[OpenMP] Introduce the `ompx_dyn_cgroup_mem(<N>)` clause

Dynamic memory allows users to allocate fast shared memory when a kernel
is launched. We support a single size for all kernels via the
`LIBOMPTARGET_SHARED_MEMORY_SIZE` environment variable but now we can
control it per kernel invocation, hence allow computed values.

Note: Only the nextgen plugins will allocate memory based on the clause,
      the old plugins will silently miscompile.

Differential Revision: https://reviews.llvm.org/D141233

diff --git a/clang/include/clang/AST/OpenMPClause.h b/clang/include/clang/AST/OpenMPClause.h
index 352e60f..627e902 100644 (file)
--- a/clang/include/clang/AST/OpenMPClause.h
+++ b/clang/include/clang/AST/OpenMPClause.h
@@ -9004,6 +9004,48 @@ public:
   }
 };
 
+/// This represents 'ompx_dyn_cgroup_mem' clause in the '#pragma omp target ...'
+/// directive.
+///
+/// \code
+/// #pragma omp target [...] ompx_dyn_cgroup_mem(N)
+/// \endcode
+class OMPXDynCGroupMemClause
+    : public OMPOneStmtClause<llvm::omp::OMPC_ompx_dyn_cgroup_mem, OMPClause>,
+      public OMPClauseWithPreInit {
+  friend class OMPClauseReader;
+
+  /// Set size.
+  void setSize(Expr *E) { setStmt(E); }
+
+public:
+  /// Build 'ompx_dyn_cgroup_mem' clause.
+  ///
+  /// \param Size Size expression.
+  /// \param HelperSize Helper Size expression
+  /// \param CaptureRegion Innermost OpenMP region where expressions in this
+  /// \param StartLoc Starting location of the clause.
+  /// \param LParenLoc Location of '('.
+  /// \param EndLoc Ending location of the clause.
+  OMPXDynCGroupMemClause(Expr *Size, Stmt *HelperSize,
+                         OpenMPDirectiveKind CaptureRegion,
+                         SourceLocation StartLoc, SourceLocation LParenLoc,
+                         SourceLocation EndLoc)
+      : OMPOneStmtClause(Size, StartLoc, LParenLoc, EndLoc),
+        OMPClauseWithPreInit(this) {
+    setPreInitStmt(HelperSize, CaptureRegion);
+  }
+
+  /// Build an empty clause.
+  OMPXDynCGroupMemClause() : OMPOneStmtClause(), OMPClauseWithPreInit(this) {}
+
+  /// Return the size expression.
+  Expr *getSize() { return getStmtAs<Expr>(); }
+
+  /// Return the size expression.
+  Expr *getSize() const { return getStmtAs<Expr>(); }
+};
+
 } // namespace clang
 
 #endif // LLVM_CLANG_AST_OPENMPCLAUSE_H

diff --git a/clang/include/clang/AST/RecursiveASTVisitor.h b/clang/include/clang/AST/RecursiveASTVisitor.h
index 7036200..5802e90 100644 (file)
--- a/clang/include/clang/AST/RecursiveASTVisitor.h
+++ b/clang/include/clang/AST/RecursiveASTVisitor.h
@@ -3852,6 +3852,14 @@ bool RecursiveASTVisitor<Derived>::VisitOMPBindClause(OMPBindClause *C) {
   return true;
 }
 
+template <typename Derived>
+bool RecursiveASTVisitor<Derived>::VisitOMPXDynCGroupMemClause(
+    OMPXDynCGroupMemClause *C) {
+  TRY_TO(VisitOMPClauseWithPreInit(C));
+  TRY_TO(TraverseStmt(C->getSize()));
+  return true;
+}
+
 // FIXME: look at the following tricky-seeming exprs to see if we
 // need to recurse on anything.  These are ones that have methods
 // returning decls or qualtypes or nestednamespecifier -- though I'm

diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index deaf152..7fbd8ef 100644 (file)
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -12177,6 +12177,11 @@ public:
                                    SourceLocation LParenLoc,
                                    SourceLocation EndLoc);
 
+  /// Called on a well-formed 'ompx_dyn_cgroup_mem' clause.
+  OMPClause *ActOnOpenMPXDynCGroupMemClause(Expr *Size, SourceLocation StartLoc,
+                                            SourceLocation LParenLoc,
+                                            SourceLocation EndLoc);
+
   /// The kind of conversion being performed.
   enum CheckedConversionKind {
     /// An implicit conversion.

diff --git a/clang/lib/AST/OpenMPClause.cpp b/clang/lib/AST/OpenMPClause.cpp
index 72a5ea6..2e88c08 100644 (file)
--- a/clang/lib/AST/OpenMPClause.cpp
+++ b/clang/lib/AST/OpenMPClause.cpp
@@ -103,6 +103,8 @@ const OMPClauseWithPreInit *OMPClauseWithPreInit::get(const OMPClause *C) {
     return static_cast<const OMPNocontextClause *>(C);
   case OMPC_filter:
     return static_cast<const OMPFilterClause *>(C);
+  case OMPC_ompx_dyn_cgroup_mem:
+    return static_cast<const OMPXDynCGroupMemClause *>(C);
   case OMPC_default:
   case OMPC_proc_bind:
   case OMPC_safelen:
@@ -2455,6 +2457,13 @@ void OMPClausePrinter::VisitOMPBindClause(OMPBindClause *Node) {
      << ")";
 }
 
+void OMPClausePrinter::VisitOMPXDynCGroupMemClause(
+    OMPXDynCGroupMemClause *Node) {
+  OS << "ompx_dyn_cgroup_mem(";
+  Node->getSize()->printPretty(OS, nullptr, Policy, 0);
+  OS << ")";
+}
+
 void OMPTraitInfo::getAsVariantMatchInfo(ASTContext &ASTCtx,
                                          VariantMatchInfo &VMI) const {
   for (const OMPTraitSet &Set : Sets) {

diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index 3820a78..960cc4f 100644 (file)
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -903,6 +903,12 @@ void OMPClauseProfiler::VisitOMPAffinityClause(const OMPAffinityClause *C) {
 }
 void OMPClauseProfiler::VisitOMPOrderClause(const OMPOrderClause *C) {}
 void OMPClauseProfiler::VisitOMPBindClause(const OMPBindClause *C) {}
+void OMPClauseProfiler::VisitOMPXDynCGroupMemClause(
+    const OMPXDynCGroupMemClause *C) {
+  VistOMPClauseWithPreInit(C);
+  if (Expr *Size = C->getSize())
+    Profiler->VisitStmt(Size);
+}
 } // namespace
 
 void

diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 5335cc0..2284aa1 100644 (file)
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -9904,6 +9904,13 @@ void CGOpenMPRuntime::emitTargetCall(
         emitTargetNumIterationsCall(CGF, D, SizeEmitter);
 
     llvm::Value *DynCGroupMem = CGF.Builder.getInt32(0);
+    if (auto *DynMemClause = D.getSingleClause<OMPXDynCGroupMemClause>()) {
+      CodeGenFunction::RunCleanupsScope DynCGroupMemScope(CGF);
+      llvm::Value *DynCGroupMemVal = CGF.EmitScalarExpr(
+          DynMemClause->getSize(), /*IgnoreResultAssign=*/true);
+      DynCGroupMem = CGF.Builder.CreateIntCast(DynCGroupMemVal, CGF.Int32Ty,
+                                               /*isSigned=*/false);
+    }
 
     llvm::Value *ZeroArray =
         llvm::Constant::getNullValue(llvm::ArrayType::get(CGF.CGM.Int32Ty, 3));

diff --git a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
index 31f2d03..e8c5f04 100644 (file)
--- a/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntimeGPU.cpp
@@ -15,6 +15,7 @@
 #include "CodeGenFunction.h"
 #include "clang/AST/Attr.h"
 #include "clang/AST/DeclOpenMP.h"
+#include "clang/AST/OpenMPClause.h"
 #include "clang/AST/StmtOpenMP.h"
 #include "clang/AST/StmtVisitor.h"
 #include "clang/Basic/Cuda.h"

diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp
index fa6e636..a31ceae 100644 (file)
--- a/clang/lib/Parse/ParseOpenMP.cpp
+++ b/clang/lib/Parse/ParseOpenMP.cpp
@@ -3195,6 +3195,7 @@ OMPClause *Parser::ParseOpenMPClause(OpenMPDirectiveKind DKind,
   case OMPC_partial:
   case OMPC_align:
   case OMPC_message:
+  case OMPC_ompx_dyn_cgroup_mem:
     // OpenMP [2.5, Restrictions]
     //  At most one num_threads clause can appear on the directive.
     // OpenMP [2.8.1, simd construct, Restrictions]

diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index 1b9d06b..eb7bd26 100644 (file)
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -15217,6 +15217,9 @@ OMPClause *Sema::ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind, Expr *Expr,
   case OMPC_align:
     Res = ActOnOpenMPAlignClause(Expr, StartLoc, LParenLoc, EndLoc);
     break;
+  case OMPC_ompx_dyn_cgroup_mem:
+    Res = ActOnOpenMPXDynCGroupMemClause(Expr, StartLoc, LParenLoc, EndLoc);
+    break;
   case OMPC_grainsize:
   case OMPC_num_tasks:
   case OMPC_device:
@@ -15933,6 +15936,26 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause(
       llvm_unreachable("Unknown OpenMP directive");
     }
     break;
+  case OMPC_ompx_dyn_cgroup_mem:
+    switch (DKind) {
+    case OMPD_target:
+    case OMPD_target_simd:
+    case OMPD_target_teams:
+    case OMPD_target_parallel:
+    case OMPD_target_teams_distribute:
+    case OMPD_target_teams_distribute_simd:
+    case OMPD_target_parallel_for:
+    case OMPD_target_parallel_for_simd:
+    case OMPD_target_parallel_loop:
+    case OMPD_target_teams_distribute_parallel_for:
+    case OMPD_target_teams_distribute_parallel_for_simd:
+    case OMPD_target_teams_loop:
+      CaptureRegion = OMPD_target;
+      break;
+    default:
+      llvm_unreachable("Unknown OpenMP directive");
+    }
+    break;
   case OMPC_device:
     switch (DKind) {
     case OMPD_target_update:
@@ -17343,6 +17366,7 @@ OMPClause *Sema::ActOnOpenMPClause(OpenMPClauseKind Kind,
   case OMPC_uses_allocators:
   case OMPC_affinity:
   case OMPC_when:
+  case OMPC_ompx_dyn_cgroup_mem:
   default:
     llvm_unreachable("Clause is not allowed.");
   }
@@ -23776,3 +23800,31 @@ OMPClause *Sema::ActOnOpenMPBindClause(OpenMPBindClauseKind Kind,
   return OMPBindClause::Create(Context, Kind, KindLoc, StartLoc, LParenLoc,
                                EndLoc);
 }
+
+OMPClause *Sema::ActOnOpenMPXDynCGroupMemClause(Expr *Size,
+                                                SourceLocation StartLoc,
+                                                SourceLocation LParenLoc,
+                                                SourceLocation EndLoc) {
+  Expr *ValExpr = Size;
+  Stmt *HelperValStmt = nullptr;
+
+  // OpenMP [2.5, Restrictions]
+  //  The ompx_dyn_cgroup_mem expression must evaluate to a positive integer
+  //  value.
+  if (!isNonNegativeIntegerValue(ValExpr, *this, OMPC_ompx_dyn_cgroup_mem,
+                                 /*StrictlyPositive=*/false))
+    return nullptr;
+
+  OpenMPDirectiveKind DKind = DSAStack->getCurrentDirective();
+  OpenMPDirectiveKind CaptureRegion = getOpenMPCaptureRegionForClause(
+      DKind, OMPC_ompx_dyn_cgroup_mem, LangOpts.OpenMP);
+  if (CaptureRegion != OMPD_unknown && !CurContext->isDependentContext()) {
+    ValExpr = MakeFullExpr(ValExpr).get();
+    llvm::MapVector<const Expr *, DeclRefExpr *> Captures;
+    ValExpr = tryBuildCapture(*this, ValExpr, Captures).get();
+    HelperValStmt = buildPreInits(Context, Captures);
+  }
+
+  return new (Context) OMPXDynCGroupMemClause(
+      ValExpr, HelperValStmt, CaptureRegion, StartLoc, LParenLoc, EndLoc);
+}

diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index 7f76d0a..6a05ecc 100644 (file)
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -2344,6 +2344,17 @@ public:
                                            EndLoc);
   }
 
+  /// Build a new OpenMP 'ompx_dyn_cgroup_mem' clause.
+  ///
+  /// By default, performs semantic analysis to build the new OpenMP clause.
+  /// Subclasses may override this routine to provide different behavior.
+  OMPClause *RebuildOMPXDynCGroupMemClause(Expr *Size, SourceLocation StartLoc,
+                                           SourceLocation LParenLoc,
+                                           SourceLocation EndLoc) {
+    return getSema().ActOnOpenMPXDynCGroupMemClause(Size, StartLoc, LParenLoc,
+                                                    EndLoc);
+  }
+
   /// Build a new OpenMP 'align' clause.
   ///
   /// By default, performs semantic analysis to build the new OpenMP clause.
@@ -10633,6 +10644,16 @@ OMPClause *TreeTransform<Derived>::TransformOMPBindClause(OMPBindClause *C) {
       C->getLParenLoc(), C->getEndLoc());
 }
 
+template <typename Derived>
+OMPClause *TreeTransform<Derived>::TransformOMPXDynCGroupMemClause(
+    OMPXDynCGroupMemClause *C) {
+  ExprResult Size = getDerived().TransformExpr(C->getSize());
+  if (Size.isInvalid())
+    return nullptr;
+  return getDerived().RebuildOMPXDynCGroupMemClause(
+      Size.get(), C->getBeginLoc(), C->getLParenLoc(), C->getEndLoc());
+}
+
 //===----------------------------------------------------------------------===//
 // Expression transformation
 //===----------------------------------------------------------------------===//

diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 241821e..4d72596 100644 (file)
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -10272,6 +10272,9 @@ OMPClause *OMPClauseReader::readClause() {
   case llvm::omp::OMPC_align:
     C = new (Context) OMPAlignClause();
     break;
+  case llvm::omp::OMPC_ompx_dyn_cgroup_mem:
+    C = new (Context) OMPXDynCGroupMemClause();
+    break;
 #define OMP_CLAUSE_NO_CLASS(Enum, Str)                                         \
   case llvm::omp::Enum:                                                        \
     break;
@@ -11342,6 +11345,12 @@ void OMPClauseReader::VisitOMPAlignClause(OMPAlignClause *C) {
   C->setLParenLoc(Record.readSourceLocation());
 }
 
+void OMPClauseReader::VisitOMPXDynCGroupMemClause(OMPXDynCGroupMemClause *C) {
+  VisitOMPClauseWithPreInit(C);
+  C->setSize(Record.readSubExpr());
+  C->setLParenLoc(Record.readSourceLocation());
+}
+
 OMPTraitInfo *ASTRecordReader::readOMPTraitInfo() {
   OMPTraitInfo &TI = getContext().getNewOMPTraitInfo();
   TI.Sets.resize(readUInt32());

diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index 3265460..bdf1100 100644 (file)
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -7122,6 +7122,12 @@ void OMPClauseWriter::VisitOMPBindClause(OMPBindClause *C) {
   Record.AddSourceLocation(C->getBindKindLoc());
 }
 
+void OMPClauseWriter::VisitOMPXDynCGroupMemClause(OMPXDynCGroupMemClause *C) {
+  VisitOMPClauseWithPreInit(C);
+  Record.AddStmt(C->getSize());
+  Record.AddSourceLocation(C->getLParenLoc());
+}
+
 void ASTRecordWriter::writeOMPTraitInfo(const OMPTraitInfo *TI) {
   writeUInt32(TI->Sets.size());
   for (const auto &Set : TI->Sets) {

diff --git a/clang/test/OpenMP/target_ompx_dyn_cgroup_mem_codegen.cpp b/clang/test/OpenMP/target_ompx_dyn_cgroup_mem_codegen.cpp
new file mode 100644 (file)
index 0000000..a023946
--- /dev/null
+++ b/clang/test/OpenMP/target_ompx_dyn_cgroup_mem_codegen.cpp
@@ -0,0 +1,2656 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
+// Test host codegen.
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
+
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+
+// Test target codegen - host bc file has to be created first.
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK9
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK11
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
+
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+
+// Test host codegen.
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
+
+// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+
+// Test target codegen - host bc file has to be created first.
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --check-prefix=CHECK9
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
+// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --check-prefix=CHECK11
+// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
+
+// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
+// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
+// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
+// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
+
+// expected-no-diagnostics
+#ifndef HEADER
+#define HEADER
+
+
+
+
+// We have 6 target regions
+
+
+
+// Check target registration is registered as a Ctor.
+
+
+template<typename tx>
+tx ftemplate(int n) {
+  tx a = 0;
+
+  #pragma omp target teams ompx_dyn_cgroup_mem(tx(20))
+  {
+  }
+
+  short b = 1;
+  #pragma omp target teams num_teams(b) ompx_dyn_cgroup_mem(1024)
+  {
+    a += b;
+  }
+
+  return a;
+}
+
+static
+int fstatic(int n) {
+
+  #pragma omp target teams distribute parallel for simd num_teams(n) ompx_dyn_cgroup_mem(n*32)
+  for (int i = 0; i < n ; ++i) {
+  }
+
+  #pragma omp target teams ompx_dyn_cgroup_mem(32+n) nowait
+  {
+  }
+
+  return n+1;
+}
+
+struct S1 {
+  double a;
+
+  int r1(int n){
+    int b = 1;
+
+    #pragma omp target teams ompx_dyn_cgroup_mem(n-b)
+    {
+      this->a = (double)b + 1.5;
+    }
+
+    #pragma omp target ompx_dyn_cgroup_mem(1024)
+    {
+      this->a = 2.5;
+    }
+
+    return (int)a;
+  }
+};
+
+int bar(int n){
+  int a = 0;
+
+  S1 S;
+  a += S.r1(n);
+
+  a += fstatic(n);
+
+  a += ftemplate<int>(n);
+
+  return a;
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+// Check that the offloading functions are emitted and that the parallel function
+// is appropriately guarded.
+
+
+
+
+
+
+#endif
+
+// CHECK1-LABEL: define {{[^@]+}}@_Z3bari
+// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
+// CHECK1-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[A]], align 4
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(ptr noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP0]])
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A]], align 4
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[A]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP2]])
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A]], align 4
+// CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
+// CHECK1-NEXT:    store i32 [[ADD2]], ptr [[A]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP4]])
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[A]], align 4
+// CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
+// CHECK1-NEXT:    store i32 [[ADD4]], ptr [[A]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[A]], align 4
+// CHECK1-NEXT:    ret i32 [[TMP6]]
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
+// CHECK1-SAME: (ptr noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    store i32 1, ptr [[B]], align 4
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B]], align 4
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
+// CHECK1-NEXT:    store i32 [[SUB]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[B]], align 4
+// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[B_CASTED]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[B_CASTED]], align 8
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
+// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[THIS1]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP6]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store ptr [[A]], ptr [[TMP7]], align 8
+// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr null, ptr [[TMP8]], align 8
+// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
+// CHECK1-NEXT:    store i64 [[TMP3]], ptr [[TMP9]], align 8
+// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
+// CHECK1-NEXT:    store i64 [[TMP3]], ptr [[TMP10]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
+// CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
+// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
+// CHECK1-NEXT:    store i64 [[TMP5]], ptr [[TMP12]], align 8
+// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
+// CHECK1-NEXT:    store i64 [[TMP5]], ptr [[TMP13]], align 8
+// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
+// CHECK1-NEXT:    store ptr null, ptr [[TMP14]], align 8
+// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
+// CHECK1-NEXT:    store i32 2, ptr [[TMP18]], align 4
+// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
+// CHECK1-NEXT:    store i32 3, ptr [[TMP19]], align 4
+// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
+// CHECK1-NEXT:    store ptr [[TMP15]], ptr [[TMP20]], align 8
+// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
+// CHECK1-NEXT:    store ptr [[TMP16]], ptr [[TMP21]], align 8
+// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
+// CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP22]], align 8
+// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
+// CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP23]], align 8
+// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
+// CHECK1-NEXT:    store ptr null, ptr [[TMP24]], align 8
+// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
+// CHECK1-NEXT:    store ptr null, ptr [[TMP25]], align 8
+// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
+// CHECK1-NEXT:    store i64 0, ptr [[TMP26]], align 8
+// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
+// CHECK1-NEXT:    store i64 0, ptr [[TMP27]], align 8
+// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP28]], align 4
+// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP29]], align 4
+// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
+// CHECK1-NEXT:    store i32 [[TMP17]], ptr [[TMP30]], align 4
+// CHECK1-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
+// CHECK1-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
+// CHECK1:       omp_offload.failed:
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(ptr [[THIS1]], i64 [[TMP3]], i64 [[TMP5]]) #[[ATTR2:[0-9]+]]
+// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
+// CHECK1:       omp_offload.cont:
+// CHECK1-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
+// CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP33]], align 8
+// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
+// CHECK1-NEXT:    store ptr [[A2]], ptr [[TMP34]], align 8
+// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr null, ptr [[TMP35]], align 8
+// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
+// CHECK1-NEXT:    [[KERNEL_ARGS6:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
+// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 0
+// CHECK1-NEXT:    store i32 2, ptr [[TMP38]], align 4
+// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 1
+// CHECK1-NEXT:    store i32 1, ptr [[TMP39]], align 4
+// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 2
+// CHECK1-NEXT:    store ptr [[TMP36]], ptr [[TMP40]], align 8
+// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 3
+// CHECK1-NEXT:    store ptr [[TMP37]], ptr [[TMP41]], align 8
+// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 4
+// CHECK1-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP42]], align 8
+// CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 5
+// CHECK1-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP43]], align 8
+// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 6
+// CHECK1-NEXT:    store ptr null, ptr [[TMP44]], align 8
+// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 7
+// CHECK1-NEXT:    store ptr null, ptr [[TMP45]], align 8
+// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 8
+// CHECK1-NEXT:    store i64 0, ptr [[TMP46]], align 8
+// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 9
+// CHECK1-NEXT:    store i64 0, ptr [[TMP47]], align 8
+// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 10
+// CHECK1-NEXT:    store [3 x i32] [i32 -1, i32 0, i32 0], ptr [[TMP48]], align 4
+// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 11
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP49]], align 4
+// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 12
+// CHECK1-NEXT:    store i32 1024, ptr [[TMP50]], align 4
+// CHECK1-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 -1, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, ptr [[KERNEL_ARGS6]])
+// CHECK1-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
+// CHECK1-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
+// CHECK1:       omp_offload.failed7:
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(ptr [[THIS1]]) #[[ATTR2]]
+// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
+// CHECK1:       omp_offload.cont8:
+// CHECK1-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP53:%.*]] = load double, ptr [[A9]], align 8
+// CHECK1-NEXT:    [[CONV:%.*]] = fptosi double [[TMP53]] to i32
+// CHECK1-NEXT:    ret i32 [[CONV]]
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
+// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED8:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x ptr], align 8
+// CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK1-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP2]], 32
+// CHECK1-NEXT:    store i32 [[MUL]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 8
+// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP4]], ptr [[TMP9]], align 8
+// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP4]], ptr [[TMP10]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
+// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
+// CHECK1-NEXT:    store i64 [[TMP6]], ptr [[TMP12]], align 8
+// CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
+// CHECK1-NEXT:    store i64 [[TMP6]], ptr [[TMP13]], align 8
+// CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
+// CHECK1-NEXT:    store ptr null, ptr [[TMP14]], align 8
+// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
+// CHECK1-NEXT:    store i64 [[TMP8]], ptr [[TMP15]], align 8
+// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
+// CHECK1-NEXT:    store i64 [[TMP8]], ptr [[TMP16]], align 8
+// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
+// CHECK1-NEXT:    store ptr null, ptr [[TMP17]], align 8
+// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP21]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP22]], 0
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK1-NEXT:    [[SUB5:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK1-NEXT:    store i32 [[SUB5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP23]], 1
+// CHECK1-NEXT:    [[TMP24:%.*]] = zext i32 [[ADD]] to i64
+// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP26:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP20]], 0
+// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
+// CHECK1-NEXT:    store i32 2, ptr [[TMP27]], align 4
+// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
+// CHECK1-NEXT:    store i32 3, ptr [[TMP28]], align 4
+// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
+// CHECK1-NEXT:    store ptr [[TMP18]], ptr [[TMP29]], align 8
+// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
+// CHECK1-NEXT:    store ptr [[TMP19]], ptr [[TMP30]], align 8
+// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
+// CHECK1-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP31]], align 8
+// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
+// CHECK1-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP32]], align 8
+// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
+// CHECK1-NEXT:    store ptr null, ptr [[TMP33]], align 8
+// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
+// CHECK1-NEXT:    store ptr null, ptr [[TMP34]], align 8
+// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
+// CHECK1-NEXT:    store i64 [[TMP24]], ptr [[TMP35]], align 8
+// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
+// CHECK1-NEXT:    store i64 0, ptr [[TMP36]], align 8
+// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
+// CHECK1-NEXT:    store [3 x i32] [[TMP26]], ptr [[TMP37]], align 4
+// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP38]], align 4
+// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
+// CHECK1-NEXT:    store i32 [[TMP25]], ptr [[TMP39]], align 4
+// CHECK1-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP20]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
+// CHECK1-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
+// CHECK1:       omp_offload.failed:
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR2]]
+// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
+// CHECK1:       omp_offload.cont:
+// CHECK1-NEXT:    [[TMP42:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 32, [[TMP42]]
+// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK1-NEXT:    [[TMP43:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK1-NEXT:    store i32 [[TMP43]], ptr [[DOTCAPTURE_EXPR__CASTED8]], align 4
+// CHECK1-NEXT:    [[TMP44:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED8]], align 8
+// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP44]], ptr [[TMP45]], align 8
+// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP44]], ptr [[TMP46]], align 8
+// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr null, ptr [[TMP47]], align 8
+// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP51:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK1-NEXT:    store i32 [[TMP51]], ptr [[TMP50]], align 4
+// CHECK1-NEXT:    [[TMP52:%.*]] = call ptr @__kmpc_omp_target_task_alloc(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i64 64, i64 4, ptr @.omp_task_entry., i64 -1)
+// CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP52]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP53]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP55:%.*]] = load ptr, ptr [[TMP54]], align 8
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[TMP55]], ptr align 4 [[AGG_CAPTURED]], i64 4, i1 false)
+// CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP52]], i32 0, i32 1
+// CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP56]], i32 0, i32 0
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP57]], ptr align 8 [[TMP48]], i64 8, i1 false)
+// CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 1
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP58]], ptr align 8 [[TMP49]], i64 8, i1 false)
+// CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 2
+// CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP59]], ptr align 8 @.offload_sizes.8, i64 8, i1 false)
+// CHECK1-NEXT:    [[TMP60:%.*]] = call i32 @__kmpc_omp_task(ptr @[[GLOB1]], i32 [[TMP0]], ptr [[TMP52]])
+// CHECK1-NEXT:    [[TMP61:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP61]], 1
+// CHECK1-NEXT:    ret i32 [[ADD12]]
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
+// CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[B:%.*]] = alloca i16, align 2
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
+// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
+// CHECK1-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[A]], align 4
+// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
+// CHECK1-NEXT:    store i32 2, ptr [[TMP0]], align 4
+// CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
+// CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
+// CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
+// CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
+// CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
+// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
+// CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
+// CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
+// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
+// CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
+// CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
+// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
+// CHECK1-NEXT:    store i64 0, ptr [[TMP8]], align 8
+// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
+// CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
+// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
+// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
+// CHECK1-NEXT:    store i32 20, ptr [[TMP12]], align 4
+// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, ptr [[KERNEL_ARGS]])
+// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
+// CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
+// CHECK1:       omp_offload.failed:
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
+// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
+// CHECK1:       omp_offload.cont:
+// CHECK1-NEXT:    store i16 1, ptr [[B]], align 2
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i16, ptr [[B]], align 2
+// CHECK1-NEXT:    store i16 [[TMP15]], ptr [[DOTCAPTURE_EXPR_]], align 2
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[A]], align 4
+// CHECK1-NEXT:    store i32 [[TMP16]], ptr [[A_CASTED]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, ptr [[A_CASTED]], align 8
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i16, ptr [[B]], align 2
+// CHECK1-NEXT:    store i16 [[TMP18]], ptr [[B_CASTED]], align 2
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[B_CASTED]], align 8
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2
+// CHECK1-NEXT:    store i16 [[TMP20]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 2
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
+// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP17]], ptr [[TMP22]], align 8
+// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    store i64 [[TMP17]], ptr [[TMP23]], align 8
+// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr null, ptr [[TMP24]], align 8
+// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
+// CHECK1-NEXT:    store i64 [[TMP19]], ptr [[TMP25]], align 8
+// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
+// CHECK1-NEXT:    store i64 [[TMP19]], ptr [[TMP26]], align 8
+// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
+// CHECK1-NEXT:    store ptr null, ptr [[TMP27]], align 8
+// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
+// CHECK1-NEXT:    store i64 [[TMP21]], ptr [[TMP28]], align 8
+// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
+// CHECK1-NEXT:    store i64 [[TMP21]], ptr [[TMP29]], align 8
+// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
+// CHECK1-NEXT:    store ptr null, ptr [[TMP30]], align 8
+// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP33:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2
+// CHECK1-NEXT:    [[TMP34:%.*]] = sext i16 [[TMP33]] to i32
+// CHECK1-NEXT:    [[TMP35:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP34]], 0
+// CHECK1-NEXT:    [[KERNEL_ARGS1:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
+// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 0
+// CHECK1-NEXT:    store i32 2, ptr [[TMP36]], align 4
+// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 1
+// CHECK1-NEXT:    store i32 3, ptr [[TMP37]], align 4
+// CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 2
+// CHECK1-NEXT:    store ptr [[TMP31]], ptr [[TMP38]], align 8
+// CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 3
+// CHECK1-NEXT:    store ptr [[TMP32]], ptr [[TMP39]], align 8
+// CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 4
+// CHECK1-NEXT:    store ptr @.offload_sizes.13, ptr [[TMP40]], align 8
+// CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 5
+// CHECK1-NEXT:    store ptr @.offload_maptypes.14, ptr [[TMP41]], align 8
+// CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 6
+// CHECK1-NEXT:    store ptr null, ptr [[TMP42]], align 8
+// CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 7
+// CHECK1-NEXT:    store ptr null, ptr [[TMP43]], align 8
+// CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 8
+// CHECK1-NEXT:    store i64 0, ptr [[TMP44]], align 8
+// CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 9
+// CHECK1-NEXT:    store i64 0, ptr [[TMP45]], align 8
+// CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 10
+// CHECK1-NEXT:    store [3 x i32] [[TMP35]], ptr [[TMP46]], align 4
+// CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 11
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4
+// CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 12
+// CHECK1-NEXT:    store i32 1024, ptr [[TMP48]], align 4
+// CHECK1-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP34]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, ptr [[KERNEL_ARGS1]])
+// CHECK1-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
+// CHECK1-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
+// CHECK1:       omp_offload.failed2:
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i64 [[TMP17]], i64 [[TMP19]], i64 [[TMP21]]) #[[ATTR2]]
+// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
+// CHECK1:       omp_offload.cont3:
+// CHECK1-NEXT:    [[TMP51:%.*]] = load i32, ptr [[A]], align 4
+// CHECK1-NEXT:    ret i32 [[TMP51]]
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
+// CHECK1-SAME: (ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[B_CASTED]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[B_CASTED]], align 8
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @.omp_outlined., ptr [[TMP0]], i64 [[TMP2]])
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
+// CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
+// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK1-NEXT:    store double [[ADD]], ptr [[A]], align 8
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
+// CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK1-NEXT:    store double 2.500000e+00, ptr [[A]], align 8
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
+// CHECK1-SAME: (i64 noundef [[N:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK1-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @.omp_outlined..3, i64 [[TMP3]])
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK1:       omp.precond.then:
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
+// CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK1:       cond.true:
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    br label [[COND_END:%.*]]
+// CHECK1:       cond.false:
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    br label [[COND_END]]
+// CHECK1:       cond.end:
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK1-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1:       omp.inner.for.cond:
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11:![0-9]+]]
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1:       omp.inner.for.body:
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP13]] to i64
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP15]] to i64
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    store i32 [[TMP17]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i64, ptr [[N_CASTED]], align 8, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 3, ptr @.omp_outlined..4, i64 [[TMP14]], i64 [[TMP16]], i64 [[TMP18]]), !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK1:       omp.inner.for.inc:
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
+// CHECK1:       omp.inner.for.end:
+// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK1:       omp.loop.exit:
+// CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
+// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK1-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
+// CHECK1-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK1:       .omp.final.then:
+// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP25]], 0
+// CHECK1-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
+// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
+// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK1:       .omp.final.done:
+// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK1:       omp.precond.end:
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK1:       omp.precond.then:
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP4]] to i32
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
+// CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
+// CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK1:       cond.true:
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    br label [[COND_END:%.*]]
+// CHECK1:       cond.false:
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    br label [[COND_END]]
+// CHECK1:       cond.end:
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1:       omp.inner.for.cond:
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1:       omp.inner.for.body:
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK1:       omp.body.continue:
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK1:       omp.inner.for.inc:
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
+// CHECK1:       omp.inner.for.end:
+// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK1:       omp.loop.exit:
+// CHECK1-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP18]])
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK1-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
+// CHECK1-NEXT:    br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK1:       .omp.final.then:
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP21]], 0
+// CHECK1-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
+// CHECK1-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
+// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
+// CHECK1-NEXT:    store i32 [[ADD11]], ptr [[I4]], align 4
+// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK1:       .omp.final.done:
+// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK1:       omp.precond.end:
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
+// CHECK1-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @.omp_outlined..7)
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
+// CHECK1-SAME: (ptr noalias noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]], ptr noalias noundef [[TMP2:%.*]], ptr noalias noundef [[TMP3:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTADDR3:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
+// CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
+// CHECK1-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2]], align 8
+// CHECK1-NEXT:    store ptr [[TMP3]], ptr [[DOTADDR3]], align 8
+// CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP4]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
+// CHECK1-NEXT:    store ptr [[TMP5]], ptr [[TMP6]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 1
+// CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2]], align 8
+// CHECK1-NEXT:    store ptr [[TMP7]], ptr [[TMP8]], align 8
+// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 2
+// CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTADDR3]], align 8
+// CHECK1-NEXT:    store ptr [[TMP9]], ptr [[TMP10]], align 8
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
+// CHECK1-SAME: (i32 noundef signext [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[KERNEL_ARGS_I:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    store i32 [[TMP0]], ptr [[DOTADDR]], align 4
+// CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTADDR]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
+// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP3]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP4]], i32 0, i32 2
+// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], ptr [[TMP4]], i32 0, i32 0
+// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
+// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP3]], i32 0, i32 1
+// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
+// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META21:![0-9]+]])
+// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META23:![0-9]+]])
+// CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META25:![0-9]+]])
+// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !27
+// CHECK1-NEXT:    store ptr [[TMP5]], ptr [[DOTPART_ID__ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    store ptr [[TMP8]], ptr [[DOTPRIVATES__ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    store ptr @.omp_task_privates_map., ptr [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    store ptr [[TMP3]], ptr [[DOTTASK_T__ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    store ptr [[TMP7]], ptr [[__CONTEXT_ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[__CONTEXT_ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTPRIVATES__ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    call void [[TMP10]](ptr [[TMP11]], ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]]) #[[ATTR2]]
+// CHECK1-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP9]], align 4
+// CHECK1-NEXT:    store i32 2, ptr [[KERNEL_ARGS_I]], align 4, !noalias !27
+// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 1
+// CHECK1-NEXT:    store i32 1, ptr [[TMP16]], align 4, !noalias !27
+// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 2
+// CHECK1-NEXT:    store ptr [[TMP12]], ptr [[TMP17]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 3
+// CHECK1-NEXT:    store ptr [[TMP13]], ptr [[TMP18]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 4
+// CHECK1-NEXT:    store ptr [[TMP14]], ptr [[TMP19]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 5
+// CHECK1-NEXT:    store ptr @.offload_maptypes.9, ptr [[TMP20]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 6
+// CHECK1-NEXT:    store ptr null, ptr [[TMP21]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 7
+// CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 8
+// CHECK1-NEXT:    store i64 0, ptr [[TMP23]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 9
+// CHECK1-NEXT:    store i64 1, ptr [[TMP24]], align 8, !noalias !27
+// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 10
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4, !noalias !27
+// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 11
+// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4, !noalias !27
+// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 12
+// CHECK1-NEXT:    store i32 [[TMP15]], ptr [[TMP27]], align 4, !noalias !27
+// CHECK1-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, ptr [[KERNEL_ARGS_I]])
+// CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
+// CHECK1-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__10_EXIT:%.*]]
+// CHECK1:       omp_offload.failed.i:
+// CHECK1-NEXT:    [[TMP30:%.*]] = load i32, ptr [[TMP9]], align 4
+// CHECK1-NEXT:    store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !27
+// CHECK1-NEXT:    [[TMP31:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !27
+// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i64 [[TMP31]]) #[[ATTR2]]
+// CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__10_EXIT]]
+// CHECK1:       .omp_outlined..10.exit:
+// CHECK1-NEXT:    ret i32 0
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
+// CHECK1-SAME: () #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @.omp_outlined..11)
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
+// CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2
+// CHECK1-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
+// CHECK1-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[A_CASTED]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[A_CASTED]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK1-NEXT:    store i16 [[TMP5]], ptr [[B_CASTED]], align 2
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[B_CASTED]], align 8
+// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @.omp_outlined..12, i64 [[TMP4]], i64 [[TMP6]])
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..12
+// CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR1]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
+// CHECK1-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK1-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[A_ADDR]], align 4
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
+// CHECK1-SAME: () #[[ATTR3]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@_Z3bari
+// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
+// CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 0, ptr [[A]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2S12r1Ei(ptr noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP0]])
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A]], align 4
+// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
+// CHECK3-NEXT:    store i32 [[ADD]], ptr [[A]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP2]])
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A]], align 4
+// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
+// CHECK3-NEXT:    store i32 [[ADD2]], ptr [[A]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[CALL3:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP4]])
+// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[A]], align 4
+// CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
+// CHECK3-NEXT:    store i32 [[ADD4]], ptr [[A]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[A]], align 4
+// CHECK3-NEXT:    ret i32 [[TMP6]]
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
+// CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 4
+// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    store i32 1, ptr [[B]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B]], align 4
+// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP0]], [[TMP1]]
+// CHECK3-NEXT:    store i32 [[SUB]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[B]], align 4
+// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[B_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[B_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
+// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
+// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[THIS1]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP6]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr [[A]], ptr [[TMP7]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr null, ptr [[TMP8]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 [[TMP3]], ptr [[TMP9]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 [[TMP3]], ptr [[TMP10]], align 4
+// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
+// CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
+// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[TMP12]], align 4
+// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
+// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[TMP13]], align 4
+// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr null, ptr [[TMP14]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 2, ptr [[TMP18]], align 4
+// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 3, ptr [[TMP19]], align 4
+// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr [[TMP15]], ptr [[TMP20]], align 4
+// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
+// CHECK3-NEXT:    store ptr [[TMP16]], ptr [[TMP21]], align 4
+// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
+// CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP22]], align 4
+// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
+// CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP23]], align 4
+// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
+// CHECK3-NEXT:    store ptr null, ptr [[TMP24]], align 4
+// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
+// CHECK3-NEXT:    store ptr null, ptr [[TMP25]], align 4
+// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
+// CHECK3-NEXT:    store i64 0, ptr [[TMP26]], align 8
+// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
+// CHECK3-NEXT:    store i64 0, ptr [[TMP27]], align 8
+// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP28]], align 4
+// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP29]], align 4
+// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
+// CHECK3-NEXT:    store i32 [[TMP17]], ptr [[TMP30]], align 4
+// CHECK3-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
+// CHECK3-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
+// CHECK3:       omp_offload.failed:
+// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121(ptr [[THIS1]], i32 [[TMP3]], i32 [[TMP5]]) #[[ATTR2:[0-9]+]]
+// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
+// CHECK3:       omp_offload.cont:
+// CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP33]], align 4
+// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr [[A2]], ptr [[TMP34]], align 4
+// CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr null, ptr [[TMP35]], align 4
+// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
+// CHECK3-NEXT:    [[KERNEL_ARGS6:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
+// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 2, ptr [[TMP38]], align 4
+// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 1, ptr [[TMP39]], align 4
+// CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr [[TMP36]], ptr [[TMP40]], align 4
+// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 3
+// CHECK3-NEXT:    store ptr [[TMP37]], ptr [[TMP41]], align 4
+// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 4
+// CHECK3-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP42]], align 4
+// CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 5
+// CHECK3-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP43]], align 4
+// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 6
+// CHECK3-NEXT:    store ptr null, ptr [[TMP44]], align 4
+// CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 7
+// CHECK3-NEXT:    store ptr null, ptr [[TMP45]], align 4
+// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 8
+// CHECK3-NEXT:    store i64 0, ptr [[TMP46]], align 8
+// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 9
+// CHECK3-NEXT:    store i64 0, ptr [[TMP47]], align 8
+// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 10
+// CHECK3-NEXT:    store [3 x i32] [i32 -1, i32 0, i32 0], ptr [[TMP48]], align 4
+// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 11
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP49]], align 4
+// CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS6]], i32 0, i32 12
+// CHECK3-NEXT:    store i32 1024, ptr [[TMP50]], align 4
+// CHECK3-NEXT:    [[TMP51:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 -1, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126.region_id, ptr [[KERNEL_ARGS6]])
+// CHECK3-NEXT:    [[TMP52:%.*]] = icmp ne i32 [[TMP51]], 0
+// CHECK3-NEXT:    br i1 [[TMP52]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
+// CHECK3:       omp_offload.failed7:
+// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126(ptr [[THIS1]]) #[[ATTR2]]
+// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT8]]
+// CHECK3:       omp_offload.cont8:
+// CHECK3-NEXT:    [[A9:%.*]] = getelementptr inbounds [[STRUCT_S1]], ptr [[THIS1]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP53:%.*]] = load double, ptr [[A9]], align 4
+// CHECK3-NEXT:    [[CONV:%.*]] = fptosi double [[TMP53]] to i32
+// CHECK3-NEXT:    ret i32 [[CONV]]
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
+// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED2:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED8:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x ptr], align 4
+// CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP2]], 32
+// CHECK3-NEXT:    store i32 [[MUL]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED2]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 [[TMP4]], ptr [[TMP9]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 [[TMP4]], ptr [[TMP10]], align 4
+// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[TMP12]], align 4
+// CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 [[TMP6]], ptr [[TMP13]], align 4
+// CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
+// CHECK3-NEXT:    store ptr null, ptr [[TMP14]], align 4
+// CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
+// CHECK3-NEXT:    store i32 [[TMP8]], ptr [[TMP15]], align 4
+// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
+// CHECK3-NEXT:    store i32 [[TMP8]], ptr [[TMP16]], align 4
+// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr null, ptr [[TMP17]], align 4
+// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP21]], ptr [[DOTCAPTURE_EXPR_3]], align 4
+// CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_3]], align 4
+// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP22]], 0
+// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK3-NEXT:    [[SUB5:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK3-NEXT:    store i32 [[SUB5]], ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
+// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP23]], 1
+// CHECK3-NEXT:    [[TMP24:%.*]] = zext i32 [[ADD]] to i64
+// CHECK3-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    [[TMP26:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP20]], 0
+// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 2, ptr [[TMP27]], align 4
+// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 3, ptr [[TMP28]], align 4
+// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr [[TMP18]], ptr [[TMP29]], align 4
+// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
+// CHECK3-NEXT:    store ptr [[TMP19]], ptr [[TMP30]], align 4
+// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
+// CHECK3-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP31]], align 4
+// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
+// CHECK3-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP32]], align 4
+// CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
+// CHECK3-NEXT:    store ptr null, ptr [[TMP33]], align 4
+// CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
+// CHECK3-NEXT:    store ptr null, ptr [[TMP34]], align 4
+// CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
+// CHECK3-NEXT:    store i64 [[TMP24]], ptr [[TMP35]], align 8
+// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
+// CHECK3-NEXT:    store i64 0, ptr [[TMP36]], align 8
+// CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
+// CHECK3-NEXT:    store [3 x i32] [[TMP26]], ptr [[TMP37]], align 4
+// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP38]], align 4
+// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
+// CHECK3-NEXT:    store i32 [[TMP25]], ptr [[TMP39]], align 4
+// CHECK3-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP20]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
+// CHECK3-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
+// CHECK3:       omp_offload.failed:
+// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104(i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR2]]
+// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
+// CHECK3:       omp_offload.cont:
+// CHECK3-NEXT:    [[TMP42:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 32, [[TMP42]]
+// CHECK3-NEXT:    store i32 [[ADD7]], ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK3-NEXT:    [[TMP43:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK3-NEXT:    store i32 [[TMP43]], ptr [[DOTCAPTURE_EXPR__CASTED8]], align 4
+// CHECK3-NEXT:    [[TMP44:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED8]], align 4
+// CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 [[TMP44]], ptr [[TMP45]], align 4
+// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 [[TMP44]], ptr [[TMP46]], align 4
+// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS11]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr null, ptr [[TMP47]], align 4
+// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP51:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK3-NEXT:    store i32 [[TMP51]], ptr [[TMP50]], align 4
+// CHECK3-NEXT:    [[TMP52:%.*]] = call ptr @__kmpc_omp_target_task_alloc(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 36, i32 4, ptr @.omp_task_entry., i64 -1)
+// CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP52]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP53]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP55:%.*]] = load ptr, ptr [[TMP54]], align 4
+// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP55]], ptr align 4 [[AGG_CAPTURED]], i32 4, i1 false)
+// CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP52]], i32 0, i32 1
+// CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP56]], i32 0, i32 0
+// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP57]], ptr align 4 @.offload_sizes.8, i32 8, i1 false)
+// CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 1
+// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP58]], ptr align 4 [[TMP48]], i32 4, i1 false)
+// CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP56]], i32 0, i32 2
+// CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[TMP59]], ptr align 4 [[TMP49]], i32 4, i1 false)
+// CHECK3-NEXT:    [[TMP60:%.*]] = call i32 @__kmpc_omp_task(ptr @[[GLOB1]], i32 [[TMP0]], ptr [[TMP52]])
+// CHECK3-NEXT:    [[TMP61:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP61]], 1
+// CHECK3-NEXT:    ret i32 [[ADD12]]
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
+// CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B:%.*]] = alloca i16, align 2
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i16, align 2
+// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
+// CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 0, ptr [[A]], align 4
+// CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 2, ptr [[TMP0]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 0, ptr [[TMP1]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
+// CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
+// CHECK3-NEXT:    store ptr null, ptr [[TMP3]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
+// CHECK3-NEXT:    store ptr null, ptr [[TMP4]], align 4
+// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
+// CHECK3-NEXT:    store ptr null, ptr [[TMP5]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
+// CHECK3-NEXT:    store ptr null, ptr [[TMP6]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
+// CHECK3-NEXT:    store ptr null, ptr [[TMP7]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
+// CHECK3-NEXT:    store i64 0, ptr [[TMP8]], align 8
+// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
+// CHECK3-NEXT:    store i64 0, ptr [[TMP9]], align 8
+// CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
+// CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
+// CHECK3-NEXT:    store i32 20, ptr [[TMP12]], align 4
+// CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88.region_id, ptr [[KERNEL_ARGS]])
+// CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
+// CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
+// CHECK3:       omp_offload.failed:
+// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88() #[[ATTR2]]
+// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
+// CHECK3:       omp_offload.cont:
+// CHECK3-NEXT:    store i16 1, ptr [[B]], align 2
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i16, ptr [[B]], align 2
+// CHECK3-NEXT:    store i16 [[TMP15]], ptr [[DOTCAPTURE_EXPR_]], align 2
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[A]], align 4
+// CHECK3-NEXT:    store i32 [[TMP16]], ptr [[A_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[A_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP18:%.*]] = load i16, ptr [[B]], align 2
+// CHECK3-NEXT:    store i16 [[TMP18]], ptr [[B_CASTED]], align 2
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[B_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2
+// CHECK3-NEXT:    store i16 [[TMP20]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 2
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
+// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 [[TMP17]], ptr [[TMP22]], align 4
+// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 [[TMP17]], ptr [[TMP23]], align 4
+// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
+// CHECK3-NEXT:    store ptr null, ptr [[TMP24]], align 4
+// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 [[TMP19]], ptr [[TMP25]], align 4
+// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 [[TMP19]], ptr [[TMP26]], align 4
+// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
+// CHECK3-NEXT:    store ptr null, ptr [[TMP27]], align 4
+// CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
+// CHECK3-NEXT:    store i32 [[TMP21]], ptr [[TMP28]], align 4
+// CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
+// CHECK3-NEXT:    store i32 [[TMP21]], ptr [[TMP29]], align 4
+// CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr null, ptr [[TMP30]], align 4
+// CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP33:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR_]], align 2
+// CHECK3-NEXT:    [[TMP34:%.*]] = sext i16 [[TMP33]] to i32
+// CHECK3-NEXT:    [[TMP35:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP34]], 0
+// CHECK3-NEXT:    [[KERNEL_ARGS1:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
+// CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 0
+// CHECK3-NEXT:    store i32 2, ptr [[TMP36]], align 4
+// CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 3, ptr [[TMP37]], align 4
+// CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr [[TMP31]], ptr [[TMP38]], align 4
+// CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 3
+// CHECK3-NEXT:    store ptr [[TMP32]], ptr [[TMP39]], align 4
+// CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 4
+// CHECK3-NEXT:    store ptr @.offload_sizes.13, ptr [[TMP40]], align 4
+// CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 5
+// CHECK3-NEXT:    store ptr @.offload_maptypes.14, ptr [[TMP41]], align 4
+// CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 6
+// CHECK3-NEXT:    store ptr null, ptr [[TMP42]], align 4
+// CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 7
+// CHECK3-NEXT:    store ptr null, ptr [[TMP43]], align 4
+// CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 8
+// CHECK3-NEXT:    store i64 0, ptr [[TMP44]], align 8
+// CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 9
+// CHECK3-NEXT:    store i64 0, ptr [[TMP45]], align 8
+// CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 10
+// CHECK3-NEXT:    store [3 x i32] [[TMP35]], ptr [[TMP46]], align 4
+// CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 11
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4
+// CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS1]], i32 0, i32 12
+// CHECK3-NEXT:    store i32 1024, ptr [[TMP48]], align 4
+// CHECK3-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 [[TMP34]], i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93.region_id, ptr [[KERNEL_ARGS1]])
+// CHECK3-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
+// CHECK3-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
+// CHECK3:       omp_offload.failed2:
+// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93(i32 [[TMP17]], i32 [[TMP19]], i32 [[TMP21]]) #[[ATTR2]]
+// CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
+// CHECK3:       omp_offload.cont3:
+// CHECK3-NEXT:    [[TMP51:%.*]] = load i32, ptr [[A]], align 4
+// CHECK3-NEXT:    ret i32 [[TMP51]]
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
+// CHECK3-SAME: (ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP1]], ptr [[B_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[B_CASTED]], align 4
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @.omp_outlined., ptr [[TMP0]], i32 [[TMP2]])
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
+// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
+// CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
+// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK3-NEXT:    store double [[ADD]], ptr [[A]], align 4
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
+// CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK3-NEXT:    store double 2.500000e+00, ptr [[A]], align 4
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
+// CHECK3-SAME: (i32 noundef [[N:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK3-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 1, ptr @.omp_outlined..3, i32 [[TMP3]])
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
+// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK3:       omp.precond.then:
+// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
+// CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK3:       cond.true:
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    br label [[COND_END:%.*]]
+// CHECK3:       cond.false:
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    br label [[COND_END]]
+// CHECK3:       cond.end:
+// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK3-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK3:       omp.inner.for.cond:
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12:![0-9]+]]
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3:       omp.inner.for.body:
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    store i32 [[TMP15]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB1]], i32 3, ptr @.omp_outlined..4, i32 [[TMP13]], i32 [[TMP14]], i32 [[TMP16]]), !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK3:       omp.inner.for.inc:
+// CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
+// CHECK3:       omp.inner.for.end:
+// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK3:       omp.loop.exit:
+// CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP20]])
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK3-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
+// CHECK3-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK3:       .omp.final.then:
+// CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP23]], 0
+// CHECK3-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
+// CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK3-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
+// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK3:       .omp.final.done:
+// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK3:       omp.precond.end:
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
+// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK3:       omp.precond.then:
+// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_LB]], align 4
+// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK3-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB3:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
+// CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK3:       cond.true:
+// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK3-NEXT:    br label [[COND_END:%.*]]
+// CHECK3:       cond.false:
+// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    br label [[COND_END]]
+// CHECK3:       cond.end:
+// CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
+// CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK3-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK3:       omp.inner.for.cond:
+// CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16:![0-9]+]]
+// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP16]]
+// CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK3:       omp.inner.for.body:
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
+// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
+// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK3-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP16]]
+// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK3:       omp.body.continue:
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK3:       omp.inner.for.inc:
+// CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
+// CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK3-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP16]]
+// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
+// CHECK3:       omp.inner.for.end:
+// CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK3:       omp.loop.exit:
+// CHECK3-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
+// CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP18]])
+// CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK3-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
+// CHECK3-NEXT:    br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK3:       .omp.final.then:
+// CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK3-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP21]], 0
+// CHECK3-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
+// CHECK3-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
+// CHECK3-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
+// CHECK3-NEXT:    store i32 [[ADD10]], ptr [[I3]], align 4
+// CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK3:       .omp.final.done:
+// CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK3:       omp.precond.end:
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
+// CHECK3-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @.omp_outlined..7)
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
+// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_task_privates_map.
+// CHECK3-SAME: (ptr noalias noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]], ptr noalias noundef [[TMP2:%.*]], ptr noalias noundef [[TMP3:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTADDR2:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTADDR3:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
+// CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
+// CHECK3-NEXT:    store ptr [[TMP2]], ptr [[DOTADDR2]], align 4
+// CHECK3-NEXT:    store ptr [[TMP3]], ptr [[DOTADDR3]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 4
+// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP4]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTADDR3]], align 4
+// CHECK3-NEXT:    store ptr [[TMP5]], ptr [[TMP6]], align 4
+// CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 1
+// CHECK3-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
+// CHECK3-NEXT:    store ptr [[TMP7]], ptr [[TMP8]], align 4
+// CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], ptr [[TMP4]], i32 0, i32 2
+// CHECK3-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTADDR2]], align 4
+// CHECK3-NEXT:    store ptr [[TMP9]], ptr [[TMP10]], align 4
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
+// CHECK3-SAME: (i32 noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[KERNEL_ARGS_I:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
+// CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTADDR]], align 4
+// CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
+// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTADDR]], align 4
+// CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP3]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP4]], i32 0, i32 2
+// CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], ptr [[TMP4]], i32 0, i32 0
+// CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
+// CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP3]], i32 0, i32 1
+// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
+// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META22:![0-9]+]])
+// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META24:![0-9]+]])
+// CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META26:![0-9]+]])
+// CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    store ptr [[TMP5]], ptr [[DOTPART_ID__ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    store ptr [[TMP8]], ptr [[DOTPRIVATES__ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    store ptr @.omp_task_privates_map., ptr [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    store ptr [[TMP3]], ptr [[DOTTASK_T__ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    store ptr [[TMP7]], ptr [[__CONTEXT_ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[__CONTEXT_ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTPRIVATES__ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    call void [[TMP10]](ptr [[TMP11]], ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]]) #[[ATTR2]]
+// CHECK3-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP9]], align 4
+// CHECK3-NEXT:    store i32 2, ptr [[KERNEL_ARGS_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 1
+// CHECK3-NEXT:    store i32 1, ptr [[TMP16]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 2
+// CHECK3-NEXT:    store ptr [[TMP12]], ptr [[TMP17]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 3
+// CHECK3-NEXT:    store ptr [[TMP13]], ptr [[TMP18]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 4
+// CHECK3-NEXT:    store ptr [[TMP14]], ptr [[TMP19]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 5
+// CHECK3-NEXT:    store ptr @.offload_maptypes.9, ptr [[TMP20]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 6
+// CHECK3-NEXT:    store ptr null, ptr [[TMP21]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 7
+// CHECK3-NEXT:    store ptr null, ptr [[TMP22]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 8
+// CHECK3-NEXT:    store i64 0, ptr [[TMP23]], align 8, !noalias !28
+// CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 9
+// CHECK3-NEXT:    store i64 1, ptr [[TMP24]], align 8, !noalias !28
+// CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 10
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 11
+// CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS_I]], i32 0, i32 12
+// CHECK3-NEXT:    store i32 [[TMP15]], ptr [[TMP27]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB1]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108.region_id, ptr [[KERNEL_ARGS_I]])
+// CHECK3-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
+// CHECK3-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__10_EXIT:%.*]]
+// CHECK3:       omp_offload.failed.i:
+// CHECK3-NEXT:    [[TMP30:%.*]] = load i32, ptr [[TMP9]], align 4
+// CHECK3-NEXT:    store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !28
+// CHECK3-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !28
+// CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108(i32 [[TMP31]]) #[[ATTR2]]
+// CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__10_EXIT]]
+// CHECK3:       .omp_outlined..10.exit:
+// CHECK3-NEXT:    ret i32 0
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
+// CHECK3-SAME: () #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 0, ptr @.omp_outlined..11)
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
+// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
+// CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]])
+// CHECK3-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2
+// CHECK3-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
+// CHECK3-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB1]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
+// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[TMP3]], ptr [[A_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[A_CASTED]], align 4
+// CHECK3-NEXT:    [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK3-NEXT:    store i16 [[TMP5]], ptr [[B_CASTED]], align 2
+// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[B_CASTED]], align 4
+// CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB1]], i32 2, ptr @.omp_outlined..12, i32 [[TMP4]], i32 [[TMP6]])
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..12
+// CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR1]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
+// CHECK3-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK3-NEXT:    [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
+// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]]
+// CHECK3-NEXT:    store i32 [[ADD]], ptr [[A_ADDR]], align 4
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
+// CHECK3-SAME: () #[[ATTR3]] {
+// CHECK3-NEXT:  entry:
+// CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
+// CHECK3-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
+// CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3:[0-9]+]])
+// CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 8
+// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK9-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
+// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK9-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
+// CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[N_CASTED]], align 8
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @.omp_outlined., i64 [[TMP3]])
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
+// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK9:       omp.precond.then:
+// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK9-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
+// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
+// CHECK9-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK9:       cond.true:
+// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    br label [[COND_END:%.*]]
+// CHECK9:       cond.false:
+// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK9-NEXT:    br label [[COND_END]]
+// CHECK9:       cond.end:
+// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK9-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK9:       omp.inner.for.cond:
+// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10:![0-9]+]]
+// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK9:       omp.inner.for.body:
+// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP13]] to i64
+// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP15]] to i64
+// CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    store i32 [[TMP17]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    [[TMP18:%.*]] = load i64, ptr [[N_CASTED]], align 8, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @.omp_outlined..1, i64 [[TMP14]], i64 [[TMP16]], i64 [[TMP18]]), !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK9:       omp.inner.for.inc:
+// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
+// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
+// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
+// CHECK9:       omp.inner.for.end:
+// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK9:       omp.loop.exit:
+// CHECK9-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
+// CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK9-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
+// CHECK9-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK9:       .omp.final.then:
+// CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP25]], 0
+// CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
+// CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK9-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
+// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK9:       .omp.final.done:
+// CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK9:       omp.precond.end:
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
+// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK9-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK9:       omp.precond.then:
+// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
+// CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP4]] to i32
+// CHECK9-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
+// CHECK9-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP5]] to i32
+// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK9-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
+// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
+// CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK9:       cond.true:
+// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK9-NEXT:    br label [[COND_END:%.*]]
+// CHECK9:       cond.false:
+// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    br label [[COND_END]]
+// CHECK9:       cond.end:
+// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
+// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK9-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
+// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK9:       omp.inner.for.cond:
+// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14:![0-9]+]]
+// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
+// CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK9:       omp.inner.for.body:
+// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
+// CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
+// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK9-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP14]]
+// CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK9:       omp.body.continue:
+// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK9:       omp.inner.for.inc:
+// CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
+// CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK9-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
+// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
+// CHECK9:       omp.inner.for.end:
+// CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK9:       omp.loop.exit:
+// CHECK9-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
+// CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP18]])
+// CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK9-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
+// CHECK9-NEXT:    br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK9:       .omp.final.then:
+// CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK9-NEXT:    [[SUB8:%.*]] = sub nsw i32 [[TMP21]], 0
+// CHECK9-NEXT:    [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
+// CHECK9-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
+// CHECK9-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
+// CHECK9-NEXT:    store i32 [[ADD11]], ptr [[I4]], align 4
+// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK9:       .omp.final.done:
+// CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK9:       omp.precond.end:
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
+// CHECK9-SAME: (i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @.omp_outlined..2)
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
+// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
+// CHECK9-SAME: (ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK9-NEXT:    store i32 [[TMP1]], ptr [[B_CASTED]], align 4
+// CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[B_CASTED]], align 8
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @.omp_outlined..3, ptr [[TMP0]], i64 [[TMP2]])
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
+// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
+// CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
+// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK9-NEXT:    store double [[ADD]], ptr [[A]], align 8
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
+// CHECK9-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
+// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK9-NEXT:    store double 2.500000e+00, ptr [[A]], align 8
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
+// CHECK9-SAME: () #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @.omp_outlined..4)
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
+// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
+// CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[B:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// CHECK9-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
+// CHECK9-NEXT:    [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2
+// CHECK9-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
+// CHECK9-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
+// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK9-NEXT:    store i32 [[TMP3]], ptr [[A_CASTED]], align 4
+// CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[A_CASTED]], align 8
+// CHECK9-NEXT:    [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK9-NEXT:    store i16 [[TMP5]], ptr [[B_CASTED]], align 2
+// CHECK9-NEXT:    [[TMP6:%.*]] = load i64, ptr [[B_CASTED]], align 8
+// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @.omp_outlined..5, i64 [[TMP4]], i64 [[TMP6]])
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
+// CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0]] {
+// CHECK9-NEXT:  entry:
+// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
+// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
+// CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
+// CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
+// CHECK9-NEXT:    store i64 [[B]], ptr [[B_ADDR]], align 8
+// CHECK9-NEXT:    [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK9-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
+// CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]]
+// CHECK9-NEXT:    store i32 [[ADD]], ptr [[A_ADDR]], align 4
+// CHECK9-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l104
+// CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3:[0-9]+]])
+// CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], ptr [[DOTCAPTURE_EXPR__ADDR2]], align 4
+// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK11-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP1]], i32 0)
+// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
+// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @.omp_outlined., i32 [[TMP3]])
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
+// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK11:       omp.precond.then:
+// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK11-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
+// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
+// CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK11:       cond.true:
+// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    br label [[COND_END:%.*]]
+// CHECK11:       cond.false:
+// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK11-NEXT:    br label [[COND_END]]
+// CHECK11:       cond.end:
+// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
+// CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
+// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
+// CHECK11-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4
+// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK11:       omp.inner.for.cond:
+// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11:![0-9]+]]
+// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
+// CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK11:       omp.inner.for.body:
+// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[N_ADDR]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    store i32 [[TMP15]], ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[N_CASTED]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @.omp_outlined..1, i32 [[TMP13]], i32 [[TMP14]], i32 [[TMP16]]), !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK11:       omp.inner.for.inc:
+// CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
+// CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
+// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
+// CHECK11:       omp.inner.for.end:
+// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK11:       omp.loop.exit:
+// CHECK11-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
+// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
+// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK11-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
+// CHECK11-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK11:       .omp.final.then:
+// CHECK11-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP23]], 0
+// CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
+// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
+// CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK11-NEXT:    store i32 [[ADD8]], ptr [[I3]], align 4
+// CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK11:       .omp.final.done:
+// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK11:       omp.precond.end:
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
+// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP1]], 0
+// CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
+// CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP2]]
+// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK11:       omp.precond.then:
+// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_UB]], align 4
+// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_LB]], align 4
+// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
+// CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK11-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
+// CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
+// CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK11:       cond.true:
+// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK11-NEXT:    br label [[COND_END:%.*]]
+// CHECK11:       cond.false:
+// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK11-NEXT:    br label [[COND_END]]
+// CHECK11:       cond.end:
+// CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
+// CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK11-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
+// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK11:       omp.inner.for.cond:
+// CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
+// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
+// CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK11:       omp.inner.for.body:
+// CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
+// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK11:       omp.body.continue:
+// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK11:       omp.inner.for.inc:
+// CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
+// CHECK11-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
+// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
+// CHECK11:       omp.inner.for.end:
+// CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK11:       omp.loop.exit:
+// CHECK11-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
+// CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP18]])
+// CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK11-NEXT:    [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
+// CHECK11-NEXT:    br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
+// CHECK11:       .omp.final.then:
+// CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK11-NEXT:    [[SUB7:%.*]] = sub nsw i32 [[TMP21]], 0
+// CHECK11-NEXT:    [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
+// CHECK11-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
+// CHECK11-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
+// CHECK11-NEXT:    store i32 [[ADD10]], ptr [[I3]], align 4
+// CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
+// CHECK11:       .omp.final.done:
+// CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK11:       omp.precond.end:
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l108
+// CHECK11-SAME: (i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @.omp_outlined..2)
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
+// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l121
+// CHECK11-SAME: (ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[TMP1]], ptr [[B_CASTED]], align 4
+// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[B_CASTED]], align 4
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @.omp_outlined..3, ptr [[TMP0]], i32 [[TMP2]])
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
+// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[B_ADDR]], align 4
+// CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to double
+// CHECK11-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
+// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK11-NEXT:    store double [[ADD]], ptr [[A]], align 4
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l126
+// CHECK11-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
+// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], ptr [[TMP0]], i32 0, i32 0
+// CHECK11-NEXT:    store double 2.500000e+00, ptr [[A]], align 4
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l88
+// CHECK11-SAME: () #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @.omp_outlined..4)
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
+// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l93
+// CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[B:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
+// CHECK11-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
+// CHECK11-NEXT:    [[TMP1:%.*]] = load i16, ptr [[DOTCAPTURE_EXPR__ADDR]], align 2
+// CHECK11-NEXT:    [[TMP2:%.*]] = sext i16 [[TMP1]] to i32
+// CHECK11-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 0)
+// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[TMP3]], ptr [[A_CASTED]], align 4
+// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[A_CASTED]], align 4
+// CHECK11-NEXT:    [[TMP5:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK11-NEXT:    store i16 [[TMP5]], ptr [[B_CASTED]], align 2
+// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[B_CASTED]], align 4
+// CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @.omp_outlined..5, i32 [[TMP4]], i32 [[TMP6]])
+// CHECK11-NEXT:    ret void
+//
+//
+// CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
+// CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[B:%.*]]) #[[ATTR0]] {
+// CHECK11-NEXT:  entry:
+// CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
+// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
+// CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
+// CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
+// CHECK11-NEXT:    store i32 [[B]], ptr [[B_ADDR]], align 4
+// CHECK11-NEXT:    [[TMP0:%.*]] = load i16, ptr [[B_ADDR]], align 2
+// CHECK11-NEXT:    [[CONV:%.*]] = sext i16 [[TMP0]] to i32
+// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
+// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CONV]]
+// CHECK11-NEXT:    store i32 [[ADD]], ptr [[A_ADDR]], align 4
+// CHECK11-NEXT:    ret void
+//

diff --git a/clang/test/OpenMP/target_ompx_dyn_cgroup_mem_messages.cpp b/clang/test/OpenMP/target_ompx_dyn_cgroup_mem_messages.cpp
new file mode 100644 (file)
index 0000000..05789f1
--- /dev/null
+++ b/clang/test/OpenMP/target_ompx_dyn_cgroup_mem_messages.cpp
@@ -0,0 +1,74 @@
+// RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 %s -Wuninitialized
+
+// RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 %s -Wuninitialized
+
+// RUN: %clang_cc1 -verify -fopenmp %s -Wuninitialized
+// RUN: %clang_cc1 -verify -fopenmp-simd %s -Wuninitialized
+
+void foo() {
+}
+
+bool foobool(int argc) {
+  return argc;
+}
+
+struct S1; // expected-note {{declared here}}
+
+template <class T, class S> // expected-note {{declared here}}
+int tmain(T argc, S **argv) {
+  T z;
+  #pragma omp target ompx_dyn_cgroup_mem // expected-error {{expected '(' after 'ompx_dyn_cgroup_mem'}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem () // expected-error {{expected expression}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc // expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc)) // expected-warning {{extra tokens at the end of '#pragma omp target' are ignored}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (foobool(argc)), ompx_dyn_cgroup_mem (true) // expected-error {{directive '#pragma omp target' cannot contain more than one 'ompx_dyn_cgroup_mem' clause}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (S) // expected-error {{'S' does not refer to a value}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem(argc+z)
+  foo();
+  return 0;
+}
+
+int main(int argc, char **argv) {
+int z;
+  #pragma omp target ompx_dyn_cgroup_mem // expected-error {{expected '(' after 'ompx_dyn_cgroup_mem'}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem ( // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem () // expected-error {{expected expression}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc // expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc)) // expected-warning {{extra tokens at the end of '#pragma omp target' are ignored}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc > 0 ? argv[1] : argv[2]) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (foobool(argc)), ompx_dyn_cgroup_mem (true) // expected-error {{directive '#pragma omp target' cannot contain more than one 'ompx_dyn_cgroup_mem' clause}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (S1) // expected-error {{'S1' does not refer to a value}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argv[1]=2) // expected-error {{expression must have integral or unscoped enumeration type, not 'char *'}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (argc argc) // expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem (1 0) // expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+  #pragma omp target ompx_dyn_cgroup_mem(ompx_dyn_cgroup_mem(tmain(argc, argv) // expected-error2 {{expected ')'}} expected-note2 {{to match this '('}} expected-note {{in instantiation of function template specialization 'tmain<int, char>' requested here}}
+  foo();
+
+  return tmain(argc, argv);
+}
+

diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
index 1b8f835..15652c4 100644 (file)
--- a/clang/tools/libclang/CIndex.cpp
+++ b/clang/tools/libclang/CIndex.cpp
@@ -2706,6 +2706,11 @@ void OMPClauseEnqueue::VisitOMPAffinityClause(const OMPAffinityClause *C) {
     Visitor->AddStmt(E);
 }
 void OMPClauseEnqueue::VisitOMPBindClause(const OMPBindClause *C) {}
+void OMPClauseEnqueue::VisitOMPXDynCGroupMemClause(
+    const OMPXDynCGroupMemClause *C) {
+  VisitOMPClauseWithPreInit(C);
+  Visitor->AddStmt(C->getSize());
+}
 
 } // namespace
 

diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index 0b87b6e..7b55acd 100644 (file)
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -436,6 +436,10 @@ def OMPC_Bind : Clause<"bind"> {
   let clangClass = "OMPBindClause";
 }
 
+def OMPC_OMPX_DynCGroupMem : Clause<"ompx_dyn_cgroup_mem"> {
+  let clangClass = "OMPXDynCGroupMemClause";
+}
+
 //===----------------------------------------------------------------------===//
 // Definition of OpenMP directives
 //===----------------------------------------------------------------------===//
@@ -637,7 +641,8 @@ def OMP_Target : Directive<"target"> {
     VersionedClause<OMPC_Device>,
     VersionedClause<OMPC_ThreadLimit, 51>,
     VersionedClause<OMPC_DefaultMap>,
-    VersionedClause<OMPC_NoWait>
+    VersionedClause<OMPC_NoWait>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 def OMP_Teams : Directive<"teams"> {
@@ -736,7 +741,8 @@ def OMP_TargetParallel : Directive<"target parallel"> {
     VersionedClause<OMPC_Device>,
     VersionedClause<OMPC_If>,
     VersionedClause<OMPC_NumThreads>,
-    VersionedClause<OMPC_ProcBind>
+    VersionedClause<OMPC_ProcBind>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 def OMP_TargetParallelFor : Directive<"target parallel for"> {
@@ -765,6 +771,9 @@ def OMP_TargetParallelFor : Directive<"target parallel for"> {
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
+  let allowedOnceClauses = [
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
+  ];
 }
 def OMP_TargetParallelDo : Directive<"target parallel do"> {
   let allowedClauses = [
@@ -1234,6 +1243,9 @@ def OMP_TargetParallelForSimd : Directive<"target parallel for simd"> {
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
+  let allowedOnceClauses = [
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
+  ];
 }
 def OMP_TargetParallelDoSimd : Directive<"target parallel do simd"> {
   let allowedClauses = [
@@ -1294,7 +1306,8 @@ def OMP_TargetSimd : Directive<"target simd"> {
     VersionedClause<OMPC_ProcBind>,
     VersionedClause<OMPC_Device>,
     VersionedClause<OMPC_DefaultMap>,
-    VersionedClause<OMPC_Schedule>
+    VersionedClause<OMPC_Schedule>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 def OMP_TeamsDistribute : Directive<"teams distribute"> {
@@ -1460,7 +1473,8 @@ def OMP_TargetTeams : Directive<"target teams"> {
     VersionedClause<OMPC_DefaultMap>,
     VersionedClause<OMPC_Default>,
     VersionedClause<OMPC_NumTeams>,
-    VersionedClause<OMPC_ThreadLimit>
+    VersionedClause<OMPC_ThreadLimit>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 def OMP_TargetTeamsDistribute : Directive<"target teams distribute"> {
@@ -1486,7 +1500,8 @@ def OMP_TargetTeamsDistribute : Directive<"target teams distribute"> {
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_ThreadLimit>,
     VersionedClause<OMPC_Collapse>,
-    VersionedClause<OMPC_DistSchedule>
+    VersionedClause<OMPC_DistSchedule>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 
@@ -1518,6 +1533,9 @@ def OMP_TargetTeamsDistributeParallelFor :
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
+  let allowedOnceClauses = [
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
+  ];
 }
 def OMP_TargetTeamsDistributeParallelDo :
     Directive<"target teams distribute parallel do"> {
@@ -1586,6 +1604,9 @@ def OMP_TargetTeamsDistributeParallelForSimd :
     VersionedClause<OMPC_Order, 50>,
     VersionedClause<OMPC_UsesAllocators, 50>
   ];
+  let allowedOnceClauses = [
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
+  ];
 }
 def OMP_TargetTeamsDistributeParallelDoSimd :
     Directive<"target teams distribute parallel do simd"> {
@@ -1655,6 +1676,7 @@ def OMP_TargetTeamsDistributeSimd :
     VersionedClause<OMPC_DistSchedule>,
     VersionedClause<OMPC_SafeLen>,
     VersionedClause<OMPC_SimdLen>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 def OMP_Allocate : Directive<"allocate"> {
@@ -2019,6 +2041,7 @@ def OMP_target_teams_loop : Directive<"target teams loop"> {
     VersionedClause<OMPC_NumTeams>,
     VersionedClause<OMPC_Order>,
     VersionedClause<OMPC_ThreadLimit>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 def OMP_parallel_loop : Directive<"parallel loop"> {
@@ -2067,6 +2090,7 @@ def OMP_target_parallel_loop : Directive<"target parallel loop"> {
     VersionedClause<OMPC_NumThreads>,
     VersionedClause<OMPC_Order>,
     VersionedClause<OMPC_ProcBind>,
+    VersionedClause<OMPC_OMPX_DynCGroupMem>,
   ];
 }
 def OMP_Metadirective : Directive<"metadirective"> {

diff --git a/openmp/docs/design/Runtimes.rst b/openmp/docs/design/Runtimes.rst
index 2db14aa..0242411 100644 (file)
--- a/openmp/docs/design/Runtimes.rst
+++ b/openmp/docs/design/Runtimes.rst
@@ -1200,8 +1200,9 @@ buffer. This pointer can be obtained using the
 ``llvm_omp_target_dynamic_shared_alloc`` extension. If this function is called
 from the host it will simply return a null pointer. In order to use this buffer
 the kernel must be launched with an adequate amount of dynamic shared memory
-allocated. Currently this is done using the ``LIBOMPTARGET_SHARED_MEMORY_SIZE``
-environment variable. An example is given below.
+allocated. This can be done using the ``LIBOMPTARGET_SHARED_MEMORY_SIZE``
+environment variable or the ``ompx_dyn_cgroup_mem(<N>)`` target directive
+clause. Examples for both are given below.
 
 .. code-block:: c++
 
@@ -1210,20 +1211,42 @@ environment variable. An example is given below.
     #pragma omp target parallel map(from : x)
       {
         int *buf = llvm_omp_target_dynamic_shared_alloc();
-    #pragma omp barrier
         if (omp_get_thread_num() == 0)
           *buf = 1;
     #pragma omp barrier
         if (omp_get_thread_num() == 1)
           x = *buf;
       }
+      assert(x == 1);
     }
 
 .. code-block:: console
 
-    $ clang++ -fopenmp -fopenmp-targets=nvptx64 shared.c
+    $ clang++ -fopenmp --offload-arch=sm_80 -O3 shared.c
     $ env LIBOMPTARGET_SHARED_MEMORY_SIZE=256 ./shared
 
+.. code-block:: c++
+
+    void foo(int N) {
+      int x;
+    #pragma omp target parallel map(from : x) ompx_dyn_cgroup_mem(N * sizeof(int))
+      {
+        int *buf = llvm_omp_target_dynamic_shared_alloc();
+        if (omp_get_thread_num() == 0)
+          buf[N - 1] = 1;
+    #pragma omp barrier
+        if (omp_get_thread_num() == 1)
+          x = buf[N - 1];
+      }
+      assert(x == 1);
+    }
+
+.. code-block:: console
+
+    $ clang++ -fopenmp --offload-arch=gfx90a -O3 shared.c
+    $ env LIBOMPTARGET_NEXTGEN_PLUGINS=1 ./shared
+
+
 .. _libomptarget_device_debugging:
 
 Debugging

diff --git a/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp b/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
index 22c28a9..457c1e5 100644 (file)
--- a/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
+++ b/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
@@ -28,6 +28,7 @@
 #include "PluginInterface.h"
 #include "Utilities.h"
 #include "UtilitiesRTL.h"
+#include "omptarget.h"
 
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/SmallVector.h"
@@ -407,10 +408,6 @@ struct AMDGPUKernelTy : public GenericKernelTy {
         return Err;
     }
 
-    // Account for user requested dynamic shared memory.
-    // TODO: This should be read from a per-kernel state flag.
-    GroupSize += Device.getDynamicMemorySize();
-
     // Make sure it is a kernel symbol.
     if (SymbolType != HSA_SYMBOL_KIND_KERNEL)
       return Plugin::error("Symbol %s is not a kernel function");
@@ -423,8 +420,8 @@ struct AMDGPUKernelTy : public GenericKernelTy {
 
   /// Launch the AMDGPU kernel function.
   Error launchImpl(GenericDeviceTy &GenericDevice, uint32_t NumThreads,
-                   uint64_t NumBlocks, uint32_t DynamicMemorySize,
-                   int32_t NumKernelArgs, void *KernelArgs,
+                   uint64_t NumBlocks, 
+                   KernelArgsTy &KernelArgs, void *Args,
                    AsyncInfoWrapperTy &AsyncInfoWrapper) const override;
 
   /// The default number of blocks is common to the whole device.
@@ -544,7 +541,7 @@ struct AMDGPUQueueTy {
   /// signal and can define an optional input signal (nullptr if none).
   Error pushKernelLaunch(const AMDGPUKernelTy &Kernel, void *KernelArgs,
                          uint32_t NumThreads, uint64_t NumBlocks,
-                         AMDGPUSignalTy *OutputSignal,
+                         uint32_t GroupSize, AMDGPUSignalTy *OutputSignal,
                          AMDGPUSignalTy *InputSignal) {
     assert(OutputSignal && "Invalid kernel output signal");
 
@@ -581,7 +578,7 @@ struct AMDGPUQueueTy {
     Packet->grid_size_y = 1;
     Packet->grid_size_z = 1;
     Packet->private_segment_size = Kernel.getPrivateSize();
-    Packet->group_segment_size = Kernel.getGroupSize();
+    Packet->group_segment_size = GroupSize;
     Packet->kernel_object = Kernel.getKernelObject();
     Packet->kernarg_address = KernelArgs;
     Packet->reserved2 = 0;
@@ -1006,6 +1003,7 @@ public:
   /// the kernel args buffer to the specified memory manager.
   Error pushKernelLaunch(const AMDGPUKernelTy &Kernel, void *KernelArgs,
                          uint32_t NumThreads, uint64_t NumBlocks,
+                         uint32_t GroupSize,
                          AMDGPUMemoryManagerTy &MemoryManager) {
     // Retrieve an available signal for the operation's output.
     AMDGPUSignalTy *OutputSignal = SignalManager.getResource();
@@ -1023,7 +1021,7 @@ public:
 
     // Push the kernel with the output signal and an input signal (optional)
     return Queue.pushKernelLaunch(Kernel, KernelArgs, NumThreads, NumBlocks,
-                                  OutputSignal, InputSignal);
+                                  GroupSize, OutputSignal, InputSignal);
   }
 
   /// Push an asynchronous memory copy between pinned memory buffers.
@@ -2456,10 +2454,9 @@ private:
 
 Error AMDGPUKernelTy::launchImpl(GenericDeviceTy &GenericDevice,
                                  uint32_t NumThreads, uint64_t NumBlocks,
-                                 uint32_t DynamicMemorySize,
-                                 int32_t NumKernelArgs, void *KernelArgs,
+                                 KernelArgsTy &KernelArgs, void *Args,
                                  AsyncInfoWrapperTy &AsyncInfoWrapper) const {
-  const uint32_t KernelArgsSize = NumKernelArgs * sizeof(void *);
+  const uint32_t KernelArgsSize = KernelArgs.NumArgs * sizeof(void *);
 
   if (ArgsSize < KernelArgsSize)
     return Plugin::error("Mismatch of kernel arguments size");
@@ -2477,6 +2474,13 @@ Error AMDGPUKernelTy::launchImpl(GenericDeviceTy &GenericDevice,
   if (auto Err = ArgsMemoryManager.allocate(AllArgsSize, &AllArgs))
     return Err;
 
+  // Account for user requested dynamic shared memory.
+  uint32_t GroupSize = getGroupSize();
+  if (uint32_t MaxDynCGroupMem = std::max(
+          KernelArgs.DynCGroupMem, GenericDevice.getDynamicMemorySize())) {
+    GroupSize += MaxDynCGroupMem;
+  }
+
   // Initialize implicit arguments.
   utils::AMDGPUImplicitArgsTy *ImplArgs =
       reinterpret_cast<utils::AMDGPUImplicitArgsTy *>(
@@ -2488,16 +2492,16 @@ Error AMDGPUKernelTy::launchImpl(GenericDeviceTy &GenericDevice,
   // Copy the explicit arguments.
   // TODO: We should expose the args memory manager alloc to the common part as
   //      alternative to copying them twice.
-  if (NumKernelArgs)
-    std::memcpy(AllArgs, *static_cast<void **>(KernelArgs),
-                sizeof(void *) * NumKernelArgs);
+  if (KernelArgs.NumArgs)
+    std::memcpy(AllArgs, *static_cast<void **>(Args),
+                sizeof(void *) * KernelArgs.NumArgs);
 
   AMDGPUDeviceTy &AMDGPUDevice = static_cast<AMDGPUDeviceTy &>(GenericDevice);
   AMDGPUStreamTy &Stream = AMDGPUDevice.getStream(AsyncInfoWrapper);
 
   // Push the kernel launch into the stream.
   return Stream.pushKernelLaunch(*this, AllArgs, NumThreads, NumBlocks,
-                                 ArgsMemoryManager);
+                                 GroupSize, ArgsMemoryManager);
 }
 
 GenericPluginTy *Plugin::createPlugin() { return new AMDGPUPluginTy(); }

diff --git a/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.cpp b/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.cpp
index 9787be4..ceece60 100644 (file)
--- a/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.cpp
+++ b/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.cpp
@@ -215,11 +215,6 @@ Error GenericKernelTy::init(GenericDeviceTy &GenericDevice,
 
   MaxNumThreads = GenericDevice.getThreadLimit();
 
-  DynamicMemorySize = GenericDevice.getDynamicMemorySize();
-
-  if (RecordReplay.isRecording())
-    RecordReplay.saveImage(Name, Image);
-
   return initImpl(GenericDevice, Image);
 }
 
@@ -242,8 +237,8 @@ Error GenericKernelTy::launch(GenericDeviceTy &GenericDevice, void **ArgPtrs,
        " blocks and %d threads in %s mode\n",
        getName(), NumBlocks, NumThreads, getExecutionModeName());
 
-  return launchImpl(GenericDevice, NumThreads, NumBlocks, DynamicMemorySize,
-                    KernelArgs.NumArgs, KernelArgsPtr, AsyncInfoWrapper);
+  return launchImpl(GenericDevice, NumThreads, NumBlocks, KernelArgs,
+                    KernelArgsPtr, AsyncInfoWrapper);
 }
 
 void *GenericKernelTy::prepareArgs(GenericDeviceTy &GenericDevice,

diff --git a/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.h b/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.h
index 1cee343..321b9bc 100644 (file)
--- a/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.h
+++ b/openmp/libomptarget/plugins-nextgen/common/PluginInterface/PluginInterface.h
@@ -162,7 +162,7 @@ public:
 struct GenericKernelTy {
   /// Construct a kernel with a name and a execution mode.
   GenericKernelTy(const char *Name, OMPTgtExecModeFlags ExecutionMode)
-      : Name(Name), ExecutionMode(ExecutionMode), DynamicMemorySize(0),
+      : Name(Name), ExecutionMode(ExecutionMode),
         PreferredNumThreads(0), MaxNumThreads(0) {}
 
   virtual ~GenericKernelTy() {}
@@ -178,8 +178,8 @@ struct GenericKernelTy {
                ptrdiff_t *ArgOffsets, KernelArgsTy &KernelArgs,
                AsyncInfoWrapperTy &AsyncInfoWrapper) const;
   virtual Error launchImpl(GenericDeviceTy &GenericDevice, uint32_t NumThreads,
-                           uint64_t NumBlocks, uint32_t DynamicMemorySize,
-                           int32_t NumKernelArgs, void *KernelArgs,
+                           uint64_t NumBlocks,
+                           KernelArgsTy &KernelArgs, void *Args,
                            AsyncInfoWrapperTy &AsyncInfoWrapper) const = 0;
 
   /// Get the kernel name.
@@ -245,9 +245,6 @@ private:
   OMPTgtExecModeFlags ExecutionMode;
 
 protected:
-  /// The dynamic memory size reserved for executing the kernel.
-  uint32_t DynamicMemorySize;
-
   /// The preferred number of threads to run the kernel.
   uint32_t PreferredNumThreads;
 

diff --git a/openmp/libomptarget/plugins-nextgen/cuda/src/rtl.cpp b/openmp/libomptarget/plugins-nextgen/cuda/src/rtl.cpp
index 6f93506..8a6479e 100644 (file)
--- a/openmp/libomptarget/plugins-nextgen/cuda/src/rtl.cpp
+++ b/openmp/libomptarget/plugins-nextgen/cuda/src/rtl.cpp
@@ -62,8 +62,7 @@ struct CUDAKernelTy : public GenericKernelTy {
 
   /// Launch the CUDA kernel function
   Error launchImpl(GenericDeviceTy &GenericDevice, uint32_t NumThreads,
-                   uint64_t NumBlocks, uint32_t DynamicMemorySize,
-                   int32_t NumKernelArgs, void *KernelArgs,
+                   uint64_t NumBlocks, KernelArgsTy &KernelArgs, void *Args,
                    AsyncInfoWrapperTy &AsyncInfoWrapper) const override;
 
   /// The default number of blocks is common to the whole device.
@@ -818,8 +817,7 @@ private:
 
 Error CUDAKernelTy::launchImpl(GenericDeviceTy &GenericDevice,
                                uint32_t NumThreads, uint64_t NumBlocks,
-                               uint32_t DynamicMemorySize,
-                               int32_t NumKernelArgs, void *KernelArgs,
+                               KernelArgsTy &KernelArgs, void *Args,
                                AsyncInfoWrapperTy &AsyncInfoWrapper) const {
   CUDADeviceTy &CUDADevice = static_cast<CUDADeviceTy &>(GenericDevice);
 
@@ -827,11 +825,14 @@ Error CUDAKernelTy::launchImpl(GenericDeviceTy &GenericDevice,
   if (!Stream)
     return Plugin::error("Failure to get stream");
 
+  uint32_t MaxDynCGroupMem =
+      std::max(KernelArgs.DynCGroupMem, GenericDevice.getDynamicMemorySize());
+
   CUresult Res =
       cuLaunchKernel(Func, NumBlocks, /* gridDimY */ 1,
                      /* gridDimZ */ 1, NumThreads,
-                     /* blockDimY */ 1, /* blockDimZ */ 1, DynamicMemorySize,
-                     Stream, (void **)KernelArgs, nullptr);
+                     /* blockDimY */ 1, /* blockDimZ */ 1, MaxDynCGroupMem,
+                     Stream, (void **)Args, nullptr);
   return Plugin::check(Res, "Error in cuLaunchKernel for '%s': %s", getName());
 }
 

diff --git a/openmp/libomptarget/plugins-nextgen/generic-elf-64bit/src/rtl.cpp b/openmp/libomptarget/plugins-nextgen/generic-elf-64bit/src/rtl.cpp
index 43bf594..cd17161 100644 (file)
--- a/openmp/libomptarget/plugins-nextgen/generic-elf-64bit/src/rtl.cpp
+++ b/openmp/libomptarget/plugins-nextgen/generic-elf-64bit/src/rtl.cpp
@@ -20,6 +20,7 @@
 #include "DeviceEnvironment.h"
 #include "GlobalHandler.h"
 #include "PluginInterface.h"
+#include "omptarget.h"
 
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Frontend/OpenMP/OMPConstants.h"
@@ -63,23 +64,22 @@ struct GenELF64KernelTy : public GenericKernelTy {
 
   /// Launch the kernel using the libffi.
   Error launchImpl(GenericDeviceTy &GenericDevice, uint32_t NumThreads,
-                   uint64_t NumBlocks, uint32_t DynamicMemorySize,
-                   int32_t NumKernelArgs, void *KernelArgs,
+                   uint64_t NumBlocks, KernelArgsTy &KernelArgs, void *Args,
                    AsyncInfoWrapperTy &AsyncInfoWrapper) const override {
     // Create a vector of ffi_types, one per argument.
-    SmallVector<ffi_type *, 16> ArgTypes(NumKernelArgs, &ffi_type_pointer);
+    SmallVector<ffi_type *, 16> ArgTypes(KernelArgs.NumArgs, &ffi_type_pointer);
     ffi_type **ArgTypesPtr = (ArgTypes.size()) ? &ArgTypes[0] : nullptr;
 
     // Prepare the cif structure before running the kernel function.
     ffi_cif Cif;
-    ffi_status Status = ffi_prep_cif(&Cif, FFI_DEFAULT_ABI, NumKernelArgs,
+    ffi_status Status = ffi_prep_cif(&Cif, FFI_DEFAULT_ABI, KernelArgs.NumArgs,
                                      &ffi_type_void, ArgTypesPtr);
     if (Status != FFI_OK)
       return Plugin::error("Error in ffi_prep_cif: %d", Status);
 
     // Call the kernel function through libffi.
     long Return;
-    ffi_call(&Cif, Func, &Return, (void **)KernelArgs);
+    ffi_call(&Cif, Func, &Return, (void **)Args);
 
     return Plugin::success();
   }

diff --git a/openmp/libomptarget/test/api/omp_dynamic_shared_memory_mixed.c b/openmp/libomptarget/test/api/omp_dynamic_shared_memory_mixed.c
new file mode 100644 (file)
index 0000000..0168361
--- /dev/null
+++ b/openmp/libomptarget/test/api/omp_dynamic_shared_memory_mixed.c
@@ -0,0 +1,51 @@
+// RUN: env LIBOMPTARGET_NEXTGEN_PLUGINS=1 \
+// RUN: %libomptarget-compileopt-run-and-check-generic
+
+#include <omp.h>
+#include <stdio.h>
+
+#define N 512
+
+int main() {
+  int Result[N], NumThreads;
+
+#pragma omp target teams num_teams(1) thread_limit(N)                          \
+                         ompx_dyn_cgroup_mem(N * sizeof(Result[0]))            \
+                         map(from : Result, NumThreads)
+  {
+    int Buffer[N];
+#pragma omp parallel
+    {
+      int *DynBuffer = (int *)llvm_omp_target_dynamic_shared_alloc();
+      int TId = omp_get_thread_num();
+      if (TId == 0)
+        NumThreads = omp_get_num_threads();
+      Buffer[TId] = 7;
+      DynBuffer[TId] = 3;
+#pragma omp barrier
+      int WrappedTId = (TId + 37) % NumThreads;
+      Result[TId] = Buffer[WrappedTId] + DynBuffer[WrappedTId];
+    }
+  }
+
+  if (llvm_omp_target_dynamic_shared_alloc())
+    return -1;
+
+  if (NumThreads < N / 2 || NumThreads > N) {
+    printf("Expected number of threads to be in [%i:%i], but got: %i", N / 2, N,
+           NumThreads);
+    return -1;
+  }
+
+  int Failed = 0;
+  for (int i = 0; i < NumThreads; ++i) {
+    if (Result[i] != 7 + 3) {
+      printf("Result[%i] is %i, expected %i\n", i, Result[i], 7 + 3);
+      ++Failed;
+    }
+  }
+
+  if (!Failed)
+    printf("PASS\n");
+  // CHECK: PASS
+}