[MLIR][OpenMP] Refactoring createTargetData in OMPIRBuilder

[TIFitis]

Summary

Key changes:

• Refactor the createTargetData function to make use of the emitOffloadingArrays and emitOffloadingArraysArgument functions to generate code.
• Added a new emitIfClause helper function to allow handling if clauses in a similar fashion to Clang.
• Updated the MLIR side of code to account for changes to createTargetData.

Depends on D149872

[TIFitis]

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4183–4193

I couldn't come up with a better way for the code generated inside the target data region to use the new AllocaPtrs when accessing map operands marked with the use_device_ptr clause.

Currently I am just comparing the Users list before and after generating the data region and replacing the new uses.

Alternative method would be to use mlir::LLVM::moduleTranslation.mapValue method to temporarily remap the mlir::Value for the use_device_ptr operands to the AllocaPtrs llvm::Value. But mapValue doesn't allow remapping of values. @kiranchandramohan Is there a good reason to prevent remapping of values?

Let me know if anyone has a better way of doing this :)

[jdoerfert]

Conceptually, I am very confused but mostly because this patch conflates things.

I would suggest to split this part:

This also refactors how the map clause is processed and separates the mlir and IRBuilder parts of the code similar to Clang.

it that is a better way of doing things.

Then, let's talks about use_device_ptr.
For one, I am not sure which test even shows the lowering.

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

1899

Everything in this file (should) come with documentation, including (most) members.

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4112

Style, also elsewhere. I know it's annoying that MLIR broke with the LLVM style, but that is what it is now.

4183–4193

This is only a last resort. We should not generate stuff that is wrong and is replaced. However, if we cannot find a better way we can talk about it.

Generally, I would have assumed that the value in the region in marked to be a use_dev_ptr/addr and that we would generate code for that as soon as we generate code for the value.

[TIFitis]

Removed use_dev_ptr clause related code. This patch only refactors the map clause prcoessing now.
Addressed reviewer comments.

[TIFitis]

Updated tests. All tests pass now.

[TIFitis]

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4183–4193

I have tried to explain this a little better in https://reviews.llvm.org/D146648

The region code is generated immediately, its that it still uses the old pointer for it instead of the newly allocated device ptr and this fixes that.

4338

getPointerDereferenceableBytes doesn't fully support Arguments and mostly returns 0.

I am looking at adding support for arguments in a separate patch, in the meantime the tests added in this patch fail because of incorrect @.offload_sizes.

[jdoerfert]

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4338

That function is not the right one for code generation. The sizes need to come from the type + data layout, and potentially the expression used in the map.

[TIFitis]

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4099–4112

I think these should just use the opaque ptr type instead of the ArrI8PtrTy it is currently using, and we can also get rid of the BitCast instructions.

However doing so would break the tests in OpenMPIRBuilderTest.cpp as that unit test still uses typed pointers.

AFAIK the consensus is to completely get rid of typed pointers in llvm 17. Should I updated this code to use opaque pointers and let the test fail until it gets converted to opaque as well?

4338

Thanks for letting me know.

Just to clarify, do you mean it should come from the MLIR::Type?

The llvm::Type doesn't let me get much information because of Opaque pointers.

[TIFitis]

Changed how size is calculated. Updated tests.

[TIFitis]

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

1360–1365

@jdoerfert Is this way of getting the size correct? It seems to work for basic types and arrays which is what we support for now.

[TIFitis]

Updated unit test.

[TIFitis]

Polite request for reviews. Thanks :)

[TIFitis]

ping for reviews. Thanks :)

[TIFitis]

Updated to use opaque pointers, removed BitCast Insts.

[TIFitis]

ping for reviews. Thanks :)

[kiranchandramohan]

Please update the summary to convey all the changes introduced in the patch.

From the tests, it looks like there is a substantial change in the IR. I was hoping this to be an NFC change.

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

1409

Please add the new argument to the documentation above.

mlir/test/Target/LLVMIR/omptarget-llvm.mlir

4–5

Why is the test changed?

[TIFitis]

In D146557#4283149, @kiranchandramohan wrote:

Please update the summary to convey all the changes introduced in the patch.

From the tests, it looks like there is a substantial change in the IR. I was hoping this to be an NFC change.

I have updated the summary highlighting the notable changes along with changes to the generated code.

mlir/test/Target/LLVMIR/omptarget-llvm.mlir

4–5

Moved the map_operand to function parameter to be consistent with other tests. I think it can remain unchanged if you prefer.

[kiranchandramohan]

mlir/test/Target/LLVMIR/omptarget-llvm.mlir

4–5

Consistency can be achieved in a follow-up NFC patch that you can submit without review. If the test can only show the differences of this patch it is easier to review.

[kiranchandramohan]

Would it be possible to break this into a few patches like suggested below or similar?

1. Changed to using inlineConvertOmpRegions to generate target data associated region code inline in same block if possible.

2. Changed to calculating map_operand size from mlir type instead of llvm.

Moved getSizeInBytes function back into OpenACC code as it is no longer shared.

3. Moved most of map_operands related codegen to OMPIRBuilder.

Moved mapper_allocas related codegen to OMPIRBuilder.

4. Changed offload_sizes parameter to a global variable same as offload_maptypes to be more consistent with Clang generated code.

[TIFitis]

Updated tests. Addressed reviewer comments.

[TIFitis]

mlir/test/Target/LLVMIR/omptarget-llvm.mlir

4–5

Thanks for letting me know. I've updated the tests.

[TIFitis]

Cleaned up how IsBegin argument is passed for createTargetData call.

[kiranchandramohan]

In D146557#4292223, @TIFitis wrote:

Cleaned up how IsBegin argument is passed for createTargetData call.

Please submit this directly as an NFC patch.

We should always work towards simple patches that are easy to review. Sometimes a big patch is required to give the full context. But it can then be broken up into small patches for faster review. I have created D149153 for the inlineConvertOmpRegions change.

[TIFitis]

In D146557#4295550, @kiranchandramohan wrote:

In D146557#4292223, @TIFitis wrote:

Cleaned up how IsBegin argument is passed for createTargetData call.

Please submit this directly as an NFC patch.

I have merged it separately.

We should always work towards simple patches that are easy to review. Sometimes a big patch is required to give the full context. But it can then be broken up into small patches for faster review. I have created D149153 for the inlineConvertOmpRegions change.

Thanks for creating your patch. I'll try to make my patches shorter in the future.

Once D149153 is merged I'll rebase this patch.

[TIFitis]

Rebased

[kiranchandramohan]

In D146557#4301955, @TIFitis wrote:

In D146557#4295550, @kiranchandramohan wrote:

In D146557#4292223, @TIFitis wrote:

Cleaned up how IsBegin argument is passed for createTargetData call.

Please submit this directly as an NFC patch.

I have merged it separately.

We should always work towards simple patches that are easy to review. Sometimes a big patch is required to give the full context. But it can then be broken up into small patches for faster review. I have created D149153 for the inlineConvertOmpRegions change.

Thanks for creating your patch. I'll try to make my patches shorter in the future.

Once D149153 is merged I'll rebase this patch.

I have submitted D149153. https://github.com/llvm/llvm-project/commit/1ed522623d95ce7dcf95e711b0f2e3844d2e6be1

[TIFitis]

Rebased

[TIFitis]

@kiranchandramohan I am in the process of submitting a new patch for review which moves the map processing from Clang Codegen to OMPIRBuilder. Once that is moved we can just use that for dealing with the map operands.
Thus putting this patch on hold for now.

[TIFitis]

Done migrating from Clang. This makes use of the clang functions now.

[TIFitis]

Small fix to code comments.

[TIFitis]

Rebased

[TIFitis]

Fix lint messages

[TIFitis]

Try to fix patch application.

[TIFitis]

Moved map processing code from functor to new static function.

[TIFitis]

Renamed from processMapOp to genMapInfos

[jdoerfert]

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

1365

why are these capitalized?

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4085

Can we not use an int for some "type". The 1 and 2 below are magic, use an enum and descriptive words.

4105

No llvm::. Also elsewhere.

4727

This function doesn't make sense to me. For one, I don't know what a "unreal block" is. Nor would I have expected a block with terminator to be silently not touched. It might just be a documentation issue (in the header). I would avoid duplicating the comment here again.

4785

CFG Utils have helpers for these things. Do we not use them on purpose?

[TIFitis]

Addressed reviewer comments.

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4727

I've removed the comment here and updated it in the header file.

The current comment was taken directly from Clang. I've updated it say exactly what it does.

4785

If you are pointing at the SplitBlockAndInsertIfThen like functions in llvm/lib/Transforms/Utils/BasicBlockUtils.cpp then they require the current BB to be well formed with a terminator which it is not in this case.

Earlier we were adding an Unreachable inst and removing it later, however this function I think does a much cleaner job of handling all the cases including region code gen.

Also, it is not apparent in this patch but future patches including some from Jan in review make use of the emitBlock and emitBranch functions directly.

[jdoerfert]

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

1364

This does not mention the deletion stuff, etc.

1887

documentation, what do they mean

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4167

Style, also elsewhere.

4734

you should not just delete BB. eraseFromParent is a better way. That said, it is also weird you would delete something in an "emit" function.

4751–4753

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

1360–1365

I don't know about mlir, if the DL has a "store size" use that.
That said, the size should potentially come from the user/front-end.

[TIFitis]

Addressed reviewer comments.

[TIFitis]

Update MLIR BodyGen callback.

[TIFitis]

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4734

I've changed it.

The emitBlock, emitBranch and emitIfClause functions are ported from Clang btw.
clang/lib/CodeGen/CGStmt.cpp

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

1360–1365

Clang seems to be doing something similar by calculating typeSize from clang::Type.

In my testing this works for what we support for now which are basic arrays, scalars, etc.

For derived types, array subscripts etc. we would need something more sophisticated. I plan on updating this as we add support.

[jdoerfert]

I think this is now style wise pretty good, I still found some potential problems below.

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

1364

This talks about creating a new block, but reading the function it seems it will just place \p BB. Which one is it?

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4134

Can we at least assert MapperFunc is non null if standalone is true. It seems like an invariant almost worth documenting, at least with an assertion + message.

4734

That might be so, but given the name and initial documentation, nobody would have expected this function to delete the block, or at least I would not have.

4754

In the call above you pass the AllocaIP here, now it is both times the saveIP. I doubt this is correct. Allocas will end up in the wrong spot, won't they? Do we have a test for that?

[TIFitis]

Addressed reviewer comments

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

1364

Fixed.

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

4134

Added

4754

The CallBacks were ignoring the AllocaIP passed and getting it from the context instead.
I've updated them to use the CallBack argument instead.

This patch https://reviews.llvm.org/D150860 changes Clang to use this function for TargetDataCall emission which tests most aspects of it.

[jdoerfert]

LG

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

public:
   /// <critical_section_name> + ".var" for "omp critical" directives; 2)
   /// <mangled_name_for_global_var> + ".cache." for cache for threadprivate
   /// variables.
   StringMap<Constant*, BumpPtrAllocator> InternalVars;
 
   /// Computes the size of type in bytes.
   Value *getSizeInBytes(Value *BasePtr);
 
+  // Emit a branch from the current block to the target one if this
+  // was a real block.  If this was just a fall-through block after a
+  // terminator, don't emit it.
+  void EmitBranch(BasicBlock *Target);
+
+  // Create a new Block and place the block after the current block, if
+  // possible, or else at the end of the function.
+  void EmitBlock(BasicBlock *BB, Function *CurFn, bool IsFinished = false);
+
+  /// Emits code for OpenMP 'if' clause using specified \a BodyGenCallbackTy
+  /// Here is the logic:
+  /// if (Cond) {
+  ///   ThenGen();
+  /// } else {
+  ///   ElseGen();
+  /// }
+  void emitIfClause(Value *Cond, BodyGenCallbackTy ThenGen,
+                    BodyGenCallbackTy ElseGen);
+
   /// Create the global variable holding the offload mappings information.
   GlobalVariable *createOffloadMaptypes(SmallVectorImpl<uint64_t> &Mappings,
                                         std::string VarName);
 
   /// Create the global variable holding the offload names information.
   GlobalVariable *
   createOffloadMapnames(SmallVectorImpl<llvm::Constant *> &Names,
                         std::string VarName);
public:
   /// \param MaptypesArg The argument types.
   /// \param MapnamesArg The argument names.
   /// \param MapperAllocas The AllocaInst used for the call.
   /// \param DeviceID Device ID for the call.
   /// \param NumOperands Number of operands in the call.
   void emitMapperCall(const LocationDescription &Loc, Function *MapperFunc,
                       Value *SrcLocInfo, Value *MaptypesArg, Value *MapnamesArg,
                       struct MapperAllocas &MapperAllocas, int64_t DeviceID,
                       unsigned NumOperands);
 
   /// Container for the arguments used to pass data to the runtime library.
   struct TargetDataRTArgs {
     explicit TargetDataRTArgs() {}
     /// The array of base pointer passed to the runtime library.
     Value *BasePointersArray = nullptr;
     /// The array of section pointers passed to the runtime library.
     Value *PointersArray = nullptr;
public:
   /// \param NumTeams Number default teams
   /// \param NumThreads Number default threads
   Constant *registerTargetRegionFunction(TargetRegionEntryInfo &EntryInfo,
                                          Function *OutlinedFunction,
                                          StringRef EntryFnName,
                                          StringRef EntryFnIDName,
                                          int32_t NumTeams, int32_t NumThreads);
 
   /// Generator for '#omp target data'
   ///
   /// \param Loc The location where the target data construct was encountered.
+  /// \param AllocaIP The insertion points to be used for alloca instructions.
   /// \param CodeGenIP The insertion point at which the target directive code
   /// should be placed.
-  /// \param MapTypeFlags BitVector storing the mapType flags for the
-  /// mapOperands.
-  /// \param MapNames Names for the mapOperands.
-  /// \param MapperAllocas Pointers to the AllocInsts for the map clause.
   /// \param IsBegin If true then emits begin mapper call otherwise emits
   /// end mapper call.
   /// \param DeviceID Stores the DeviceID from the device clause.
   /// \param IfCond Value which corresponds to the if clause condition.
-  /// \param ProcessMapOpCB Callback that generates code for the map clause.
-  /// \param BodyGenCB Callback that will generate the region code.
+  /// \param Info Stores all information realted to the Target Data directive.
+  /// \param GenMapInfoCB Callback that populates the MapInfos and returns.
+  /// \param BodyGenCB Optional Callback to generate the region code.
   OpenMPIRBuilder::InsertPointTy createTargetData(
-      const LocationDescription &Loc, OpenMPIRBuilder::InsertPointTy CodeGenIP,
-      SmallVectorImpl<uint64_t> &MapTypeFlags,
-      SmallVectorImpl<Constant *> &MapNames,
-      struct MapperAllocas &MapperAllocas, bool IsBegin, int64_t DeviceID,
-      Value *IfCond, BodyGenCallbackTy ProcessMapOpCB,
-      BodyGenCallbackTy BodyGenCB = {});
+      const LocationDescription &Loc, InsertPointTy AllocaIP,
+      InsertPointTy CodeGenIP, Value *DeviceID, Value *IfCond,
+      TargetDataInfo &Info,
+      function_ref<MapInfosTy &(InsertPointTy CodeGenIP)> GenMapInfoCB,
+      omp::RuntimeFunction *MapperFunc = nullptr,
+      function_ref<InsertPointTy(InsertPointTy CodeGenIP, int BodyGenType)>
+          BodyGenCB = nullptr);
 
   using TargetBodyGenCallbackTy = function_ref<InsertPointTy(
       InsertPointTy AllocaIP, InsertPointTy CodeGenIP)>;
 
   /// Generator for '#omp target'
   ///
   /// \param Loc where the target data construct was encountered.
   /// \param CodeGenIP The insertion point where the call to the outlined

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

Constant *OpenMPIRBuilder::registerTargetRegionFunction(
   auto EntryAddr = createTargetRegionEntryAddr(OutlinedFn, EntryFnName);
   OffloadInfoManager.registerTargetRegionEntryInfo(
       EntryInfo, EntryAddr, OutlinedFnID,
       OffloadEntriesInfoManager::OMPTargetRegionEntryTargetRegion);
   return OutlinedFnID;
 }
 
 OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::createTargetData(
-    const LocationDescription &Loc, OpenMPIRBuilder::InsertPointTy CodeGenIP,
-    SmallVectorImpl<uint64_t> &MapTypeFlags,
-    SmallVectorImpl<Constant *> &MapNames, struct MapperAllocas &MapperAllocas,
-    bool IsBegin, int64_t DeviceID, Value *IfCond,
-    BodyGenCallbackTy ProcessMapOpCB, BodyGenCallbackTy BodyGenCB) {
+    const LocationDescription &Loc, InsertPointTy AllocaIP,
+    InsertPointTy CodeGenIP, Value *DeviceID, Value *IfCond,
+    TargetDataInfo &Info,
+    function_ref<MapInfosTy &(InsertPointTy CodeGenIP)> GenMapInfoCB,
+    omp::RuntimeFunction *MapperFunc,
+    function_ref<InsertPointTy(InsertPointTy CodeGenIP, int BodyGenType)>
+        BodyGenCB) {
   if (!updateToLocation(Loc))
     return InsertPointTy();
 
   Builder.restoreIP(CodeGenIP);
+  bool IsStandAlone = !BodyGenCB;
 
-  // LLVM utilities like blocks with terminators.
-  // The UI acts as a resume point for code insertion after the BodyGen
-  auto *UI = Builder.CreateUnreachable();
-  if (IfCond) {
-    auto *ThenTI =
-        SplitBlockAndInsertIfThen(IfCond, UI, /* Unreachable */ false);
-    ThenTI->getParent()->setName("omp_if.then");
-    Builder.SetInsertPoint(ThenTI);
-  } else {
-    Builder.SetInsertPoint(UI);
-  }
+  // Generate the code for the opening of the data environment. Capture all the
+  // arguments of the runtime call by reference because they are used in the
+  // closing of the region.
+  auto BeginThenGen = [&](InsertPointTy UnusedIP, InsertPointTy CodeGenIP) {
+    emitOffloadingArrays(AllocaIP, Builder.saveIP(),
+                         GenMapInfoCB(Builder.saveIP()), Info,
+                         /*IsNonContiguous=*/true);
+
+    TargetDataRTArgs RTArgs;
+    emitOffloadingArraysArgument(Builder, RTArgs, Info);
 
-  ProcessMapOpCB(Builder.saveIP(), Builder.saveIP());
+    // Emit the number of elements in the offloading arrays.
+    llvm::Value *PointerNum = Builder.getInt32(Info.NumberOfPtrs);
 
+    // Source location for the ident struct
     uint32_t SrcLocStrSize;
     Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
     Value *srcLocInfo = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
 
-  GlobalVariable *MapTypesGV =
-      createOffloadMaptypes(MapTypeFlags, ".offload_maptypes");
-  Value *MapTypesArg = Builder.CreateConstInBoundsGEP2_32(
-      ArrayType::get(Builder.getInt64Ty(), MapTypeFlags.size()), MapTypesGV,
-      /*Idx0=*/0, /*Idx1=*/0);
+    llvm::Value *OffloadingArgs[] = {
+        srcLocInfo,           DeviceID,
+        PointerNum,           RTArgs.BasePointersArray,
+        RTArgs.PointersArray, RTArgs.SizesArray,
+        RTArgs.MapTypesArray, RTArgs.MapNamesArray,
+        RTArgs.MappersArray};
 
-  GlobalVariable *MapNamesGV =
-      createOffloadMapnames(MapNames, ".offload_mapnames");
-  Value *MapNamesArg = Builder.CreateConstInBoundsGEP2_32(
-      ArrayType::get(Builder.getInt8PtrTy(), MapNames.size()), MapNamesGV,
-      /*Idx0=*/0, /*Idx1=*/0);
+    if (IsStandAlone) {
+      Builder.CreateCall(getOrCreateRuntimeFunctionPtr(*MapperFunc),
+                         OffloadingArgs);
+    } else {
+      Function *beginMapperFunc = getOrCreateRuntimeFunctionPtr(
+          omp::OMPRTL___tgt_target_data_begin_mapper);
+
+      Builder.CreateCall(beginMapperFunc, OffloadingArgs);
+
+      // If device pointer privatization is required, emit the body of the
+      // region here. It will have to be duplicated: with and without
+      // privatization.
+      Builder.restoreIP(BodyGenCB(Builder.saveIP(), 1));
+    }
+  };
+
+  // If we need device pointer privatization, we need to emit the body of the
+  // region with no privatization in the 'else' branch of the conditional.
+  // Otherwise, we don't have to do anything.
+  auto BeginElseGen = [&](InsertPointTy UnusedIP, InsertPointTy CodeGenIP) {
+    Builder.restoreIP(BodyGenCB(Builder.saveIP(), 2));
+  };
+
+  // Generate code for the closing of the data region.
+  auto EndThenGen = [&](InsertPointTy UnusedIP, InsertPointTy CodeGenIP) {
+    TargetDataRTArgs RTArgs;
+    emitOffloadingArraysArgument(Builder, RTArgs, Info, /*EmitDebug=*/false,
+                                 /*ForEndCall=*/true);
 
-  Function *beginMapperFunc =
-      getOrCreateRuntimeFunctionPtr(omp::OMPRTL___tgt_target_data_begin_mapper);
+    // Emit the number of elements in the offloading arrays.
+    llvm::Value *PointerNum = Builder.getInt32(Info.NumberOfPtrs);
+
+    // Source location for the ident struct
+    uint32_t SrcLocStrSize;
+    Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
+    Value *srcLocInfo = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
+
+    llvm::Value *OffloadingArgs[] = {
+        srcLocInfo,           DeviceID,
+        PointerNum,           RTArgs.BasePointersArray,
+        RTArgs.PointersArray, RTArgs.SizesArray,
+        RTArgs.MapTypesArray, RTArgs.MapNamesArray,
+        RTArgs.MappersArray};
     Function *endMapperFunc =
         getOrCreateRuntimeFunctionPtr(omp::OMPRTL___tgt_target_data_end_mapper);
 
+    Builder.CreateCall(endMapperFunc, OffloadingArgs);
+  };
+
+  // We don't have to do anything to close the region if the if clause evaluates
+  // to false.
+  auto EndElseGen = [&](InsertPointTy UnusedIP, InsertPointTy CodeGenIP) {};
+
   if (BodyGenCB) {
-    // Create call to start the data region.
-    emitMapperCall(Builder.saveIP(), beginMapperFunc, srcLocInfo, MapTypesArg,
-                   MapNamesArg, MapperAllocas, DeviceID, MapTypeFlags.size());
+    if (IfCond) {
+      emitIfClause(IfCond, BeginThenGen, BeginElseGen);
+    } else {
+      BeginThenGen(AllocaIP, Builder.saveIP());
+    }
 
-    BodyGenCB(Builder.saveIP(), Builder.saveIP());
+    // If we don't require privatization of device pointers, we emit the body in
+    // between the runtime calls. This avoids duplicating the body code.
+    Builder.restoreIP(BodyGenCB(Builder.saveIP(), 3));
 
-    Builder.SetInsertPoint(UI->getParent());
-    // Create call to end the data region.
-    emitMapperCall(Builder.saveIP(), endMapperFunc, srcLocInfo, MapTypesArg,
-                   MapNamesArg, MapperAllocas, DeviceID, MapTypeFlags.size());
+    if (IfCond) {
+      emitIfClause(IfCond, EndThenGen, EndElseGen);
+    } else {
+      EndThenGen(AllocaIP, Builder.saveIP());
+    }
+  } else {
+    if (IfCond) {
+      emitIfClause(IfCond, BeginThenGen, EndElseGen);
     } else {
-    emitMapperCall(Builder.saveIP(), IsBegin ? beginMapperFunc : endMapperFunc,
-                   srcLocInfo, MapTypesArg, MapNamesArg, MapperAllocas,
-                   DeviceID, MapTypeFlags.size());
+      BeginThenGen(AllocaIP, Builder.saveIP());
+    }
   }
 
-  // Update the insertion point and remove the terminator we introduced.
-  Builder.SetInsertPoint(UI->getParent());
-  if (IfCond)
-    UI->getParent()->setName("omp_if.end");
-  UI->eraseFromParent();
   return Builder.saveIP();
 }
 
 static Function *
 createOutlinedFunction(IRBuilderBase &Builder, StringRef FuncName,
                        SmallVectorImpl<Value *> &Inputs,
                        OpenMPIRBuilder::TargetBodyGenCallbackTy &CBFunc) {
   SmallVector<Type *> ParameterTypes;

 Value *OpenMPIRBuilder::getSizeInBytes(Value *BasePtr) {
   LLVMContext &Ctx = Builder.getContext();
   Value *Null = Constant::getNullValue(BasePtr->getType()->getPointerTo());
   Value *SizeGep =
       Builder.CreateGEP(BasePtr->getType(), Null, Builder.getInt32(1));
   Value *SizePtrToInt = Builder.CreatePtrToInt(SizeGep, Type::getInt64Ty(Ctx));
   return SizePtrToInt;
 }
 
 GlobalVariable *
 OpenMPIRBuilder::createOffloadMaptypes(SmallVectorImpl<uint64_t> &Mappings,
                                        std::string VarName) {
   llvm::Constant *MaptypesArrayInit =
       llvm::ConstantDataArray::get(M.getContext(), Mappings);
   auto *MaptypesArrayGlobal = new llvm::GlobalVariable(
       M, MaptypesArrayInit->getType(),
       /*isConstant=*/true, llvm::GlobalValue::PrivateLinkage, MaptypesArrayInit,
void OpenMPIRBuilder::emitOffloadingArrays(
   }
 
   if (!IsNonContiguous || CombinedInfo.NonContigInfo.Offsets.empty() ||
       Info.NumberOfPtrs == 0)
     return;
   emitNonContiguousDescriptor(AllocaIP, CodeGenIP, CombinedInfo, Info);
 }
 
+void OpenMPIRBuilder::EmitBranch(BasicBlock *Target) {
+  // Emit a branch from the current block to the target one if this
+  // was a real block.  If this was just a fall-through block after a
+  // terminator, don't emit it.
+  BasicBlock *CurBB = Builder.GetInsertBlock();
+
+  if (!CurBB || CurBB->getTerminator()) {
+    // If there is no insert point or the previous block is already
+    // terminated, don't touch it.
+  } else {
+    // Otherwise, create a fall-through branch.
+    Builder.CreateBr(Target);
+  }
+
+  Builder.ClearInsertionPoint();
+}
+
+void OpenMPIRBuilder::EmitBlock(BasicBlock *BB, Function *CurFn,
+                                bool IsFinished) {
+  BasicBlock *CurBB = Builder.GetInsertBlock();
+
+  // Fall out of the current block (if necessary).
+  EmitBranch(BB);
+
+  if (IsFinished && BB->use_empty()) {
+    delete BB;
+    return;
+  }
+
+  // Place the block after the current block, if possible, or else at
+  // the end of the function.
+  if (CurBB && CurBB->getParent())
+    CurFn->insert(std::next(CurBB->getIterator()), BB);
+  else
+    CurFn->insert(CurFn->end(), BB);
+  Builder.SetInsertPoint(BB);
+}
+
+void OpenMPIRBuilder::emitIfClause(Value *Cond, BodyGenCallbackTy ThenGen,
+                                   BodyGenCallbackTy ElseGen) {
+  // If the condition constant folds and can be elided, try to avoid emitting
+  // the condition and the dead arm of the if/else.
+  if (isa<ConstantInt>(Cond)) {
+    bool CondConstant =
+        cast<ConstantInt>(Cond)->getUniqueInteger().getSExtValue();
+    if (CondConstant)
+      ThenGen(Builder.saveIP(), Builder.saveIP());
+    else
+      ElseGen(Builder.saveIP(), Builder.saveIP());
+    return;
+  }
+
+  Function *CurFn = Builder.GetInsertBlock()->getParent();
+
+  // Otherwise, the condition did not fold, or we couldn't elide it.  Just
+  // emit the conditional branch.
+  BasicBlock *ThenBlock = BasicBlock::Create(M.getContext(), "omp_if.then");
+  BasicBlock *ElseBlock = BasicBlock::Create(M.getContext(), "omp_if.else");
+  BasicBlock *ContBlock = BasicBlock::Create(M.getContext(), "omp_if.end");
+  Builder.CreateCondBr(Cond, ThenBlock, ElseBlock);
+
+  // Emit the 'then' code.
+  EmitBlock(ThenBlock, CurFn);
+  ThenGen(Builder.saveIP(), Builder.saveIP());
+  EmitBranch(ContBlock);
+  // Emit the 'else' code if present.
+  // There is no need to emit line number for unconditional branch.
+  EmitBlock(ElseBlock, CurFn);
+  ElseGen(Builder.saveIP(), Builder.saveIP());
+  // There is no need to emit line number for unconditional branch.
+  EmitBranch(ContBlock);
+  // Emit the continuation block for code after the if.
+  EmitBlock(ContBlock, CurFn, /*IsFinished=*/true);
+}
+
 bool OpenMPIRBuilder::checkAndEmitFlushAfterAtomic(
     const LocationDescription &Loc, llvm::AtomicOrdering AO, AtomicKind AK) {
   assert(!(AO == AtomicOrdering::NotAtomic ||
            AO == llvm::AtomicOrdering::Unordered) &&
          "Unexpected Atomic Ordering.");
 
   bool Flush = false;
   llvm::AtomicOrdering FlushAO = AtomicOrdering::Monotonic;

llvm/unittests/Frontend/OpenMPIRBuilderTest.cpp

TEST_F(OpenMPIRBuilderTest, EmitMapperCall) {
   EXPECT_TRUE(MapperCall->getOperand(8)->getType()->isPointerTy());
 }
 
 TEST_F(OpenMPIRBuilderTest, TargetEnterData) {
   OpenMPIRBuilder OMPBuilder(*M);
   OMPBuilder.initialize();
   F->setName("func");
   IRBuilder<> Builder(BB);
-
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
-  unsigned NumDataOperands = 1;
   int64_t DeviceID = 2;
-  struct OpenMPIRBuilder::MapperAllocas MapperAllocas;
-  SmallVector<uint64_t> MapTypeFlagsTo = {1};
-  SmallVector<Constant *> MapNames;
-  auto *I8PtrTy = Builder.getInt8PtrTy();
-  auto *ArrI8PtrTy = ArrayType::get(I8PtrTy, NumDataOperands);
-  auto *I64Ty = Builder.getInt64Ty();
-  auto *ArrI64Ty = ArrayType::get(I64Ty, NumDataOperands);
 
   AllocaInst *Val1 =
       Builder.CreateAlloca(Builder.getInt32Ty(), Builder.getInt64(1));
   ASSERT_NE(Val1, nullptr);
 
   IRBuilder<>::InsertPoint AllocaIP(&F->getEntryBlock(),
                                     F->getEntryBlock().getFirstInsertionPt());
-  OMPBuilder.createMapperAllocas(Builder.saveIP(), AllocaIP, NumDataOperands,
-                                 MapperAllocas);
 
+  llvm::OpenMPIRBuilder::MapInfosTy CombinedInfo;
   using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
-  auto ProcessMapOpCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
-    Value *DataValue = Val1;
-    Value *DataPtrBase;
-    Value *DataPtr;
-    DataPtrBase = DataValue;
-    DataPtr = DataValue;
-    Builder.restoreIP(CodeGenIP);
+  auto GenMapInfoCB =
+      [&](InsertPointTy codeGenIP) -> llvm::OpenMPIRBuilder::MapInfosTy & {
+    // Get map clause information.
+    Builder.restoreIP(codeGenIP);
 
-    Value *Null = Constant::getNullValue(DataValue->getType()->getPointerTo());
-    Value *SizeGep =
-        Builder.CreateGEP(DataValue->getType(), Null, Builder.getInt32(1));
-    Value *SizePtrToInt = Builder.CreatePtrToInt(SizeGep, I64Ty);
-
-    Value *PtrBaseGEP =
-        Builder.CreateInBoundsGEP(ArrI8PtrTy, MapperAllocas.ArgsBase,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Value *PtrBaseCast = Builder.CreateBitCast(
-        PtrBaseGEP, DataPtrBase->getType()->getPointerTo());
-    Builder.CreateStore(DataPtrBase, PtrBaseCast);
-    Value *PtrGEP =
-        Builder.CreateInBoundsGEP(ArrI8PtrTy, MapperAllocas.Args,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Value *PtrCast =
-        Builder.CreateBitCast(PtrGEP, DataPtr->getType()->getPointerTo());
-    Builder.CreateStore(DataPtr, PtrCast);
-    Value *SizeGEP =
-        Builder.CreateInBoundsGEP(ArrI64Ty, MapperAllocas.ArgSizes,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Builder.CreateStore(SizePtrToInt, SizeGEP);
-  };
+    CombinedInfo.BasePointers.emplace_back(Val1);
+    CombinedInfo.Pointers.emplace_back(Val1);
+    CombinedInfo.Sizes.emplace_back(Builder.getInt64(4));
+    CombinedInfo.Types.emplace_back(llvm::omp::OpenMPOffloadMappingFlags(1));
+    uint32_t temp;
+    CombinedInfo.Names.emplace_back(
+        OMPBuilder.getOrCreateSrcLocStr("unknown", temp));
+    return CombinedInfo;
+  };
+
+  llvm::OpenMPIRBuilder::TargetDataInfo Info(
+      /*RequiresDevicePointerInfo=*/false,
+      /*SeparateBeginEndCalls=*/true);
 
+  OMPBuilder.Config.setIsTargetCodegen(true);
+
+  llvm::omp::RuntimeFunction RTLFunc = OMPRTL___tgt_target_data_begin_mapper;
   Builder.restoreIP(OMPBuilder.createTargetData(
-      Loc, Builder.saveIP(), MapTypeFlagsTo, MapNames, MapperAllocas,
-      /* IsBegin= */ true, DeviceID, /* IfCond= */ nullptr, ProcessMapOpCB));
+      Loc, AllocaIP, Builder.saveIP(), Builder.getInt64(DeviceID),
+      /* IfCond= */ nullptr, Info, GenMapInfoCB, &RTLFunc));
 
   CallInst *TargetDataCall = dyn_cast<CallInst>(&BB->back());
   EXPECT_NE(TargetDataCall, nullptr);
   EXPECT_EQ(TargetDataCall->arg_size(), 9U);
   EXPECT_EQ(TargetDataCall->getCalledFunction()->getName(),
             "__tgt_target_data_begin_mapper");
   EXPECT_TRUE(TargetDataCall->getOperand(1)->getType()->isIntegerTy(64));
   EXPECT_TRUE(TargetDataCall->getOperand(2)->getType()->isIntegerTy(32));
   EXPECT_TRUE(TargetDataCall->getOperand(8)->getType()->isPointerTy());
 
   Builder.CreateRetVoid();
   EXPECT_FALSE(verifyModule(*M, &errs()));
 }
 
 TEST_F(OpenMPIRBuilderTest, TargetExitData) {
   OpenMPIRBuilder OMPBuilder(*M);
   OMPBuilder.initialize();
   F->setName("func");
   IRBuilder<> Builder(BB);
-
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
-  unsigned NumDataOperands = 1;
   int64_t DeviceID = 2;
-  struct OpenMPIRBuilder::MapperAllocas MapperAllocas;
-  SmallVector<uint64_t> MapTypeFlagsFrom = {2};
-  SmallVector<Constant *> MapNames;
-  auto *I8PtrTy = Builder.getInt8PtrTy();
-  auto *ArrI8PtrTy = ArrayType::get(I8PtrTy, NumDataOperands);
-  auto *I64Ty = Builder.getInt64Ty();
-  auto *ArrI64Ty = ArrayType::get(I64Ty, NumDataOperands);
 
   AllocaInst *Val1 =
       Builder.CreateAlloca(Builder.getInt32Ty(), Builder.getInt64(1));
   ASSERT_NE(Val1, nullptr);
 
   IRBuilder<>::InsertPoint AllocaIP(&F->getEntryBlock(),
                                     F->getEntryBlock().getFirstInsertionPt());
-  OMPBuilder.createMapperAllocas(Builder.saveIP(), AllocaIP, NumDataOperands,
-                                 MapperAllocas);
 
+  llvm::OpenMPIRBuilder::MapInfosTy CombinedInfo;
   using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
-  auto ProcessMapOpCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
-    Value *DataValue = Val1;
-    Value *DataPtrBase;
-    Value *DataPtr;
-    DataPtrBase = DataValue;
-    DataPtr = DataValue;
-    Builder.restoreIP(CodeGenIP);
+  auto GenMapInfoCB =
+      [&](InsertPointTy codeGenIP) -> llvm::OpenMPIRBuilder::MapInfosTy & {
+    // Get map clause information.
+    Builder.restoreIP(codeGenIP);
 
-    Value *Null = Constant::getNullValue(DataValue->getType()->getPointerTo());
-    Value *SizeGep =
-        Builder.CreateGEP(DataValue->getType(), Null, Builder.getInt32(1));
-    Value *SizePtrToInt = Builder.CreatePtrToInt(SizeGep, I64Ty);
-
-    Value *PtrBaseGEP =
-        Builder.CreateInBoundsGEP(ArrI8PtrTy, MapperAllocas.ArgsBase,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Value *PtrBaseCast = Builder.CreateBitCast(
-        PtrBaseGEP, DataPtrBase->getType()->getPointerTo());
-    Builder.CreateStore(DataPtrBase, PtrBaseCast);
-    Value *PtrGEP =
-        Builder.CreateInBoundsGEP(ArrI8PtrTy, MapperAllocas.Args,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Value *PtrCast =
-        Builder.CreateBitCast(PtrGEP, DataPtr->getType()->getPointerTo());
-    Builder.CreateStore(DataPtr, PtrCast);
-    Value *SizeGEP =
-        Builder.CreateInBoundsGEP(ArrI64Ty, MapperAllocas.ArgSizes,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Builder.CreateStore(SizePtrToInt, SizeGEP);
-  };
+    CombinedInfo.BasePointers.emplace_back(Val1);
+    CombinedInfo.Pointers.emplace_back(Val1);
+    CombinedInfo.Sizes.emplace_back(Builder.getInt64(4));
+    CombinedInfo.Types.emplace_back(llvm::omp::OpenMPOffloadMappingFlags(2));
+    uint32_t temp;
+    CombinedInfo.Names.emplace_back(
+        OMPBuilder.getOrCreateSrcLocStr("unknown", temp));
+    return CombinedInfo;
+  };
+
+  llvm::OpenMPIRBuilder::TargetDataInfo Info(
+      /*RequiresDevicePointerInfo=*/false,
+      /*SeparateBeginEndCalls=*/true);
+
+  OMPBuilder.Config.setIsTargetCodegen(true);
 
+  llvm::omp::RuntimeFunction RTLFunc = OMPRTL___tgt_target_data_end_mapper;
   Builder.restoreIP(OMPBuilder.createTargetData(
-      Loc, Builder.saveIP(), MapTypeFlagsFrom, MapNames, MapperAllocas,
-      /* IsBegin= */ false, DeviceID, /* IfCond= */ nullptr, ProcessMapOpCB));
+      Loc, AllocaIP, Builder.saveIP(), Builder.getInt64(DeviceID),
+      /* IfCond= */ nullptr, Info, GenMapInfoCB, &RTLFunc));
 
   CallInst *TargetDataCall = dyn_cast<CallInst>(&BB->back());
   EXPECT_NE(TargetDataCall, nullptr);
   EXPECT_EQ(TargetDataCall->arg_size(), 9U);
   EXPECT_EQ(TargetDataCall->getCalledFunction()->getName(),
             "__tgt_target_data_end_mapper");
   EXPECT_TRUE(TargetDataCall->getOperand(1)->getType()->isIntegerTy(64));
   EXPECT_TRUE(TargetDataCall->getOperand(2)->getType()->isIntegerTy(32));
   EXPECT_TRUE(TargetDataCall->getOperand(8)->getType()->isPointerTy());
 
   Builder.CreateRetVoid();
   EXPECT_FALSE(verifyModule(*M, &errs()));
 }
 
 TEST_F(OpenMPIRBuilderTest, TargetDataRegion) {
   OpenMPIRBuilder OMPBuilder(*M);
   OMPBuilder.initialize();
   F->setName("func");
   IRBuilder<> Builder(BB);
-
   OpenMPIRBuilder::LocationDescription Loc({Builder.saveIP(), DL});
 
-  unsigned NumDataOperands = 1;
   int64_t DeviceID = 2;
-  struct OpenMPIRBuilder::MapperAllocas MapperAllocas;
-  SmallVector<uint64_t> MapTypeFlagsToFrom = {3};
-  SmallVector<Constant *> MapNames;
-  auto *I8PtrTy = Builder.getInt8PtrTy();
-  auto *ArrI8PtrTy = ArrayType::get(I8PtrTy, NumDataOperands);
-  auto *I64Ty = Builder.getInt64Ty();
-  auto *ArrI64Ty = ArrayType::get(I64Ty, NumDataOperands);
 
   AllocaInst *Val1 =
       Builder.CreateAlloca(Builder.getInt32Ty(), Builder.getInt64(1));
   ASSERT_NE(Val1, nullptr);
 
   IRBuilder<>::InsertPoint AllocaIP(&F->getEntryBlock(),
                                     F->getEntryBlock().getFirstInsertionPt());
-  OMPBuilder.createMapperAllocas(Builder.saveIP(), AllocaIP, NumDataOperands,
-                                 MapperAllocas);
 
+  llvm::OpenMPIRBuilder::MapInfosTy CombinedInfo;
   using InsertPointTy = OpenMPIRBuilder::InsertPointTy;
-  auto ProcessMapOpCB = [&](InsertPointTy AllocaIP, InsertPointTy CodeGenIP) {
-    Value *DataValue = Val1;
-    Value *DataPtrBase;
-    Value *DataPtr;
-    DataPtrBase = DataValue;
-    DataPtr = DataValue;
-    Builder.restoreIP(CodeGenIP);
+  auto GenMapInfoCB =
+      [&](InsertPointTy codeGenIP) -> llvm::OpenMPIRBuilder::MapInfosTy & {
+    // Get map clause information.
+    Builder.restoreIP(codeGenIP);
 
-    Value *Null = Constant::getNullValue(DataValue->getType()->getPointerTo());
-    Value *SizeGep =
-        Builder.CreateGEP(DataValue->getType(), Null, Builder.getInt32(1));
-    Value *SizePtrToInt = Builder.CreatePtrToInt(SizeGep, I64Ty);
-
-    Value *PtrBaseGEP =
-        Builder.CreateInBoundsGEP(ArrI8PtrTy, MapperAllocas.ArgsBase,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Value *PtrBaseCast = Builder.CreateBitCast(
-        PtrBaseGEP, DataPtrBase->getType()->getPointerTo());
-    Builder.CreateStore(DataPtrBase, PtrBaseCast);
-    Value *PtrGEP =
-        Builder.CreateInBoundsGEP(ArrI8PtrTy, MapperAllocas.Args,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Value *PtrCast =
-        Builder.CreateBitCast(PtrGEP, DataPtr->getType()->getPointerTo());
-    Builder.CreateStore(DataPtr, PtrCast);
-    Value *SizeGEP =
-        Builder.CreateInBoundsGEP(ArrI64Ty, MapperAllocas.ArgSizes,
-                                  {Builder.getInt32(0), Builder.getInt32(0)});
-    Builder.CreateStore(SizePtrToInt, SizeGEP);
-  };
+    CombinedInfo.BasePointers.emplace_back(Val1);
+    CombinedInfo.Pointers.emplace_back(Val1);
+    CombinedInfo.Sizes.emplace_back(Builder.getInt64(4));
+    CombinedInfo.Types.emplace_back(llvm::omp::OpenMPOffloadMappingFlags(3));
+    uint32_t temp;
+    CombinedInfo.Names.emplace_back(
+        OMPBuilder.getOrCreateSrcLocStr("unknown", temp));
+    return CombinedInfo;
+  };
+
+  llvm::OpenMPIRBuilder::TargetDataInfo Info(
+      /*RequiresDevicePointerInfo=*/false,
+      /*SeparateBeginEndCalls=*/true);
 
-  auto BodyCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP) {
-    Builder.restoreIP(codeGenIP);
-    auto *SI = Builder.CreateStore(Builder.getInt32(99), Val1);
-    auto *newBB = SplitBlock(Builder.GetInsertBlock(), SI);
-    Builder.SetInsertPoint(newBB);
-    auto *UI = &Builder.GetInsertBlock()->back();
-    SplitBlock(Builder.GetInsertBlock(), UI);
+  OMPBuilder.Config.setIsTargetCodegen(true);
+
+  auto BodyCB = [&](InsertPointTy CodeGenIP, int BodyGenType) {
+    if (BodyGenType == 3) {
+      Builder.restoreIP(CodeGenIP);
+      CallInst *TargetDataCall = dyn_cast<CallInst>(&BB->back());
+      EXPECT_NE(TargetDataCall, nullptr);
+      EXPECT_EQ(TargetDataCall->arg_size(), 9U);
+      EXPECT_EQ(TargetDataCall->getCalledFunction()->getName(),
+                "__tgt_target_data_begin_mapper");
+      EXPECT_TRUE(TargetDataCall->getOperand(1)->getType()->isIntegerTy(64));
+      EXPECT_TRUE(TargetDataCall->getOperand(2)->getType()->isIntegerTy(32));
+      EXPECT_TRUE(TargetDataCall->getOperand(8)->getType()->isPointerTy());
+      Builder.restoreIP(CodeGenIP);
+      Builder.CreateStore(Builder.getInt32(99), Val1);
+    }
+    return Builder.saveIP();
   };
 
   Builder.restoreIP(OMPBuilder.createTargetData(
-      Loc, Builder.saveIP(), MapTypeFlagsToFrom, MapNames, MapperAllocas,
-      /* IsBegin= */ false, DeviceID, /* IfCond= */ nullptr, ProcessMapOpCB,
-      BodyCB));
+      Loc, AllocaIP, Builder.saveIP(), Builder.getInt64(DeviceID),
+      /* IfCond= */ nullptr, Info, GenMapInfoCB, nullptr, BodyCB));
 
-  CallInst *TargetDataCall =
-      dyn_cast<CallInst>(&Builder.GetInsertBlock()->back());
+  CallInst *TargetDataCall = dyn_cast<CallInst>(&BB->back());
   EXPECT_NE(TargetDataCall, nullptr);
   EXPECT_EQ(TargetDataCall->arg_size(), 9U);
   EXPECT_EQ(TargetDataCall->getCalledFunction()->getName(),
             "__tgt_target_data_end_mapper");
   EXPECT_TRUE(TargetDataCall->getOperand(1)->getType()->isIntegerTy(64));
   EXPECT_TRUE(TargetDataCall->getOperand(2)->getType()->isIntegerTy(32));
   EXPECT_TRUE(TargetDataCall->getOperand(8)->getType()->isPointerTy());
 

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

convertOmpThreadprivate(Operation &opInst, llvm::IRBuilderBase &builder,
   llvm::Value *callInst =
       moduleTranslation.getOpenMPBuilder()->createCachedThreadPrivate(
           ompLoc, data, size, cacheName);
   llvm::Value *result = builder.CreateBitCast(callInst, globalValue->getType());
   moduleTranslation.mapValue(opInst.getResult(0), result);
   return success();
 }
 
-/// Process MapOperands for Target Data directives.
-static LogicalResult processMapOperand(
-    llvm::IRBuilderBase &builder, LLVM::ModuleTranslation &moduleTranslation,
-    const SmallVector<Value> &mapOperands, const ArrayAttr &mapTypes,
-    SmallVector<uint64_t> &mapTypeFlags,
-    SmallVectorImpl<llvm::Constant *> &mapNames,
-    struct llvm::OpenMPIRBuilder::MapperAllocas &mapperAllocas) {
-  auto numMapOperands = mapOperands.size();
-  llvm::OpenMPIRBuilder *ompBuilder = moduleTranslation.getOpenMPBuilder();
-  llvm::PointerType *i8PtrTy = builder.getInt8PtrTy();
-  llvm::ArrayType *arrI8PtrTy = llvm::ArrayType::get(i8PtrTy, numMapOperands);
-  llvm::IntegerType *i64Ty = builder.getInt64Ty();
-  llvm::ArrayType *arrI64Ty = llvm::ArrayType::get(i64Ty, numMapOperands);
+int64_t getSizeInBytes(DataLayout &DL, const mlir::Type &type) {
+  if (isa<LLVM::LLVMPointerType>(type))
+    return DL.getTypeSize(cast<LLVM::LLVMPointerType>(type).getElementType());
+
+  return 0;
+}
 
+static void genMapInfos(llvm::IRBuilderBase &builder,
+                        LLVM::ModuleTranslation &moduleTranslation,
+                        DataLayout &DL,
+                        llvm::OpenMPIRBuilder::MapInfosTy &combinedInfo,
+                        const SmallVector<Value> &mapOperands,
+                        const ArrayAttr &mapTypes) {
+  // Get map clause information.
+  llvm::OpenMPIRBuilder *ompBuilder = moduleTranslation.getOpenMPBuilder();
   unsigned index = 0;
   for (const auto &mapOp : mapOperands) {
-    const auto &mapTypeOp = mapTypes[index];
-
-    llvm::Value *mapOpValue = moduleTranslation.lookupValue(mapOp);
-    llvm::Value *mapOpPtrBase;
-    llvm::Value *mapOpPtr;
-    llvm::Value *mapOpSize;
-
-    if (isa<LLVM::LLVMPointerType>(mapOp.getType())) {
-      mapOpPtrBase = mapOpValue;
-      mapOpPtr = mapOpValue;
-      mapOpSize = ompBuilder->getSizeInBytes(mapOpValue);
-    } else {
-      return failure();
+    if (!mapOp.getType().isa<LLVM::LLVMPointerType>()) {
+      // TODO: Only LLVMPointerTypes are handled.
+      combinedInfo.BasePointers.clear();
+      combinedInfo.Pointers.clear();
+      combinedInfo.Sizes.clear();
+      combinedInfo.Types.clear();
+      combinedInfo.Names.clear();
+      return;
     }
 
-    // Store base pointer extracted from operand into the i-th position of
-    // argBase.
-    llvm::Value *ptrBaseGEP = builder.CreateInBoundsGEP(
-        arrI8PtrTy, mapperAllocas.ArgsBase,
-        {builder.getInt32(0), builder.getInt32(index)});
-    llvm::Value *ptrBaseCast = builder.CreateBitCast(
-        ptrBaseGEP, mapOpPtrBase->getType()->getPointerTo());
-    builder.CreateStore(mapOpPtrBase, ptrBaseCast);
-
-    // Store pointer extracted from operand into the i-th position of args.
-    llvm::Value *ptrGEP = builder.CreateInBoundsGEP(
-        arrI8PtrTy, mapperAllocas.Args,
-        {builder.getInt32(0), builder.getInt32(index)});
-    llvm::Value *ptrCast =
-        builder.CreateBitCast(ptrGEP, mapOpPtr->getType()->getPointerTo());
-    builder.CreateStore(mapOpPtr, ptrCast);
-
-    // Store size extracted from operand into the i-th position of argSizes.
-    llvm::Value *sizeGEP = builder.CreateInBoundsGEP(
-        arrI64Ty, mapperAllocas.ArgSizes,
-        {builder.getInt32(0), builder.getInt32(index)});
-    builder.CreateStore(mapOpSize, sizeGEP);
-
-    mapTypeFlags.push_back(dyn_cast<mlir::IntegerAttr>(mapTypeOp).getInt());
-    llvm::Constant *mapName =
-        mlir::LLVM::createMappingInformation(mapOp.getLoc(), *ompBuilder);
-    mapNames.push_back(mapName);
-    ++index;
+    llvm::Value *mapOpValue = moduleTranslation.lookupValue(mapOp);
+    combinedInfo.BasePointers.emplace_back(mapOpValue);
+    combinedInfo.Pointers.emplace_back(mapOpValue);
+    combinedInfo.Names.emplace_back(
+        mlir::LLVM::createMappingInformation(mapOp.getLoc(), *ompBuilder));
+    combinedInfo.Types.emplace_back(llvm::omp::OpenMPOffloadMappingFlags(
+        mapTypes[index].dyn_cast<mlir::IntegerAttr>().getInt()));
+    combinedInfo.Sizes.emplace_back(
+        builder.getInt64(getSizeInBytes(DL, mapOp.getType())));
+    index++;
   }
-
-  return success();
 }
 
 static LogicalResult
 convertOmpTargetData(Operation *op, llvm::IRBuilderBase &builder,
                      LLVM::ModuleTranslation &moduleTranslation) {
-  unsigned numMapOperands;
   llvm::Value *ifCond = nullptr;
   int64_t deviceID = llvm::omp::OMP_DEVICEID_UNDEF;
   SmallVector<Value> mapOperands;
   ArrayAttr mapTypes;
+  llvm::omp::RuntimeFunction RTLFn;
+  DataLayout DL = DataLayout(op->getParentOfType<ModuleOp>());
 
   llvm::OpenMPIRBuilder *ompBuilder = moduleTranslation.getOpenMPBuilder();
 
   LogicalResult result =
       llvm::TypeSwitch<Operation *, LogicalResult>(op)
           .Case([&](omp::DataOp dataOp) {
             if (!dataOp.getUseDeviceAddr().empty() ||
                 !dataOp.getUseDevicePtr().empty())
               return failure();
 
             if (auto ifExprVar = dataOp.getIfExpr())
               ifCond = moduleTranslation.lookupValue(ifExprVar);
 
             if (auto devId = dataOp.getDevice())
               if (auto constOp = mlir::dyn_cast<mlir::LLVM::ConstantOp>(
                       devId.getDefiningOp()))
                 if (auto intAttr =
                         dyn_cast<mlir::IntegerAttr>(constOp.getValue()))
                   deviceID = intAttr.getInt();
 
-            numMapOperands = dataOp.getMapOperands().size();
             mapOperands = dataOp.getMapOperands();
             mapTypes = dataOp.getMapTypes();
             return success();
           })
           .Case([&](omp::EnterDataOp enterDataOp) {
             if (enterDataOp.getNowait())
               return failure();
 
             if (auto ifExprVar = enterDataOp.getIfExpr())
               ifCond = moduleTranslation.lookupValue(ifExprVar);
 
             if (auto devId = enterDataOp.getDevice())
               if (auto constOp = mlir::dyn_cast<mlir::LLVM::ConstantOp>(
                       devId.getDefiningOp()))
                 if (auto intAttr =
                         dyn_cast<mlir::IntegerAttr>(constOp.getValue()))
                   deviceID = intAttr.getInt();
-
-            numMapOperands = enterDataOp.getMapOperands().size();
+            RTLFn = llvm::omp::OMPRTL___tgt_target_data_begin_mapper;
             mapOperands = enterDataOp.getMapOperands();
             mapTypes = enterDataOp.getMapTypes();
             return success();
           })
           .Case([&](omp::ExitDataOp exitDataOp) {
             if (exitDataOp.getNowait())
               return failure();
 
             if (auto ifExprVar = exitDataOp.getIfExpr())
               ifCond = moduleTranslation.lookupValue(ifExprVar);
 
             if (auto devId = exitDataOp.getDevice())
               if (auto constOp = mlir::dyn_cast<mlir::LLVM::ConstantOp>(
                       devId.getDefiningOp()))
                 if (auto intAttr =
                         dyn_cast<mlir::IntegerAttr>(constOp.getValue()))
                   deviceID = intAttr.getInt();
 
-            numMapOperands = exitDataOp.getMapOperands().size();
+            RTLFn = llvm::omp::OMPRTL___tgt_target_data_end_mapper;
             mapOperands = exitDataOp.getMapOperands();
             mapTypes = exitDataOp.getMapTypes();
             return success();
           })
           .Default([&](Operation *op) {
             return op->emitError("unsupported OpenMP operation: ")
                    << op->getName();
           });
 
   if (failed(result))
     return failure();
 
-  llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
-  llvm::OpenMPIRBuilder::InsertPointTy allocaIP =
-      findAllocaInsertPoint(builder, moduleTranslation);
-
-  struct llvm::OpenMPIRBuilder::MapperAllocas mapperAllocas;
-  SmallVector<uint64_t> mapTypeFlags;
-  SmallVector<llvm::Constant *> mapNames;
-  ompBuilder->createMapperAllocas(builder.saveIP(), allocaIP, numMapOperands,
-                                  mapperAllocas);
-
   using InsertPointTy = llvm::OpenMPIRBuilder::InsertPointTy;
-  LogicalResult processMapOpStatus = success();
-  auto processMapOpCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP) {
+
+  // Fill up the arrays with all the mapped variables.
+  llvm::OpenMPIRBuilder::MapInfosTy combinedInfo;
+  auto genMapInfoCB =
+      [&](InsertPointTy codeGenIP) -> llvm::OpenMPIRBuilder::MapInfosTy & {
     builder.restoreIP(codeGenIP);
-    processMapOpStatus =
-        processMapOperand(builder, moduleTranslation, mapOperands, mapTypes,
-                          mapTypeFlags, mapNames, mapperAllocas);
+    genMapInfos(builder, moduleTranslation, DL, combinedInfo, mapOperands,
+                mapTypes);
+    return combinedInfo;
   };
 
   LogicalResult bodyGenStatus = success();
-  auto bodyCB = [&](InsertPointTy allocaIP, InsertPointTy codeGenIP) {
+  auto bodyGenCB = [&](InsertPointTy codeGenIP, int bodyGenType) {
+    if (bodyGenType == 3) {
       // DataOp has only one region associated with it.
       auto &region = cast<omp::DataOp>(op).getRegion();
       builder.restoreIP(codeGenIP);
-    bodyGenStatus = inlineConvertOmpRegions(region, "omp.data.region", builder,
-                                            moduleTranslation);
+      bodyGenStatus = inlineConvertOmpRegions(region, "omp.data.region",
+                                              builder, moduleTranslation);
+    }
+    return builder.saveIP();
   };
 
+  llvm::OpenMPIRBuilder::LocationDescription ompLoc(builder);
+  llvm::OpenMPIRBuilder::InsertPointTy allocaIP =
+      findAllocaInsertPoint(builder, moduleTranslation);
+
+  // TODO: Add support for DevicePointerInfo
+  llvm::OpenMPIRBuilder::TargetDataInfo info(
+      /*RequiresDevicePointerInfo=*/false,
+      /*SeparateBeginEndCalls=*/true);
   if (isa<omp::DataOp>(op)) {
     builder.restoreIP(ompBuilder->createTargetData(
-        ompLoc, builder.saveIP(), mapTypeFlags, mapNames, mapperAllocas,
-        /*IsBegin=*/false, deviceID, ifCond, processMapOpCB, bodyCB));
+        ompLoc, allocaIP, builder.saveIP(), builder.getInt64(deviceID), ifCond,
+        info, genMapInfoCB, nullptr, bodyGenCB));
   } else {
     builder.restoreIP(ompBuilder->createTargetData(
-        ompLoc, builder.saveIP(), mapTypeFlags, mapNames, mapperAllocas,
-        isa<omp::EnterDataOp>(op), deviceID, ifCond, processMapOpCB));
+        ompLoc, allocaIP, builder.saveIP(), builder.getInt64(deviceID), ifCond,
+        info, genMapInfoCB, &RTLFn));
   }
 
-  if (failed(processMapOpStatus))
-    return processMapOpStatus;
   return bodyGenStatus;
 }
 
 /// Lowers the FlagsAttr which is applied to the module on the device
 /// pass when offloading, this attribute contains OpenMP RTL globals that can
 /// be passed as flags to the frontend, otherwise they are set to default
 LogicalResult convertFlagsAttr(Operation *op, mlir::omp::FlagsAttr attribute,
                                LLVM::ModuleTranslation &moduleTranslation) {
   if (!cast<mlir::ModuleOp>(op))
     return failure();
-  
   llvm::OpenMPIRBuilder *ompBuilder = moduleTranslation.getOpenMPBuilder();
 
   ompBuilder->createGlobalFlag(
       attribute.getDebugKind() /*LangOpts().OpenMPTargetDebug*/,
       "__omp_rtl_debug_kind");
   ompBuilder->createGlobalFlag(
       attribute
           .getAssumeTeamsOversubscription() /*LangOpts().OpenMPTeamSubscription*/

mlir/test/Target/LLVMIR/omptarget-llvm.mlir

 // RUN: mlir-translate -mlir-to-llvmir -split-input-file %s | FileCheck %s
 
 llvm.func @_QPopenmp_target_data() {
   %0 = llvm.mlir.constant(1 : i64) : i64
   %1 = llvm.alloca %0 x i32 {bindc_name = "i", in_type = i32, operand_segment_sizes = array<i32: 0, 0>, uniq_name = "_QFopenmp_target_dataEi"} : (i64) -> !llvm.ptr<i32>
   omp.target_data   map((tofrom -> %1 : !llvm.ptr<i32>)) {
     %2 = llvm.mlir.constant(99 : i32) : i32
     llvm.store %2, %1 : !llvm.ptr<i32>
     omp.terminator
   }
   llvm.return
 }
 
+// CHECK:         @.offload_sizes = private unnamed_addr constant [1 x i64] [i64 4]
 // CHECK:         @.offload_maptypes = private unnamed_addr constant [1 x i64] [i64 3]
 // CHECK-LABEL: define void @_QPopenmp_target_data() {
 // CHECK:         %[[VAL_0:.*]] = alloca [1 x ptr], align 8
 // CHECK:         %[[VAL_1:.*]] = alloca [1 x ptr], align 8
-// CHECK:         %[[VAL_2:.*]] = alloca [1 x i64], align 8
+// CHECK:         %[[VAL_2:.*]] = alloca [1 x ptr], align 8
 // CHECK:         %[[VAL_3:.*]] = alloca i32, i64 1, align 4
 // CHECK:         br label %[[VAL_4:.*]]
-// CHECK:       [[VAL_4]]:
-// CHECK:         %[[VAL_5:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
-// CHECK:         store ptr %[[VAL_3]], ptr %[[VAL_5]], align 8
-// CHECK:         %[[VAL_6:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
+// CHECK:       entry:                                            ; preds = %[[VAL_5:.*]]
+// CHECK:         %[[VAL_6:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
 // CHECK:         store ptr %[[VAL_3]], ptr %[[VAL_6]], align 8
-// CHECK:         %[[VAL_7:.*]] = getelementptr inbounds [1 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         store i64 ptrtoint (ptr getelementptr (ptr, ptr null, i32 1) to i64), ptr %[[VAL_7]], align 4
-// CHECK:         %[[VAL_8:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
-// CHECK:         %[[VAL_9:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
-// CHECK:         %[[VAL_10:.*]] = getelementptr inbounds [1 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         call void @__tgt_target_data_begin_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_8]], ptr %[[VAL_9]], ptr %[[VAL_10]], ptr @.offload_maptypes, ptr @.offload_mapnames, ptr null)
+// CHECK:         %[[VAL_7:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
+// CHECK:         store ptr %[[VAL_3]], ptr %[[VAL_7]], align 8
+// CHECK:         %[[VAL_8:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_2]], i64 0, i64 0
+// CHECK:         store ptr null, ptr %[[VAL_8]], align 8
+// CHECK:         %[[VAL_9:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
+// CHECK:         %[[VAL_10:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
+// CHECK:         call void @__tgt_target_data_begin_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_9]], ptr %[[VAL_10]], ptr @.offload_sizes, ptr @.offload_maptypes, ptr null, ptr null)
 // CHECK:         store i32 99, ptr %[[VAL_3]], align 4
 // CHECK:         %[[VAL_11:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
 // CHECK:         %[[VAL_12:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
-// CHECK:         %[[VAL_13:.*]] = getelementptr inbounds [1 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         call void @__tgt_target_data_end_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_11]], ptr %[[VAL_12]], ptr %[[VAL_13]], ptr @.offload_maptypes, ptr @.offload_mapnames, ptr null)
+// CHECK:         call void @__tgt_target_data_end_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_11]], ptr %[[VAL_12]], ptr @.offload_sizes, ptr @.offload_maptypes, ptr null, ptr null)
 // CHECK:         ret void
 
 // -----
 
 llvm.func @_QPopenmp_target_data_region(%1 : !llvm.ptr<array<1024 x i32>>) {
   omp.target_data   map((from -> %1 : !llvm.ptr<array<1024 x i32>>)) {
     %2 = llvm.mlir.constant(99 : i32) : i32
     %3 = llvm.mlir.constant(1 : i64) : i64
     %4 = llvm.mlir.constant(1 : i64) : i64
     %5 = llvm.mlir.constant(0 : i64) : i64
     %6 = llvm.getelementptr %1[0, %5] : (!llvm.ptr<array<1024 x i32>>, i64) -> !llvm.ptr<i32>
     llvm.store %2, %6 : !llvm.ptr<i32>
     omp.terminator
   }
   llvm.return
 }
 
+// CHECK:         @.offload_sizes = private unnamed_addr constant [1 x i64] [i64 4096]
 // CHECK:         @.offload_maptypes = private unnamed_addr constant [1 x i64] [i64 2]
 // CHECK-LABEL: define void @_QPopenmp_target_data_region
 // CHECK:         (ptr %[[ARG_0:.*]]) {
 // CHECK:         %[[VAL_0:.*]] = alloca [1 x ptr], align 8
 // CHECK:         %[[VAL_1:.*]] = alloca [1 x ptr], align 8
-// CHECK:         %[[VAL_2:.*]] = alloca [1 x i64], align 8
+// CHECK:         %[[VAL_2:.*]] = alloca [1 x ptr], align 8
 // CHECK:         br label %[[VAL_3:.*]]
-// CHECK:       [[VAL_3]]:
-// CHECK:         %[[VAL_4:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
-// CHECK:         store ptr %[[VAL_5:.*]], ptr %[[VAL_4]], align 8
-// CHECK:         %[[VAL_6:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
-// CHECK:         store ptr %[[VAL_5]], ptr %[[VAL_6]], align 8
-// CHECK:         %[[VAL_7:.*]] = getelementptr inbounds [1 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         store i64 ptrtoint (ptr getelementptr (ptr, ptr null, i32 1) to i64), ptr %[[VAL_7]], align 4
-// CHECK:         %[[VAL_8:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
-// CHECK:         %[[VAL_9:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
-// CHECK:         %[[VAL_10:.*]] = getelementptr inbounds [1 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         call void @__tgt_target_data_begin_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_8]], ptr %[[VAL_9]], ptr %[[VAL_10]], ptr @.offload_maptypes, ptr @.offload_mapnames, ptr null)
-// CHECK:         %[[VAL_11:.*]] = getelementptr [1024 x i32], ptr %[[VAL_5]], i32 0, i64 0
+// CHECK:       entry:                                            ; preds = %[[VAL_4:.*]]
+// CHECK:         %[[VAL_5:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
+// CHECK:         store ptr %[[VAL_6:.*]], ptr %[[VAL_5]], align 8
+// CHECK:         %[[VAL_7:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
+// CHECK:         store ptr %[[VAL_6]], ptr %[[VAL_7]], align 8
+// CHECK:         %[[VAL_8:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_2]], i64 0, i64 0
+// CHECK:         store ptr null, ptr %[[VAL_8]], align 8
+// CHECK:         %[[VAL_9:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
+// CHECK:         %[[VAL_10:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
+// CHECK:         call void @__tgt_target_data_begin_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_9]], ptr %[[VAL_10]], ptr @.offload_sizes, ptr @.offload_maptypes, ptr null, ptr null)
+// CHECK:         %[[VAL_11:.*]] = getelementptr [1024 x i32], ptr %[[VAL_6]], i32 0, i64 0
 // CHECK:         store i32 99, ptr %[[VAL_11]], align 4
 // CHECK:         %[[VAL_12:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_0]], i32 0, i32 0
 // CHECK:         %[[VAL_13:.*]] = getelementptr inbounds [1 x ptr], ptr %[[VAL_1]], i32 0, i32 0
-// CHECK:         %[[VAL_14:.*]] = getelementptr inbounds [1 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         call void @__tgt_target_data_end_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_12]], ptr %[[VAL_13]], ptr %[[VAL_14]], ptr @.offload_maptypes, ptr @.offload_mapnames, ptr null)
+// CHECK:         call void @__tgt_target_data_end_mapper(ptr @2, i64 -1, i32 1, ptr %[[VAL_12]], ptr %[[VAL_13]], ptr @.offload_sizes, ptr @.offload_maptypes, ptr null, ptr null)
 // CHECK:         ret void
 
 // -----
 
-llvm.func @_QPomp_target_enter_exit(%1 : !llvm.ptr<array<1024 x i32>>, %3 : !llvm.ptr<array<1024 x i32>>) {
+llvm.func @_QPomp_target_enter_exit(%1 : !llvm.ptr<array<1024 x i32>>, %3 : !llvm.ptr<array<512 x i32>>) {
   %4 = llvm.mlir.constant(1 : i64) : i64
   %5 = llvm.alloca %4 x i32 {bindc_name = "dvc", in_type = i32, operand_segment_sizes = array<i32: 0, 0>, uniq_name = "_QFomp_target_enter_exitEdvc"} : (i64) -> !llvm.ptr<i32>
   %6 = llvm.mlir.constant(1 : i64) : i64
   %7 = llvm.alloca %6 x i32 {bindc_name = "i", in_type = i32, operand_segment_sizes = array<i32: 0, 0>, uniq_name = "_QFomp_target_enter_exitEi"} : (i64) -> !llvm.ptr<i32>
   %8 = llvm.mlir.constant(5 : i32) : i32
   llvm.store %8, %7 : !llvm.ptr<i32>
   %9 = llvm.mlir.constant(2 : i32) : i32
   llvm.store %9, %5 : !llvm.ptr<i32>
   %10 = llvm.load %7 : !llvm.ptr<i32>
   %11 = llvm.mlir.constant(10 : i32) : i32
   %12 = llvm.icmp "slt" %10, %11 : i32
   %13 = llvm.load %5 : !llvm.ptr<i32>
-  omp.target_enter_data   if(%12 : i1) device(%13 : i32) map((to -> %1 : !llvm.ptr<array<1024 x i32>>), (alloc -> %3 : !llvm.ptr<array<1024 x i32>>))
+  omp.target_enter_data   if(%12 : i1) device(%13 : i32) map((to -> %1 : !llvm.ptr<array<1024 x i32>>), (alloc -> %3 : !llvm.ptr<array<512 x i32>>))
   %14 = llvm.load %7 : !llvm.ptr<i32>
   %15 = llvm.mlir.constant(10 : i32) : i32
   %16 = llvm.icmp "sgt" %14, %15 : i32
   %17 = llvm.load %5 : !llvm.ptr<i32>
-  omp.target_exit_data   if(%16 : i1) device(%17 : i32) map((from -> %1 : !llvm.ptr<array<1024 x i32>>), (release -> %3 : !llvm.ptr<array<1024 x i32>>))
+  omp.target_exit_data   if(%16 : i1) device(%17 : i32) map((from -> %1 : !llvm.ptr<array<1024 x i32>>), (release -> %3 : !llvm.ptr<array<512 x i32>>))
   llvm.return
 }
 
+// CHECK:         @.offload_sizes = private unnamed_addr constant [2 x i64] [i64 4096, i64 2048]
 // CHECK:         @.offload_maptypes = private unnamed_addr constant [2 x i64] [i64 1, i64 0]
-// CHECK:         @.offload_maptypes.1 = private unnamed_addr constant [2 x i64] [i64 2, i64 0]
+// CHECK:         @.offload_sizes.1 = private unnamed_addr constant [2 x i64] [i64 4096, i64 2048]
+// CHECK:         @.offload_maptypes.2 = private unnamed_addr constant [2 x i64] [i64 2, i64 0]
 // CHECK-LABEL: define void @_QPomp_target_enter_exit
 // CHECK:         (ptr %[[ARG_0:.*]], ptr %[[ARG_1:.*]]) {
 // CHECK:         %[[VAL_0:.*]] = alloca [2 x ptr], align 8
 // CHECK:         %[[VAL_1:.*]] = alloca [2 x ptr], align 8
-// CHECK:         %[[VAL_2:.*]] = alloca [2 x i64], align 8
+// CHECK:         %[[VAL_2:.*]] = alloca [2 x ptr], align 8
 // CHECK:         %[[VAL_3:.*]] = alloca [2 x ptr], align 8
 // CHECK:         %[[VAL_4:.*]] = alloca [2 x ptr], align 8
-// CHECK:         %[[VAL_5:.*]] = alloca [2 x i64], align 8
+// CHECK:         %[[VAL_5:.*]] = alloca [2 x ptr], align 8
 // CHECK:         %[[VAL_6:.*]] = alloca i32, i64 1, align 4
 // CHECK:         %[[VAL_7:.*]] = alloca i32, i64 1, align 4
 // CHECK:         store i32 5, ptr %[[VAL_7]], align 4
 // CHECK:         store i32 2, ptr %[[VAL_6]], align 4
 // CHECK:         %[[VAL_8:.*]] = load i32, ptr %[[VAL_7]], align 4
 // CHECK:         %[[VAL_9:.*]] = icmp slt i32 %[[VAL_8]], 10
 // CHECK:         %[[VAL_10:.*]] = load i32, ptr %[[VAL_6]], align 4
 // CHECK:         br label %[[VAL_11:.*]]
 // CHECK:       entry:                                            ; preds = %[[VAL_12:.*]]
 // CHECK:         br i1 %[[VAL_9]], label %[[VAL_13:.*]], label %[[VAL_14:.*]]
 // CHECK:       omp_if.then:                                      ; preds = %[[VAL_11]]
 // CHECK:         %[[VAL_15:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_3]], i32 0, i32 0
 // CHECK:         store ptr %[[VAL_16:.*]], ptr %[[VAL_15]], align 8
 // CHECK:         %[[VAL_17:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_4]], i32 0, i32 0
 // CHECK:         store ptr %[[VAL_16]], ptr %[[VAL_17]], align 8
-// CHECK:         %[[VAL_18:.*]] = getelementptr inbounds [2 x i64], ptr %[[VAL_5]], i32 0, i32 0
-// CHECK:         store i64 ptrtoint (ptr getelementptr (ptr, ptr null, i32 1) to i64), ptr %[[VAL_18]], align 4
+// CHECK:         %[[VAL_18:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_5]], i64 0, i64 0
+// CHECK:         store ptr null, ptr %[[VAL_18]], align 8
 // CHECK:         %[[VAL_19:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_3]], i32 0, i32 1
 // CHECK:         store ptr %[[VAL_20:.*]], ptr %[[VAL_19]], align 8
 // CHECK:         %[[VAL_21:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_4]], i32 0, i32 1
 // CHECK:         store ptr %[[VAL_20]], ptr %[[VAL_21]], align 8
-// CHECK:         %[[VAL_22:.*]] = getelementptr inbounds [2 x i64], ptr %[[VAL_5]], i32 0, i32 1
-// CHECK:         store i64 ptrtoint (ptr getelementptr (ptr, ptr null, i32 1) to i64), ptr %[[VAL_22]], align 4
+// CHECK:         %[[VAL_22:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_5]], i64 0, i64 1
+// CHECK:         store ptr null, ptr %[[VAL_22]], align 8
 // CHECK:         %[[VAL_23:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_3]], i32 0, i32 0
 // CHECK:         %[[VAL_24:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_4]], i32 0, i32 0
-// CHECK:         %[[VAL_25:.*]] = getelementptr inbounds [2 x i64], ptr %[[VAL_5]], i32 0, i32 0
-// CHECK:         call void @__tgt_target_data_begin_mapper(ptr @3, i64 -1, i32 2, ptr %[[VAL_23]], ptr %[[VAL_24]], ptr %[[VAL_25]], ptr @.offload_maptypes, ptr @.offload_mapnames, ptr null)
-// CHECK:         br label %[[VAL_14]]
-// CHECK:       omp_if.end:                                       ; preds = %[[VAL_11]], %[[VAL_13]]
+// CHECK:         call void @__tgt_target_data_begin_mapper(ptr @3, i64 -1, i32 2, ptr %[[VAL_23]], ptr %[[VAL_24]], ptr @.offload_sizes, ptr @.offload_maptypes, ptr null, ptr null)
+// CHECK:         br label %[[VAL_25:.*]]
+// CHECK:       omp_if.else:                                      ; preds = %[[VAL_11]]
+// CHECK:         br label %[[VAL_25]]
+// CHECK:       omp_if.end:                                       ; preds = %[[VAL_14]], %[[VAL_13]]
 // CHECK:         %[[VAL_26:.*]] = load i32, ptr %[[VAL_7]], align 4
 // CHECK:         %[[VAL_27:.*]] = icmp sgt i32 %[[VAL_26]], 10
 // CHECK:         %[[VAL_28:.*]] = load i32, ptr %[[VAL_6]], align 4
 // CHECK:         br i1 %[[VAL_27]], label %[[VAL_29:.*]], label %[[VAL_30:.*]]
-// CHECK:       omp_if.then4:                                     ; preds = %[[VAL_14]]
+// CHECK:       omp_if.then1:                                     ; preds = %[[VAL_25]]
 // CHECK:         %[[VAL_31:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_0]], i32 0, i32 0
 // CHECK:         store ptr %[[VAL_16]], ptr %[[VAL_31]], align 8
 // CHECK:         %[[VAL_32:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_1]], i32 0, i32 0
 // CHECK:         store ptr %[[VAL_16]], ptr %[[VAL_32]], align 8
-// CHECK:         %[[VAL_33:.*]] = getelementptr inbounds [2 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         store i64 ptrtoint (ptr getelementptr (ptr, ptr null, i32 1) to i64), ptr %[[VAL_33]], align 4
+// CHECK:         %[[VAL_33:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_2]], i64 0, i64 0
+// CHECK:         store ptr null, ptr %[[VAL_33]], align 8
 // CHECK:         %[[VAL_34:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_0]], i32 0, i32 1
 // CHECK:         store ptr %[[VAL_20]], ptr %[[VAL_34]], align 8
 // CHECK:         %[[VAL_35:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_1]], i32 0, i32 1
 // CHECK:         store ptr %[[VAL_20]], ptr %[[VAL_35]], align 8
-// CHECK:         %[[VAL_36:.*]] = getelementptr inbounds [2 x i64], ptr %[[VAL_2]], i32 0, i32 1
-// CHECK:         store i64 ptrtoint (ptr getelementptr (ptr, ptr null, i32 1) to i64), ptr %[[VAL_36]], align 4
+// CHECK:         %[[VAL_36:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_2]], i64 0, i64 1
+// CHECK:         store ptr null, ptr %[[VAL_36]], align 8
 // CHECK:         %[[VAL_37:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_0]], i32 0, i32 0
 // CHECK:         %[[VAL_38:.*]] = getelementptr inbounds [2 x ptr], ptr %[[VAL_1]], i32 0, i32 0
-// CHECK:         %[[VAL_39:.*]] = getelementptr inbounds [2 x i64], ptr %[[VAL_2]], i32 0, i32 0
-// CHECK:         call void @__tgt_target_data_end_mapper(ptr @3, i64 -1, i32 2, ptr %[[VAL_37]], ptr %[[VAL_38]], ptr %[[VAL_39]], ptr @.offload_maptypes.1, ptr @.offload_mapnames.2, ptr null)
-// CHECK:         br label %[[VAL_30]]
-// CHECK:       omp_if.end5:                                      ; preds = %[[VAL_14]], %[[VAL_29]]
+// CHECK:         call void @__tgt_target_data_end_mapper(ptr @3, i64 -1, i32 2, ptr %[[VAL_37]], ptr %[[VAL_38]], ptr @.offload_sizes.1, ptr @.offload_maptypes.2, ptr null, ptr null)
+// CHECK:         br label %[[VAL_39:.*]]
+// CHECK:       omp_if.else5:                                     ; preds = %[[VAL_25]]
+// CHECK:         br label %[[VAL_39]]
+// CHECK:       omp_if.end6:                                      ; preds = %[[VAL_30]], %[[VAL_29]]
 // CHECK:         ret void
 
 // -----
-