-          llvm::ConstantInt::get(SizeTy, size.getQuantity());
         CGM.getObjCRuntime().EmitGCMemmoveCollectable(*this, DestPtr, SrcPtr, 
                                                       SizeVal);
         return;
       }
     }
   }
 
   // Determine the metadata to describe the position of any padding in this
   // memcpy, as well as the TBAA tags for the members of the struct, in case
   // the optimizer wishes to expand it in to scalar memory operations.
   llvm::MDNode *TBAAStructTag = CGM.getTBAAStructInfo(Ty);
 
-  Builder.CreateMemCpy(DestPtr, SrcPtr,
-                       llvm::ConstantInt::get(IntPtrTy, 
-                                              TypeInfo.first.getQuantity()),
-                       alignment.getQuantity(), isVolatile,
-                       /*TBAATag=*/nullptr, TBAAStructTag);
+  Builder.CreateMemCpy(DestPtr, SrcPtr, SizeVal, alignment.getQuantity(),
+                       isVolatile, /*TBAATag=*/nullptr, TBAAStructTag);
 }

cfe/trunk/lib/Sema/SemaOpenMP.cpp

 StmtResult Sema::ActOnOpenMPRegionEnd(StmtResult S,
                                       ArrayRef<OMPClause *> Clauses) {
   if (!S.isUsable()) {
     ActOnCapturedRegionError();
     return StmtError();
   }
   // This is required for proper codegen.
   for (auto *Clause : Clauses) {
-    if (isOpenMPPrivate(Clause->getClauseKind())) {
+    if (isOpenMPPrivate(Clause->getClauseKind()) ||
+        Clause->getClauseKind() == OMPC_copyprivate) {
       // Mark all variables in private list clauses as used in inner region.
       for (auto *VarRef : Clause->children()) {
         if (auto *E = cast_or_null<Expr>(VarRef)) {
           MarkDeclarationsReferencedInExpr(E);
         }
       }
     } else if (isParallelOrTaskRegion(DSAStack->getCurrentDirective()) &&
                Clause->getClauseKind() == OMPC_schedule) {

cfe/trunk/test/OpenMP/parallel_firstprivate_codegen.cpp

 // ARRAY-LABEL: array_func
 struct St {
   int a, b;
   St() : a(0), b(0) {}
   St(const St &) { }
   ~St() {}
 };
 
-void array_func(int a[3], St s[2]) {
+void array_func(int a[3], St s[2], int n, long double vla1[n]) {
+  double vla2[n];
 // ARRAY: @__kmpc_fork_call(
 // ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* %{{.+}}, i64 12, i32 4, i1 false)
 // ARRAY: call void @_ZN2StC1ERKS_(%struct.St* %{{.+}}, %struct.St* dereferenceable(8) %{{.+}}
-#pragma omp parallel firstprivate(a, s)
+// ARRAY: call i8* @llvm.stacksave()
+// ARRAY: [[SIZE:%.+]] = mul nuw i64 %{{.+}}, 16
+// ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* %{{.+}}, i64 [[SIZE]], i32 16, i1 false)
+// ARRAY: [[SIZE:%.+]] = mul nuw i64 %{{.+}}, 8
+// ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* %{{.+}}, i64 [[SIZE]], i32 8, i1 false)
+// ARRAY: call void @llvm.stackrestore(i8*
+// ARRAY: call void @_ZN2StD1Ev(%struct.St* %{{.+}})
+// ARRAY: br i1
+#pragma omp parallel firstprivate(a, s, vla1, vla2)
   ;
 }
 #endif