[OPENMP 4.0] Codegen for array sections.

Added codegen for array section in 'depend' clause of 'task' directive. It emits to pointers, one for the begin of array section and another for the end of array section. Size of the section is calculated as (end + 1 - start) * sizeof(basic_element_type).

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@246422 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/CGExpr.cpp b/lib/CodeGen/CGExpr.cpp
index a0d004d315c8b90890bd6fe111805762a8951f72..f6cb3e304b7d8d335d2a57d00efdf318606547b2 100644 (file)
--- a/lib/CodeGen/CGExpr.cpp
+++ b/lib/CodeGen/CGExpr.cpp
@@ -911,6 +911,8 @@ LValue CodeGenFunction::EmitLValue(const Expr *E) {
     return EmitUnaryOpLValue(cast<UnaryOperator>(E));
   case Expr::ArraySubscriptExprClass:
     return EmitArraySubscriptExpr(cast<ArraySubscriptExpr>(E));
+  case Expr::OMPArraySectionExprClass:
+    return EmitOMPArraySectionExpr(cast<OMPArraySectionExpr>(E));
   case Expr::ExtVectorElementExprClass:
     return EmitExtVectorElementExpr(cast<ExtVectorElementExpr>(E));
   case Expr::MemberExprClass:
@@ -2559,6 +2561,155 @@ LValue CodeGenFunction::EmitArraySubscriptExpr(const ArraySubscriptExpr *E,
   return LV;
 }
 
+LValue CodeGenFunction::EmitOMPArraySectionExpr(const OMPArraySectionExpr *E,
+                                                bool IsLowerBound) {
+  LValue Base;
+  if (auto *ASE =
+          dyn_cast<OMPArraySectionExpr>(E->getBase()->IgnoreParenImpCasts()))
+    Base = EmitOMPArraySectionExpr(ASE, IsLowerBound);
+  else
+    Base = EmitLValue(E->getBase());
+  QualType BaseTy = Base.getType();
+  llvm::Value *Idx = nullptr;
+  QualType ResultExprTy;
+  if (auto *AT = getContext().getAsArrayType(BaseTy))
+    ResultExprTy = AT->getElementType();
+  else
+    ResultExprTy = BaseTy->getPointeeType();
+  if (IsLowerBound || (!IsLowerBound && E->getColonLoc().isInvalid())) {
+    // Requesting lower bound or upper bound, but without provided length and
+    // without ':' symbol for the default length -> length = 1.
+    // Idx = LowerBound ?: 0;
+    if (auto *LowerBound = E->getLowerBound()) {
+      Idx = Builder.CreateIntCast(
+          EmitScalarExpr(LowerBound), IntPtrTy,
+          LowerBound->getType()->hasSignedIntegerRepresentation());
+    } else
+      Idx = llvm::ConstantInt::getNullValue(IntPtrTy);
+  } else {
+    // Try to emit length or lower bound as constant. If this is possible, 1 is
+    // subtracted from constant length or lower bound. Otherwise, emit LLVM IR
+    // (LB + Len) - 1.
+    auto &C = CGM.getContext();
+    auto *Length = E->getLength();
+    llvm::APSInt ConstLength;
+    if (Length) {
+      // Idx = LowerBound + Length - 1;
+      if (Length->isIntegerConstantExpr(ConstLength, C)) {
+        ConstLength = ConstLength.zextOrTrunc(PointerWidthInBits);
+        Length = nullptr;
+      }
+      auto *LowerBound = E->getLowerBound();
+      llvm::APSInt ConstLowerBound(PointerWidthInBits, /*isUnsigned=*/false);
+      if (LowerBound && LowerBound->isIntegerConstantExpr(ConstLowerBound, C)) {
+        ConstLowerBound = ConstLowerBound.zextOrTrunc(PointerWidthInBits);
+        LowerBound = nullptr;
+      }
+      if (!Length)
+        --ConstLength;
+      else if (!LowerBound)
+        --ConstLowerBound;
+
+      if (Length || LowerBound) {
+        auto *LowerBoundVal =
+            LowerBound
+                ? Builder.CreateIntCast(
+                      EmitScalarExpr(LowerBound), IntPtrTy,
+                      LowerBound->getType()->hasSignedIntegerRepresentation())
+                : llvm::ConstantInt::get(IntPtrTy, ConstLowerBound);
+        auto *LengthVal =
+            Length
+                ? Builder.CreateIntCast(
+                      EmitScalarExpr(Length), IntPtrTy,
+                      Length->getType()->hasSignedIntegerRepresentation())
+                : llvm::ConstantInt::get(IntPtrTy, ConstLength);
+        Idx = Builder.CreateAdd(LowerBoundVal, LengthVal, "lb_add_len",
+                                /*HasNUW=*/false,
+                                !getLangOpts().isSignedOverflowDefined());
+        if (Length && LowerBound) {
+          Idx = Builder.CreateSub(
+              Idx, llvm::ConstantInt::get(IntPtrTy, /*V=*/1), "idx_sub_1",
+              /*HasNUW=*/false, !getLangOpts().isSignedOverflowDefined());
+        }
+      } else
+        Idx = llvm::ConstantInt::get(IntPtrTy, ConstLength + ConstLowerBound);
+    } else {
+      // Idx = ArraySize - 1;
+      if (auto *VAT = C.getAsVariableArrayType(BaseTy)) {
+        Length = VAT->getSizeExpr();
+        if (Length->isIntegerConstantExpr(ConstLength, C))
+          Length = nullptr;
+      } else {
+        auto *CAT = C.getAsConstantArrayType(BaseTy);
+        ConstLength = CAT->getSize();
+      }
+      if (Length) {
+        auto *LengthVal = Builder.CreateIntCast(
+            EmitScalarExpr(Length), IntPtrTy,
+            Length->getType()->hasSignedIntegerRepresentation());
+        Idx = Builder.CreateSub(
+            LengthVal, llvm::ConstantInt::get(IntPtrTy, /*V=*/1), "len_sub_1",
+            /*HasNUW=*/false, !getLangOpts().isSignedOverflowDefined());
+      } else {
+        ConstLength = ConstLength.zextOrTrunc(PointerWidthInBits);
+        --ConstLength;
+        Idx = llvm::ConstantInt::get(IntPtrTy, ConstLength);
+      }
+    }
+  }
+  assert(Idx);
+
+  llvm::Value *Address;
+  CharUnits ArrayAlignment;
+  if (auto *VLA = getContext().getAsVariableArrayType(ResultExprTy)) {
+    // The element count here is the total number of non-VLA elements.
+    llvm::Value *numElements = getVLASize(VLA).first;
+
+    // Effectively, the multiply by the VLA size is part of the GEP.
+    // GEP indexes are signed, and scaling an index isn't permitted to
+    // signed-overflow, so we use the same semantics for our explicit
+    // multiply.  We suppress this if overflow is not undefined behavior.
+    if (getLangOpts().isSignedOverflowDefined()) {
+      Idx = Builder.CreateMul(Idx, numElements);
+      Address = Builder.CreateGEP(Base.getAddress(), Idx, "arrayidx");
+    } else {
+      Idx = Builder.CreateNSWMul(Idx, numElements);
+      Address = Builder.CreateInBoundsGEP(Base.getAddress(), Idx, "arrayidx");
+    }
+  } else if (BaseTy->isConstantArrayType()) {
+    llvm::Value *ArrayPtr = Base.getAddress();
+    llvm::Value *Zero = llvm::ConstantInt::getNullValue(IntPtrTy);
+    llvm::Value *Args[] = {Zero, Idx};
+
+    // Propagate the alignment from the array itself to the result.
+    ArrayAlignment = Base.getAlignment();
+
+    if (getLangOpts().isSignedOverflowDefined())
+      Address = Builder.CreateGEP(ArrayPtr, Args, "arrayidx");
+    else
+      Address = Builder.CreateInBoundsGEP(ArrayPtr, Args, "arrayidx");
+  } else {
+    // The base must be a pointer, which is not an aggregate.  Emit it.
+    if (getLangOpts().isSignedOverflowDefined())
+      Address = Builder.CreateGEP(Base.getAddress(), Idx, "arrayidx");
+    else
+      Address = Builder.CreateInBoundsGEP(Base.getAddress(), Idx, "arrayidx");
+  }
+
+  // Limit the alignment to that of the result type.
+  LValue LV;
+  if (!ArrayAlignment.isZero()) {
+    CharUnits Align = getContext().getTypeAlignInChars(ResultExprTy);
+    ArrayAlignment = std::min(Align, ArrayAlignment);
+    LV = MakeAddrLValue(Address, ResultExprTy, ArrayAlignment);
+  } else
+    LV = MakeNaturalAlignAddrLValue(Address, ResultExprTy);
+
+  LV.getQuals().setAddressSpace(BaseTy.getAddressSpace());
+
+  return LV;
+}
+
 LValue CodeGenFunction::
 EmitExtVectorElementExpr(const ExtVectorElementExpr *E) {
   // Emit the base vector as an l-value.

diff --git a/lib/CodeGen/CGOpenMPRuntime.cpp b/lib/CodeGen/CGOpenMPRuntime.cpp
index f734298360b8910fc68b95aabcb242eaed817673..f4148db7519468f75366c8f1216ab7e521ca04ff 100644 (file)
--- a/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -2091,6 +2091,27 @@ emitTaskPrivateMappingFunction(CodeGenModule &CGM, SourceLocation Loc,
   return TaskPrivatesMap;
 }
 
+llvm::Value *getTypeSize(CodeGenFunction &CGF, QualType Ty) {
+  auto &C = CGF.getContext();
+  llvm::Value *Size;
+  auto SizeInChars = C.getTypeSizeInChars(Ty);
+  if (SizeInChars.isZero()) {
+    // getTypeSizeInChars() returns 0 for a VLA.
+    Size = nullptr;
+    while (auto *VAT = C.getAsVariableArrayType(Ty)) {
+      llvm::Value *ArraySize;
+      std::tie(ArraySize, Ty) = CGF.getVLASize(VAT);
+      Size = Size ? CGF.Builder.CreateNUWMul(Size, ArraySize) : ArraySize;
+    }
+    SizeInChars = C.getTypeSizeInChars(Ty);
+    assert(!SizeInChars.isZero());
+    Size = CGF.Builder.CreateNUWMul(
+        Size, llvm::ConstantInt::get(CGF.SizeTy, SizeInChars.getQuantity()));
+  } else
+    Size = llvm::ConstantInt::get(CGF.SizeTy, SizeInChars.getQuantity());
+  return Size;
+}
+
 static int array_pod_sort_comparator(const PrivateDataTy *P1,
                                      const PrivateDataTy *P2) {
   return P1->first < P2->first ? 1 : (P2->first < P1->first ? -1 : 0);
@@ -2328,19 +2349,30 @@ void CGOpenMPRuntime::emitTaskCall(
     // kmp_depend_info[<Dependences.size()>] deps;
     DependInfo = CGF.CreateMemTemp(KmpDependInfoArrayTy);
     for (unsigned i = 0; i < DependencesNumber; ++i) {
-      auto Addr = CGF.EmitLValue(Dependences[i].second);
-      auto *Size = llvm::ConstantInt::get(
-          CGF.SizeTy,
-          C.getTypeSizeInChars(Dependences[i].second->getType()).getQuantity());
+      auto *E = Dependences[i].second;
+      LValue Addr = CGF.EmitLValue(E);
+      llvm::Value *Size;
+      QualType Ty = E->getType();
+      auto *DestAddr = Addr.getAddress();
+      if (auto *ASE = dyn_cast<OMPArraySectionExpr>(E->IgnoreParenImpCasts())) {
+        LValue UpAddrLVal =
+            CGF.EmitOMPArraySectionExpr(ASE, /*LowerBound=*/false);
+        llvm::Value *UpAddr =
+            CGF.Builder.CreateConstGEP1_32(UpAddrLVal.getAddress(), /*Idx0=*/1);
+        llvm::Value *LowIntPtr =
+            CGF.Builder.CreatePtrToInt(DestAddr, CGM.SizeTy);
+        llvm::Value *UpIntPtr = CGF.Builder.CreatePtrToInt(UpAddr, CGM.SizeTy);
+        Size = CGF.Builder.CreateNUWSub(UpIntPtr, LowIntPtr);
+      } else
+        Size = getTypeSize(CGF, Ty);
       auto Base = CGF.MakeNaturalAlignAddrLValue(
           CGF.Builder.CreateStructGEP(/*Ty=*/nullptr, DependInfo, i),
           KmpDependInfoTy);
       // deps[i].base_addr = &<Dependences[i].second>;
       auto BaseAddrLVal = CGF.EmitLValueForField(
           Base, *std::next(KmpDependInfoRD->field_begin(), BaseAddr));
-      CGF.EmitStoreOfScalar(
-          CGF.Builder.CreatePtrToInt(Addr.getAddress(), CGF.IntPtrTy),
-          BaseAddrLVal);
+      CGF.EmitStoreOfScalar(CGF.Builder.CreatePtrToInt(DestAddr, CGF.IntPtrTy),
+                            BaseAddrLVal);
       // deps[i].len = sizeof(<Dependences[i].second>);
       auto LenLVal = CGF.EmitLValueForField(
           Base, *std::next(KmpDependInfoRD->field_begin(), Len));

diff --git a/lib/CodeGen/CodeGenFunction.h b/lib/CodeGen/CodeGenFunction.h
index 95a512c48e2a22db46598a0309f9eca7d29c5d29..4852d3a393f0f076fdbe36453894d8e45cbea1a3 100644 (file)
--- a/lib/CodeGen/CodeGenFunction.h
+++ b/lib/CodeGen/CodeGenFunction.h
@@ -24,6 +24,7 @@
 #include "clang/AST/CharUnits.h"
 #include "clang/AST/ExprCXX.h"
 #include "clang/AST/ExprObjC.h"
+#include "clang/AST/ExprOpenMP.h"
 #include "clang/AST/Type.h"
 #include "clang/Basic/ABI.h"
 #include "clang/Basic/CapturedStmt.h"
@@ -2412,6 +2413,8 @@ public:
   LValue EmitUnaryOpLValue(const UnaryOperator *E);
   LValue EmitArraySubscriptExpr(const ArraySubscriptExpr *E,
                                 bool Accessed = false);
+  LValue EmitOMPArraySectionExpr(const OMPArraySectionExpr *E,
+                                 bool IsLowerBound = true);
   LValue EmitExtVectorElementExpr(const ExtVectorElementExpr *E);
   LValue EmitMemberExpr(const MemberExpr *E);
   LValue EmitObjCIsaExpr(const ObjCIsaExpr *E);

diff --git a/test/OpenMP/task_codegen.cpp b/test/OpenMP/task_codegen.cpp
index 139ac505a9f65e9a1e935323f30e2b319b9d7508..02141d1cc69eccbec4cf197b0f7a78b03d753ba7 100644 (file)
--- a/test/OpenMP/task_codegen.cpp
+++ b/test/OpenMP/task_codegen.cpp
@@ -24,6 +24,7 @@ int main() {
 // CHECK: [[S:%.+]] = alloca [2 x [[STRUCT_S]]]
   char b;
   S s[2];
+  int arr[10][a];
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T]]* @{{.+}})
 // CHECK: [[B_REF:%.+]] = getelementptr inbounds [[STRUCT_SHAREDS]], [[STRUCT_SHAREDS]]* [[CAPTURES:%.+]], i32 0, i32 0
 // CHECK: store i8* [[B]], i8** [[B_REF]]
@@ -52,14 +53,14 @@ int main() {
 // CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[SHAREDS_REF]], i8* [[BITCAST]], i64 8, i32 8, i1 false)
 // CHECK: [[DESTRUCTORS_REF_PTR:%.+]] = getelementptr inbounds [[KMP_TASK_T]], [[KMP_TASK_T]]* [[TASK_PTR]], i32 0, i32 3
 // CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** [[DESTRUCTORS_REF_PTR]]
-// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: getelementptr inbounds [4 x [[KMP_DEPEND_INFO]]], [4 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: store i64 ptrtoint (i32* @{{.+}} to i64), i64*
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 1
 // CHECK: store i64 4, i64*
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
 // CHECK: store i8 1, i8*
-// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 1
+// CHECK: getelementptr inbounds [4 x [[KMP_DEPEND_INFO]]], [4 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 1
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: ptrtoint i8* [[B]] to i64
 // CHECK: store i64 %{{[^,]+}}, i64*
@@ -67,7 +68,7 @@ int main() {
 // CHECK: store i64 1, i64*
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
 // CHECK: store i8 1, i8*
-// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 2
+// CHECK: getelementptr inbounds [4 x [[KMP_DEPEND_INFO]]], [4 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 2
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: ptrtoint [2 x [[STRUCT_S]]]* [[S]] to i64
 // CHECK: store i64 %{{[^,]+}}, i64*
@@ -75,10 +76,26 @@ int main() {
 // CHECK: store i64 8, i64*
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
 // CHECK: store i8 1, i8*
-// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: [[IDX1:%.+]] = mul nsw i64 0, [[A_VAL:%.+]]
+// CHECK: [[START:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
+// CHECK: [[IDX1:%.+]] = mul nsw i64 9, [[A_VAL]]
+// CHECK: [[END:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
+// CHECK: [[END1:%.+]] = getelementptr i32, i32* [[END]], i32 1
+// CHECK: [[START_INT:%.+]] = ptrtoint i32* [[START]] to i64
+// CHECK: [[END_INT:%.+]] = ptrtoint i32* [[END1]] to i64
+// CHECK: [[SIZEOF:%.+]] = sub nuw i64 [[END_INT]], [[START_INT]]
+// CHECK: getelementptr inbounds [4 x [[KMP_DEPEND_INFO]]], [4 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 3
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: ptrtoint i32* [[START]] to i64
+// CHECK: store i64 %{{[^,]+}}, i64*
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 1
+// CHECK: store i64 [[SIZEOF]], i64*
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
+// CHECK: store i8 1, i8*
+// CHECK: getelementptr inbounds [4 x [[KMP_DEPEND_INFO]]], [4 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: bitcast [[KMP_DEPEND_INFO]]* %{{.+}} to i8*
-// CHECK: call i32 @__kmpc_omp_task_with_deps([[IDENT_T]]* @{{.+}}, i32 [[GTID]], i8* [[ORIG_TASK_PTR]], i32 3, i8* %{{[^,]+}}, i32 0, i8* null)
-#pragma omp task shared(a, s) depend(in : a, b, s)
+// CHECK: call i32 @__kmpc_omp_task_with_deps([[IDENT_T]]* @{{.+}}, i32 [[GTID]], i8* [[ORIG_TASK_PTR]], i32 4, i8* %{{[^,]+}}, i32 0, i8* null)
+#pragma omp task shared(a, s) depend(in : a, b, s, arr[:])
   {
     a = 15;
     s[1].a = 10;
@@ -95,7 +112,7 @@ int main() {
 // CHECK: [[DESTRUCTORS_REF_PTR:%.+]] = getelementptr inbounds [[KMP_TASK_T]]{{.*}}* {{%.+}}, i32 0, i32 3
 // CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** [[DESTRUCTORS_REF_PTR]]
 // CHECK: getelementptr inbounds [2 x [[STRUCT_S]]], [2 x [[STRUCT_S]]]* [[S]], i32 0, i64 0
-// CHECK: getelementptr inbounds [1 x [[KMP_DEPEND_INFO]]], [1 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: getelementptr inbounds [2 x [[KMP_DEPEND_INFO]]], [2 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: ptrtoint [[STRUCT_S]]* %{{.+}} to i64
 // CHECK: store i64 %{{[^,]+}}, i64*
@@ -103,17 +120,39 @@ int main() {
 // CHECK: store i64 4, i64*
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
 // CHECK: store i8 2, i8*
-// CHECK: getelementptr inbounds [1 x [[KMP_DEPEND_INFO]]], [1 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: [[IDX1:%.+]] = mul nsw i64 4, [[A_VAL]]
+// CHECK: [[START:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
+// CHECK: [[B_VAL:%.+]] = load i8, i8* [[B]]
+// CHECK: [[IDX2:%.+]] = sext i8 [[B_VAL]] to i64
+// CHECK: [[START1:%.+]] = getelementptr inbounds i32, i32* [[START]], i64 [[IDX2]]
+// CHECK: [[IDX1:%.+]] = mul nsw i64 9, [[A_VAL]]
+// CHECK: [[END:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
+// CHECK: [[B_VAL:%.+]] = load i8, i8* [[B]]
+// CHECK: [[IDX2:%.+]] = sext i8 [[B_VAL]] to i64
+// CHECK: [[END1:%.+]] = getelementptr inbounds i32, i32* [[END]], i64 [[IDX2]]
+// CHECK: [[END2:%.+]] = getelementptr i32, i32* [[END1]], i32 1
+// CHECK: [[START_INT:%.+]] = ptrtoint i32* [[START1]] to i64
+// CHECK: [[END_INT:%.+]] = ptrtoint i32* [[END2]] to i64
+// CHECK: [[SIZEOF:%.+]] = sub nuw i64 [[END_INT]], [[START_INT]]
+// CHECK: getelementptr inbounds [2 x [[KMP_DEPEND_INFO]]], [2 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 1
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: ptrtoint i32* [[START1]] to i64
+// CHECK: store i64 %{{[^,]+}}, i64*
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 1
+// CHECK: store i64 [[SIZEOF]], i64*
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
+// CHECK: store i8 2, i8*
+// CHECK: getelementptr inbounds [2 x [[KMP_DEPEND_INFO]]], [2 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: bitcast [[KMP_DEPEND_INFO]]* %{{.+}} to i8*
-// CHECK: call i32 @__kmpc_omp_task_with_deps([[IDENT_T]]* @{{.+}}, i32 [[GTID]], i8* [[ORIG_TASK_PTR]], i32 1, i8* %{{[^,]+}}, i32 0, i8* null)
-#pragma omp task untied depend(out : s[0])
+// CHECK: call i32 @__kmpc_omp_task_with_deps([[IDENT_T]]* @{{.+}}, i32 [[GTID]], i8* [[ORIG_TASK_PTR]], i32 2, i8* %{{[^,]+}}, i32 0, i8* null)
+#pragma omp task untied depend(out : s[0], arr[4:][b])
   {
     a = 1;
   }
 // CHECK: [[ORIG_TASK_PTR:%.+]] = call i8* @__kmpc_omp_task_alloc([[IDENT_T]]* @{{.+}}, i32 [[GTID]], i32 3, i64 32, i64 1,
 // CHECK: [[DESTRUCTORS_REF_PTR:%.+]] = getelementptr inbounds [[KMP_TASK_T]]{{.*}}* {{%.+}}, i32 0, i32 3
 // CHECK: store i32 (i32, i8*)* null, i32 (i32, i8*)** [[DESTRUCTORS_REF_PTR]]
-// CHECK: getelementptr inbounds [2 x [[KMP_DEPEND_INFO]]], [2 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: store i64 ptrtoint (i32* @{{.+}} to i64), i64*
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 1
@@ -121,7 +160,7 @@ int main() {
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
 // CHECK: store i8 3, i8*
 // CHECK: getelementptr inbounds [2 x [[STRUCT_S]]], [2 x [[STRUCT_S]]]* [[S]], i32 0, i64 1
-// CHECK: getelementptr inbounds [2 x [[KMP_DEPEND_INFO]]], [2 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 1
+// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 1
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: ptrtoint [[STRUCT_S]]* %{{.+}} to i64
 // CHECK: store i64 %{{[^,]+}}, i64*
@@ -129,10 +168,34 @@ int main() {
 // CHECK: store i64 4, i64*
 // CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
 // CHECK: store i8 3, i8*
-// CHECK: getelementptr inbounds [2 x [[KMP_DEPEND_INFO]]], [2 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: [[IDX1:%.+]] = mul nsw i64 0, [[A_VAL]]
+// CHECK: [[START:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
+// CHECK: [[START1:%.+]] = getelementptr inbounds i32, i32* [[START]], i64 3
+// CHECK: [[NEW_A_VAL:%.+]] = load i32, i32* @{{.+}},
+// CHECK: [[NEW_A_VAL_I64:%.+]] = sext i32 [[NEW_A_VAL]] to i64
+// CHECK: [[SUB:%.+]] = add nsw i64 -1, [[NEW_A_VAL_I64]]
+// CHECK: [[IDX1:%.+]] = mul nsw i64 [[SUB]], [[A_VAL]]
+// CHECK: [[END:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
+// CHECK: [[NEW_A_VAL:%.+]] = load i32, i32* @{{.+}},
+// CHECK: [[NEW_A_VAL_I64:%.+]] = sext i32 [[NEW_A_VAL]] to i64
+// CHECK: [[IDX2:%.+]] = sub nsw i64 [[NEW_A_VAL_I64]], 1
+// CHECK: [[END1:%.+]] = getelementptr inbounds i32, i32* [[END]], i64 [[IDX2]]
+// CHECK: [[END2:%.+]] = getelementptr i32, i32* [[END1]], i32 1
+// CHECK: [[START_INT:%.+]] = ptrtoint i32* [[START1]] to i64
+// CHECK: [[END_INT:%.+]] = ptrtoint i32* [[END2]] to i64
+// CHECK: [[SIZEOF:%.+]] = sub nuw i64 [[END_INT]], [[START_INT]]
+// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 2
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 0
+// CHECK: ptrtoint i32* [[START1]] to i64
+// CHECK: store i64 %{{[^,]+}}, i64*
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 1
+// CHECK: store i64 [[SIZEOF]], i64*
+// CHECK: getelementptr inbounds [[KMP_DEPEND_INFO]], [[KMP_DEPEND_INFO]]* %{{[^,]+}}, i32 0, i32 2
+// CHECK: store i8 3, i8*
+// CHECK: getelementptr inbounds [3 x [[KMP_DEPEND_INFO]]], [3 x [[KMP_DEPEND_INFO]]]* %{{[^,]+}}, i32 0, i32 0
 // CHECK: bitcast [[KMP_DEPEND_INFO]]* %{{.+}} to i8*
-// CHECK: call i32 @__kmpc_omp_task_with_deps([[IDENT_T]]* @{{.+}}, i32 [[GTID]], i8* [[ORIG_TASK_PTR]], i32 2, i8* %{{[^,]+}}, i32 0, i8* null)
-#pragma omp task final(true) depend(inout: a, s[1])
+// CHECK: call i32 @__kmpc_omp_task_with_deps([[IDENT_T]]* @{{.+}}, i32 [[GTID]], i8* [[ORIG_TASK_PTR]], i32 3, i8* %{{[^,]+}}, i32 0, i8* null)
+#pragma omp task final(true) depend(inout: a, s[1], arr[:a][3:])
   {
     a = 2;
   }