[OPENMP] Additional attributes for the pointer parameters.

Added attributes for better optimization of the OpenMP code.

llvm-svn: 329751

8 files changed, 133 insertions, 95 deletions

diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index d23989140ba..547876bf018 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -2982,6 +2982,7 @@ static llvm::Value *emitCopyprivateCopyFunction(
       CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
       ".omp.copyprivate.copy_func", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
   // Dest = (void*[n])(LHSArg);
@@ -4211,6 +4212,7 @@ emitProxyTaskFunction(CodeGenModule &CGM, SourceLocation Loc,
       llvm::Function::Create(TaskEntryTy, llvm::GlobalValue::InternalLinkage,
                              ".omp_task_entry.", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), TaskEntry, TaskEntryFnInfo);
+  TaskEntry->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), KmpInt32Ty, TaskEntry, TaskEntryFnInfo, Args,
                     Loc, Loc);
@@ -4312,6 +4314,7 @@ static llvm::Value *emitDestructorsFunction(CodeGenModule &CGM,
                              ".omp_task_destructor.", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), DestructorFn,
                                     DestructorFnInfo);
+  DestructorFn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), KmpInt32Ty, DestructorFn, DestructorFnInfo,
                     Args, Loc, Loc);
@@ -4589,6 +4592,7 @@ emitTaskDupFunction(CodeGenModule &CGM, SourceLocation Loc,
       llvm::Function::Create(TaskDupTy, llvm::GlobalValue::InternalLinkage,
                              ".omp_task_dup.", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), TaskDup, TaskDupFnInfo);
+  TaskDup->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, TaskDup, TaskDupFnInfo, Args, Loc,
                     Loc);
@@ -5231,6 +5235,7 @@ llvm::Value *CGOpenMPRuntime::emitReductionFunction(
       CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
       ".omp.reduction.reduction_func", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
 
@@ -5653,6 +5658,7 @@ static llvm::Value *emitReduceInitFunction(CodeGenModule &CGM,
   auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
                                     ".red_init.", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
   Address PrivateAddr = CGF.EmitLoadOfPointer(
@@ -5728,6 +5734,7 @@ static llvm::Value *emitReduceCombFunction(CodeGenModule &CGM,
   auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
                                     ".red_comb.", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
   llvm::Value *Size = nullptr;
@@ -5796,6 +5803,7 @@ static llvm::Value *emitReduceFiniFunction(CodeGenModule &CGM,
   auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
                                     ".red_fini.", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
   Address PrivateAddr = CGF.EmitLoadOfPointer(
diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
index 0d7386490df..1127562b51e 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeNVPTX.cpp
@@ -566,6 +566,7 @@ void CGOpenMPRuntimeNVPTX::WorkerFunctionState::createWorkerFunction(
       CGM.getTypes().GetFunctionType(*CGFI), llvm::GlobalValue::InternalLinkage,
       /*placeholder=*/"_worker", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), WorkerFn, *CGFI);
+  WorkerFn->setDoesNotRecurse();
 }
 
 bool CGOpenMPRuntimeNVPTX::isInSpmdExecutionMode() const {
@@ -1868,6 +1869,7 @@ static llvm::Value *emitReduceScratchpadFunction(
       CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
       "_omp_reduction_load_and_reduce", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
 
@@ -1989,6 +1991,7 @@ static llvm::Value *emitCopyToScratchpad(CodeGenModule &CGM,
       CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
       "_omp_reduction_copy_to_scratchpad", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
 
@@ -2070,6 +2073,7 @@ static llvm::Value *emitInterWarpCopyFunction(CodeGenModule &CGM,
       CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
       "_omp_reduction_inter_warp_copy_func", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
 
@@ -2320,6 +2324,7 @@ static llvm::Value *emitShuffleAndReduceFunction(
       CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
       "_omp_reduction_shuffle_and_reduce_func", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+  Fn->setDoesNotRecurse();
   CodeGenFunction CGF(CGM);
   CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
 
@@ -2948,6 +2953,7 @@ llvm::Function *CGOpenMPRuntimeNVPTX::createParallelDataSharingWrapper(
       OutlinedParallelFn->getName() + "_wrapper", &CGM.getModule());
   CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
   Fn->setLinkage(llvm::GlobalValue::InternalLinkage);
+  Fn->setDoesNotRecurse();
 
   CodeGenFunction CGF(CGM, /*suppressNewContext=*/true);
   CGF.StartFunction(GlobalDecl(), Ctx.VoidTy, Fn, CGFI, WrapperArgs,
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 455a4fff1ae..5cfece71254 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -456,6 +456,7 @@ static llvm::Function *emitOutlinedFunctionPrologue(
   CGM.SetInternalFunctionAttributes(CD, F, FuncInfo);
   if (CD->isNothrow())
     F->setDoesNotThrow();
+  F->setDoesNotRecurse();
 
   // Generate the function.
   CGF.StartFunction(CD, Ctx.VoidTy, F, FuncInfo, TargetArgs,
@@ -3418,7 +3419,7 @@ static llvm::Function *emitOutlinedOrderedFunction(CodeGenModule &CGM,
   CodeGenFunction::CGCapturedStmtInfo CapStmtInfo;
   CGF.CapturedStmtInfo = &CapStmtInfo;
   auto *Fn = CGF.GenerateOpenMPCapturedStmtFunction(*S);
-  Fn->addFnAttr(llvm::Attribute::NoInline);
+  Fn->setDoesNotRecurse();
   return Fn;
 }
 
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index ca9dc32ea40..35f30b3727d 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -2172,7 +2172,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
   case OMPD_teams:
   case OMPD_teams_distribute:
   case OMPD_teams_distribute_simd: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
     Sema::CapturedParamNameType Params[] = {
@@ -2190,17 +2190,21 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
   case OMPD_target_parallel_for_simd:
   case OMPD_target_teams_distribute:
   case OMPD_target_teams_distribute_simd: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
-    QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
+    QualType VoidPtrTy = Context.VoidPtrTy.withConst().withRestrict();
+    QualType KmpInt32PtrTy =
+        Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
+    QualType Args[] = {VoidPtrTy};
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
     Sema::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
-        std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
-        std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
-        std::make_pair(".copy_fn.",
-                       Context.getPointerType(CopyFnType).withConst()),
+        std::make_pair(".part_id.", KmpInt32PtrTy),
+        std::make_pair(".privates.", VoidPtrTy),
+        std::make_pair(
+            ".copy_fn.",
+            Context.getPointerType(CopyFnType).withConst().withRestrict()),
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
@@ -2210,15 +2214,13 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     // function directly.
     getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
-            Context, AlwaysInlineAttr::Keyword_forceinline, SourceRange()));
+            Context, AlwaysInlineAttr::Keyword_forceinline));
     Sema::CapturedParamNameType ParamsTarget[] = {
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     // Start a captured region for 'target' with no implicit parameters.
     ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
                              ParamsTarget);
-    QualType KmpInt32PtrTy =
-        Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
     Sema::CapturedParamNameType ParamsTeamsOrParallel[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
@@ -2232,17 +2234,21 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
   }
   case OMPD_target:
   case OMPD_target_simd: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
-    QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
+    QualType VoidPtrTy = Context.VoidPtrTy.withConst().withRestrict();
+    QualType KmpInt32PtrTy =
+        Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
+    QualType Args[] = {VoidPtrTy};
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
     Sema::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
-        std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
-        std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
-        std::make_pair(".copy_fn.",
-                       Context.getPointerType(CopyFnType).withConst()),
+        std::make_pair(".part_id.", KmpInt32PtrTy),
+        std::make_pair(".privates.", VoidPtrTy),
+        std::make_pair(
+            ".copy_fn.",
+            Context.getPointerType(CopyFnType).withConst().withRestrict()),
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
@@ -2252,7 +2258,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     // function directly.
     getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
-            Context, AlwaysInlineAttr::Keyword_forceinline, SourceRange()));
+            Context, AlwaysInlineAttr::Keyword_forceinline));
     ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
                              std::make_pair(StringRef(), QualType()));
     break;
@@ -2279,17 +2285,21 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     break;
   }
   case OMPD_task: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
-    QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
+    QualType VoidPtrTy = Context.VoidPtrTy.withConst().withRestrict();
+    QualType KmpInt32PtrTy =
+        Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
+    QualType Args[] = {VoidPtrTy};
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
     Sema::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
-        std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
-        std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
-        std::make_pair(".copy_fn.",
-                       Context.getPointerType(CopyFnType).withConst()),
+        std::make_pair(".part_id.", KmpInt32PtrTy),
+        std::make_pair(".privates.", VoidPtrTy),
+        std::make_pair(
+            ".copy_fn.",
+            Context.getPointerType(CopyFnType).withConst().withRestrict()),
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
@@ -2299,35 +2309,40 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     // function directly.
     getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
-            Context, AlwaysInlineAttr::Keyword_forceinline, SourceRange()));
+            Context, AlwaysInlineAttr::Keyword_forceinline));
     break;
   }
   case OMPD_taskloop:
   case OMPD_taskloop_simd: {
     QualType KmpInt32Ty =
-        Context.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
+        Context.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1)
+            .withConst();
     QualType KmpUInt64Ty =
-        Context.getIntTypeForBitwidth(/*DestWidth=*/64, /*Signed=*/0);
+        Context.getIntTypeForBitwidth(/*DestWidth=*/64, /*Signed=*/0)
+            .withConst();
     QualType KmpInt64Ty =
-        Context.getIntTypeForBitwidth(/*DestWidth=*/64, /*Signed=*/1);
-    QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+        Context.getIntTypeForBitwidth(/*DestWidth=*/64, /*Signed=*/1)
+            .withConst();
+    QualType VoidPtrTy = Context.VoidPtrTy.withConst().withRestrict();
+    QualType KmpInt32PtrTy =
+        Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
+    QualType Args[] = {VoidPtrTy};
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
     Sema::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
-        std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
-        std::make_pair(".privates.",
-                       Context.VoidPtrTy.withConst().withRestrict()),
+        std::make_pair(".part_id.", KmpInt32PtrTy),
+        std::make_pair(".privates.", VoidPtrTy),
         std::make_pair(
             ".copy_fn.",
             Context.getPointerType(CopyFnType).withConst().withRestrict()),
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(".lb.", KmpUInt64Ty),
-        std::make_pair(".ub.", KmpUInt64Ty), std::make_pair(".st.", KmpInt64Ty),
+        std::make_pair(".ub.", KmpUInt64Ty),
+        std::make_pair(".st.", KmpInt64Ty),
         std::make_pair(".liter.", KmpInt32Ty),
-        std::make_pair(".reductions.",
-                       Context.VoidPtrTy.withConst().withRestrict()),
+        std::make_pair(".reductions.", VoidPtrTy),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
@@ -2336,19 +2351,19 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     // function directly.
     getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
-            Context, AlwaysInlineAttr::Keyword_forceinline, SourceRange()));
+            Context, AlwaysInlineAttr::Keyword_forceinline));
     break;
   }
   case OMPD_distribute_parallel_for_simd:
   case OMPD_distribute_parallel_for: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
     Sema::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
-        std::make_pair(".previous.lb.", Context.getSizeType()),
-        std::make_pair(".previous.ub.", Context.getSizeType()),
+        std::make_pair(".previous.lb.", Context.getSizeType().withConst()),
+        std::make_pair(".previous.ub.", Context.getSizeType().withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
@@ -2357,20 +2372,22 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
   }
   case OMPD_target_teams_distribute_parallel_for:
   case OMPD_target_teams_distribute_parallel_for_simd: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
+    QualType VoidPtrTy = Context.VoidPtrTy.withConst().withRestrict();
 
-    QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+    QualType Args[] = {VoidPtrTy};
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
     Sema::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
-        std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
-        std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
-        std::make_pair(".copy_fn.",
-                       Context.getPointerType(CopyFnType).withConst()),
+        std::make_pair(".part_id.", KmpInt32PtrTy),
+        std::make_pair(".privates.", VoidPtrTy),
+        std::make_pair(
+            ".copy_fn.",
+            Context.getPointerType(CopyFnType).withConst().withRestrict()),
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
@@ -2380,7 +2397,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     // function directly.
     getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
-            Context, AlwaysInlineAttr::Keyword_forceinline, SourceRange()));
+            Context, AlwaysInlineAttr::Keyword_forceinline));
     Sema::CapturedParamNameType ParamsTarget[] = {
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
@@ -2400,8 +2417,8 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     Sema::CapturedParamNameType ParamsParallel[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
-        std::make_pair(".previous.lb.", Context.getSizeType()),
-        std::make_pair(".previous.ub.", Context.getSizeType()),
+        std::make_pair(".previous.lb.", Context.getSizeType().withConst()),
+        std::make_pair(".previous.ub.", Context.getSizeType().withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     // Start a captured region for 'teams' or 'parallel'.  Both regions have
@@ -2413,7 +2430,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
 
   case OMPD_teams_distribute_parallel_for:
   case OMPD_teams_distribute_parallel_for_simd: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
     QualType KmpInt32PtrTy =
         Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
 
@@ -2429,8 +2446,8 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     Sema::CapturedParamNameType ParamsParallel[] = {
         std::make_pair(".global_tid.", KmpInt32PtrTy),
         std::make_pair(".bound_tid.", KmpInt32PtrTy),
-        std::make_pair(".previous.lb.", Context.getSizeType()),
-        std::make_pair(".previous.ub.", Context.getSizeType()),
+        std::make_pair(".previous.lb.", Context.getSizeType().withConst()),
+        std::make_pair(".previous.ub.", Context.getSizeType().withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
     // Start a captured region for 'teams' or 'parallel'.  Both regions have
@@ -2442,17 +2459,21 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
   case OMPD_target_update:
   case OMPD_target_enter_data:
   case OMPD_target_exit_data: {
-    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
-    QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+    QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1).withConst();
+    QualType VoidPtrTy = Context.VoidPtrTy.withConst().withRestrict();
+    QualType KmpInt32PtrTy =
+        Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
+    QualType Args[] = {VoidPtrTy};
     FunctionProtoType::ExtProtoInfo EPI;
     EPI.Variadic = true;
     QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
     Sema::CapturedParamNameType Params[] = {
         std::make_pair(".global_tid.", KmpInt32Ty),
-        std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
-        std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
-        std::make_pair(".copy_fn.",
-                       Context.getPointerType(CopyFnType).withConst()),
+        std::make_pair(".part_id.", KmpInt32PtrTy),
+        std::make_pair(".privates.", VoidPtrTy),
+        std::make_pair(
+            ".copy_fn.",
+            Context.getPointerType(CopyFnType).withConst().withRestrict()),
         std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
         std::make_pair(StringRef(), QualType()) // __context with shared vars
     };
@@ -2462,7 +2483,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
     // function directly.
     getCurCapturedRegion()->TheCapturedDecl->addAttr(
         AlwaysInlineAttr::CreateImplicit(
-            Context, AlwaysInlineAttr::Keyword_forceinline, SourceRange()));
+            Context, AlwaysInlineAttr::Keyword_forceinline));
     break;
   }
   case OMPD_threadprivate:
diff --git a/clang/lib/Sema/SemaStmt.cpp b/clang/lib/Sema/SemaStmt.cpp
index 7f1e4d527d2..8963f4c564e 100644
--- a/clang/lib/Sema/SemaStmt.cpp
+++ b/clang/lib/Sema/SemaStmt.cpp
@@ -4135,7 +4135,9 @@ void Sema::ActOnCapturedRegionStart(SourceLocation Loc, Scope *CurScope,
       assert(!ContextIsFound &&
              "null type has been found already for '__context' parameter");
       IdentifierInfo *ParamName = &Context.Idents.get("__context");
-      QualType ParamType = Context.getPointerType(Context.getTagDeclType(RD));
+      QualType ParamType = Context.getPointerType(Context.getTagDeclType(RD))
+                               .withConst()
+                               .withRestrict();
       auto *Param =
           ImplicitParamDecl::Create(Context, DC, Loc, ParamName, ParamType,
                                     ImplicitParamDecl::CapturedContext);
diff --git a/clang/test/OpenMP/target_enter_data_depend_codegen.cpp b/clang/test/OpenMP/target_enter_data_depend_codegen.cpp
index 2e4440cd405..8f4c50fc21a 100644
--- a/clang/test/OpenMP/target_enter_data_depend_codegen.cpp
+++ b/clang/test/OpenMP/target_enter_data_depend_codegen.cpp
@@ -336,9 +336,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPBP]] = getelementptr inbounds {{.+}}[[BP:%[^,]+]]
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1: ret i32 0
 // CK1: }
@@ -348,9 +348,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPBP]] = getelementptr inbounds {{.+}}[[BP:%[^,]+]]
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1: ret i32 0
 // CK1: }
@@ -361,9 +361,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
 
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1-NOT: __tgt_target_data_end
 // CK1: ret i32 0
@@ -375,9 +375,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
 
-// CK1-DAG: [[BP]] = load [2 x i8*]*, [2 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [2 x i8*]*, [2 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [2 x i[[sz]]]*, [2 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [2 x i8*]*, [2 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [2 x i8*]*, [2 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [2 x i[[sz]]]*, [2 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [2 x i8*]** [[BP_PRIV]], [2 x i8*]** [[P_PRIV]], [2 x i[[sz]]]** [[S_PRIV]])
 // CK1-NOT: __tgt_target_data_end
 // CK1: ret i32 0
diff --git a/clang/test/OpenMP/target_exit_data_depend_codegen.cpp b/clang/test/OpenMP/target_exit_data_depend_codegen.cpp
index ad8c47522d4..1d005d55cfc 100644
--- a/clang/test/OpenMP/target_exit_data_depend_codegen.cpp
+++ b/clang/test/OpenMP/target_exit_data_depend_codegen.cpp
@@ -336,9 +336,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPBP]] = getelementptr inbounds {{.+}}[[BP:%[^,]+]]
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1: ret i32 0
 // CK1: }
@@ -348,9 +348,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPBP]] = getelementptr inbounds {{.+}}[[BP:%[^,]+]]
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1: ret i32 0
 // CK1: }
@@ -361,9 +361,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
 
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1-NOT: __tgt_target_data_end
 // CK1: ret i32 0
@@ -375,9 +375,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
 
-// CK1-DAG: [[BP]] = load [2 x i8*]*, [2 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [2 x i8*]*, [2 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [2 x i[[sz]]]*, [2 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [2 x i8*]*, [2 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [2 x i8*]*, [2 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [2 x i[[sz]]]*, [2 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [2 x i8*]** [[BP_PRIV]], [2 x i8*]** [[P_PRIV]], [2 x i[[sz]]]** [[S_PRIV]])
 // CK1-NOT: __tgt_target_data_end
 // CK1: ret i32 0
diff --git a/clang/test/OpenMP/target_update_depend_codegen.cpp b/clang/test/OpenMP/target_update_depend_codegen.cpp
index 550d31f9de6..b2152c082f2 100644
--- a/clang/test/OpenMP/target_update_depend_codegen.cpp
+++ b/clang/test/OpenMP/target_update_depend_codegen.cpp
@@ -336,9 +336,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPBP]] = getelementptr inbounds {{.+}}[[BP:%[^,]+]]
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1: ret i32 0
 // CK1: }
@@ -348,9 +348,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPBP]] = getelementptr inbounds {{.+}}[[BP:%[^,]+]]
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1: ret i32 0
 // CK1: }
@@ -361,9 +361,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
 
-// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [1 x i8*]*, [1 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [1 x i8*]*, [1 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [1 x i[[sz]]]*, [1 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [1 x i8*]** [[BP_PRIV]], [1 x i8*]** [[P_PRIV]], [1 x i[[sz]]]** [[S_PRIV]])
 // CK1-NOT: __tgt_target_data_end
 // CK1: ret i32 0
@@ -375,9 +375,9 @@ void foo(int arg) {
 // CK1-DAG: [[GEPP]] = getelementptr inbounds {{.+}}[[P:%[^,]+]]
 // CK1-DAG: [[GEPS]] = getelementptr inbounds {{.+}}[[S:%[^,]+]]
 
-// CK1-DAG: [[BP]] = load [2 x i8*]*, [2 x i8*]** [[BP_PRIV:%.+]],
-// CK1-DAG: [[P]] = load [2 x i8*]*, [2 x i8*]** [[P_PRIV:%.+]],
-// CK1-DAG: [[S]] = load [2 x i[[sz]]]*, [2 x i[[sz]]]** [[S_PRIV:%.+]],
+// CK1-DAG: [[BP]] = load [2 x i8*]*, [2 x i8*]** [[BP_PRIV:%[^,]+]],
+// CK1-DAG: [[P]] = load [2 x i8*]*, [2 x i8*]** [[P_PRIV:%[^,]+]],
+// CK1-DAG: [[S]] = load [2 x i[[sz]]]*, [2 x i[[sz]]]** [[S_PRIV:%[^,]+]],
 // CK1-DAG: call void (i8*, ...) %{{.+}}(i8* %{{[^,]+}}, [2 x i8*]** [[BP_PRIV]], [2 x i8*]** [[P_PRIV]], [2 x i[[sz]]]** [[S_PRIV]])
 // CK1-NOT: __tgt_target_data_end
 // CK1: ret i32 0