[llvm-branch-commits] [clang] [llvm] [openmp] [Clang][OpenMP] Add interchange directive (PR #93022)

Michael Kruse via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Wed May 22 04:43:25 PDT 2024


https://github.com/Meinersbur created https://github.com/llvm/llvm-project/pull/93022

Add the interchange directive which will be introduced in the upcoming OpenMP 6.0 specification. A preview has been published in [Technical Report 12](https://www.openmp.org/wp-content/uploads/openmp-TR12.pdf).

>From 76634ade083cde2881bb072fe821f1d434f99196 Mon Sep 17 00:00:00 2001
From: Michael Kruse <llvm-project at meinersbur.de>
Date: Wed, 22 May 2024 13:22:05 +0200
Subject: [PATCH] [Clang][OpenMP] Add interchange directive

---
 clang/include/clang-c/Index.h                 |    4 +
 clang/include/clang/AST/RecursiveASTVisitor.h |    3 +
 clang/include/clang/AST/StmtOpenMP.h          |   76 +-
 clang/include/clang/Basic/StmtNodes.td        |    1 +
 clang/include/clang/Sema/SemaOpenMP.h         |    6 +
 .../include/clang/Serialization/ASTBitCodes.h |    1 +
 clang/lib/AST/StmtOpenMP.cpp                  |   20 +
 clang/lib/AST/StmtPrinter.cpp                 |    5 +
 clang/lib/AST/StmtProfile.cpp                 |    5 +
 clang/lib/Basic/OpenMPKinds.cpp               |    4 +-
 clang/lib/CodeGen/CGStmt.cpp                  |    3 +
 clang/lib/CodeGen/CGStmtOpenMP.cpp            |   10 +
 clang/lib/CodeGen/CodeGenFunction.h           |    1 +
 clang/lib/Parse/ParseOpenMP.cpp               |    2 +
 clang/lib/Sema/SemaExceptionSpec.cpp          |    1 +
 clang/lib/Sema/SemaOpenMP.cpp                 |  162 ++
 clang/lib/Sema/TreeTransform.h                |   11 +
 clang/lib/Serialization/ASTReaderStmt.cpp     |   11 +
 clang/lib/Serialization/ASTWriterStmt.cpp     |    5 +
 clang/test/OpenMP/interchange_ast_print.cpp   |  135 ++
 clang/test/OpenMP/interchange_codegen.cpp     | 1990 +++++++++++++++++
 clang/test/OpenMP/interchange_messages.cpp    |   77 +
 clang/tools/libclang/CIndex.cpp               |    8 +
 clang/tools/libclang/CXCursor.cpp             |    3 +
 llvm/include/llvm/Frontend/OpenMP/OMP.td      |    3 +
 .../test/transform/interchange/foreach.cpp    |  216 ++
 .../test/transform/interchange/intfor.c       |   38 +
 .../test/transform/interchange/iterfor.cpp    |  222 ++
 .../parallel-wsloop-collapse-foreach.cpp      |  340 +++
 .../parallel-wsloop-collapse-intfor.cpp       |  106 +
 30 files changed, 3467 insertions(+), 2 deletions(-)
 create mode 100644 clang/test/OpenMP/interchange_ast_print.cpp
 create mode 100644 clang/test/OpenMP/interchange_codegen.cpp
 create mode 100644 clang/test/OpenMP/interchange_messages.cpp
 create mode 100644 openmp/runtime/test/transform/interchange/foreach.cpp
 create mode 100644 openmp/runtime/test/transform/interchange/intfor.c
 create mode 100644 openmp/runtime/test/transform/interchange/iterfor.cpp
 create mode 100644 openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-foreach.cpp
 create mode 100644 openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-intfor.cpp

diff --git a/clang/include/clang-c/Index.h b/clang/include/clang-c/Index.h
index c7d63818ece23..a79aafbf20222 100644
--- a/clang/include/clang-c/Index.h
+++ b/clang/include/clang-c/Index.h
@@ -2150,6 +2150,10 @@ enum CXCursorKind {
    */
   CXCursor_OMPReverseDirective = 307,
 
+  /** OpenMP interchange directive.
+   */
+  CXCursor_OMPInterchangeDirective = 308,
+
   /** OpenACC Compute Construct.
    */
   CXCursor_OpenACCComputeConstruct = 320,
diff --git a/clang/include/clang/AST/RecursiveASTVisitor.h b/clang/include/clang/AST/RecursiveASTVisitor.h
index 06b29d59785f6..1bb167d7ddc3c 100644
--- a/clang/include/clang/AST/RecursiveASTVisitor.h
+++ b/clang/include/clang/AST/RecursiveASTVisitor.h
@@ -3024,6 +3024,9 @@ DEF_TRAVERSE_STMT(OMPUnrollDirective,
 DEF_TRAVERSE_STMT(OMPReverseDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })
 
+DEF_TRAVERSE_STMT(OMPInterchangeDirective,
+                  { TRY_TO(TraverseOMPExecutableDirective(S)); })
+
 DEF_TRAVERSE_STMT(OMPForDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })
 
diff --git a/clang/include/clang/AST/StmtOpenMP.h b/clang/include/clang/AST/StmtOpenMP.h
index fb7f413162fad..01c8b8e1a9f5e 100644
--- a/clang/include/clang/AST/StmtOpenMP.h
+++ b/clang/include/clang/AST/StmtOpenMP.h
@@ -1009,7 +1009,7 @@ class OMPLoopTransformationDirective : public OMPLoopBasedDirective {
   static bool classof(const Stmt *T) {
     Stmt::StmtClass C = T->getStmtClass();
     return C == OMPTileDirectiveClass || C == OMPUnrollDirectiveClass ||
-           C == OMPReverseDirectiveClass;
+           C == OMPReverseDirectiveClass || C == OMPInterchangeDirectiveClass;
   }
 };
 
@@ -5777,6 +5777,80 @@ class OMPReverseDirective final : public OMPLoopTransformationDirective {
   }
 };
 
+/// Represents the '#pragma omp interchange' loop transformation directive.
+///
+/// \code{c}
+///   #pragma omp interchange
+///   for (int i = 0; i < m; ++i)
+///     for (int j = 0; j < n; ++j)
+///       ..
+/// \endcode
+class OMPInterchangeDirective final : public OMPLoopTransformationDirective {
+  friend class ASTStmtReader;
+  friend class OMPExecutableDirective;
+
+  /// Offsets of child members.
+  enum {
+    PreInitsOffset = 0,
+    TransformedStmtOffset,
+  };
+
+  explicit OMPInterchangeDirective(SourceLocation StartLoc,
+                                   SourceLocation EndLoc, unsigned NumLoops)
+      : OMPLoopTransformationDirective(OMPInterchangeDirectiveClass,
+                                       llvm::omp::OMPD_interchange, StartLoc,
+                                       EndLoc, NumLoops) {
+    setNumGeneratedLoops(3 * NumLoops);
+  }
+
+  void setPreInits(Stmt *PreInits) {
+    Data->getChildren()[PreInitsOffset] = PreInits;
+  }
+
+  void setTransformedStmt(Stmt *S) {
+    Data->getChildren()[TransformedStmtOffset] = S;
+  }
+
+public:
+  /// Create a new AST node representation for '#pragma omp interchange'.
+  ///
+  /// \param C         Context of the AST.
+  /// \param StartLoc  Location of the introducer (e.g. the 'omp' token).
+  /// \param EndLoc    Location of the directive's end (e.g. the tok::eod).
+  /// \param Clauses   The directive's clauses.
+  /// \param NumLoops  Number of affected loops
+  ///                  (number of items in the 'permutation' clause if present).
+  /// \param AssociatedStmt  The outermost associated loop.
+  /// \param TransformedStmt The loop nest after tiling, or nullptr in
+  ///                        dependent contexts.
+  /// \param PreInits  Helper preinits statements for the loop nest.
+  static OMPInterchangeDirective *
+  Create(const ASTContext &C, SourceLocation StartLoc, SourceLocation EndLoc,
+         ArrayRef<OMPClause *> Clauses, unsigned NumLoops, Stmt *AssociatedStmt,
+         Stmt *TransformedStmt, Stmt *PreInits);
+
+  /// Build an empty '#pragma omp interchange' AST node for deserialization.
+  ///
+  /// \param C          Context of the AST.
+  /// \param NumClauses Number of clauses to allocate.
+  /// \param NumLoops   Number of associated loops to allocate.
+  static OMPInterchangeDirective *
+  CreateEmpty(const ASTContext &C, unsigned NumClauses, unsigned NumLoops);
+
+  /// Gets the associated loops after the transformation. This is the de-sugared
+  /// replacement or nullptr in dependent contexts.
+  Stmt *getTransformedStmt() const {
+    return Data->getChildren()[TransformedStmtOffset];
+  }
+
+  /// Return preinits statement.
+  Stmt *getPreInits() const { return Data->getChildren()[PreInitsOffset]; }
+
+  static bool classof(const Stmt *T) {
+    return T->getStmtClass() == OMPInterchangeDirectiveClass;
+  }
+};
+
 /// This represents '#pragma omp scan' directive.
 ///
 /// \code
diff --git a/clang/include/clang/Basic/StmtNodes.td b/clang/include/clang/Basic/StmtNodes.td
index b2e2be5c998bb..b445ea225eac5 100644
--- a/clang/include/clang/Basic/StmtNodes.td
+++ b/clang/include/clang/Basic/StmtNodes.td
@@ -230,6 +230,7 @@ def OMPLoopTransformationDirective : StmtNode<OMPLoopBasedDirective, 1>;
 def OMPTileDirective : StmtNode<OMPLoopTransformationDirective>;
 def OMPUnrollDirective : StmtNode<OMPLoopTransformationDirective>;
 def OMPReverseDirective : StmtNode<OMPLoopTransformationDirective>;
+def OMPInterchangeDirective : StmtNode<OMPLoopTransformationDirective>;
 def OMPForDirective : StmtNode<OMPLoopDirective>;
 def OMPForSimdDirective : StmtNode<OMPLoopDirective>;
 def OMPSectionsDirective : StmtNode<OMPExecutableDirective>;
diff --git a/clang/include/clang/Sema/SemaOpenMP.h b/clang/include/clang/Sema/SemaOpenMP.h
index ca91bffe24f6f..06376f173e8df 100644
--- a/clang/include/clang/Sema/SemaOpenMP.h
+++ b/clang/include/clang/Sema/SemaOpenMP.h
@@ -425,6 +425,12 @@ class SemaOpenMP : public SemaBase {
   /// Called on well-formed '#pragma omp reverse'.
   StmtResult ActOnOpenMPReverseDirective(Stmt *AStmt, SourceLocation StartLoc,
                                          SourceLocation EndLoc);
+  /// Called on well-formed '#pragma omp interchange' after parsing of its
+  /// clauses and the associated statement.
+  StmtResult ActOnOpenMPInterchangeDirective(ArrayRef<OMPClause *> Clauses,
+                                             Stmt *AStmt,
+                                             SourceLocation StartLoc,
+                                             SourceLocation EndLoc);
   /// Called on well-formed '\#pragma omp for' after parsing
   /// of the associated statement.
   StmtResult
diff --git a/clang/include/clang/Serialization/ASTBitCodes.h b/clang/include/clang/Serialization/ASTBitCodes.h
index dee0d073557cc..5fbdfd7a496fe 100644
--- a/clang/include/clang/Serialization/ASTBitCodes.h
+++ b/clang/include/clang/Serialization/ASTBitCodes.h
@@ -1857,6 +1857,7 @@ enum StmtCode {
   STMT_OMP_TILE_DIRECTIVE,
   STMT_OMP_UNROLL_DIRECTIVE,
   STMT_OMP_REVERSE_DIRECTIVE,
+  STMT_OMP_INTERCHANGE_DIRECTIVE,
   STMT_OMP_FOR_DIRECTIVE,
   STMT_OMP_FOR_SIMD_DIRECTIVE,
   STMT_OMP_SECTIONS_DIRECTIVE,
diff --git a/clang/lib/AST/StmtOpenMP.cpp b/clang/lib/AST/StmtOpenMP.cpp
index 83b8a08e9af73..24d8eb25c59ba 100644
--- a/clang/lib/AST/StmtOpenMP.cpp
+++ b/clang/lib/AST/StmtOpenMP.cpp
@@ -467,6 +467,26 @@ OMPReverseDirective *OMPReverseDirective::CreateEmpty(const ASTContext &C,
       SourceLocation(), SourceLocation());
 }
 
+OMPInterchangeDirective *OMPInterchangeDirective::Create(
+    const ASTContext &C, SourceLocation StartLoc, SourceLocation EndLoc,
+    ArrayRef<OMPClause *> Clauses, unsigned NumLoops, Stmt *AssociatedStmt,
+    Stmt *TransformedStmt, Stmt *PreInits) {
+  OMPInterchangeDirective *Dir = createDirective<OMPInterchangeDirective>(
+      C, Clauses, AssociatedStmt, TransformedStmtOffset + 1, StartLoc, EndLoc,
+      NumLoops);
+  Dir->setTransformedStmt(TransformedStmt);
+  Dir->setPreInits(PreInits);
+  return Dir;
+}
+
+OMPInterchangeDirective *
+OMPInterchangeDirective::CreateEmpty(const ASTContext &C, unsigned NumClauses,
+                                     unsigned NumLoops) {
+  return createEmptyDirective<OMPInterchangeDirective>(
+      C, NumClauses, /*HasAssociatedStmt=*/true, TransformedStmtOffset + 1,
+      SourceLocation(), SourceLocation(), NumLoops);
+}
+
 OMPForSimdDirective *
 OMPForSimdDirective::Create(const ASTContext &C, SourceLocation StartLoc,
                             SourceLocation EndLoc, unsigned CollapsedNum,
diff --git a/clang/lib/AST/StmtPrinter.cpp b/clang/lib/AST/StmtPrinter.cpp
index 64b481f680311..64bee75b205ae 100644
--- a/clang/lib/AST/StmtPrinter.cpp
+++ b/clang/lib/AST/StmtPrinter.cpp
@@ -768,6 +768,11 @@ void StmtPrinter::VisitOMPReverseDirective(OMPReverseDirective *Node) {
   PrintOMPExecutableDirective(Node);
 }
 
+void StmtPrinter::VisitOMPInterchangeDirective(OMPInterchangeDirective *Node) {
+  Indent() << "#pragma omp interchange";
+  PrintOMPExecutableDirective(Node);
+}
+
 void StmtPrinter::VisitOMPForDirective(OMPForDirective *Node) {
   Indent() << "#pragma omp for";
   PrintOMPExecutableDirective(Node);
diff --git a/clang/lib/AST/StmtProfile.cpp b/clang/lib/AST/StmtProfile.cpp
index 7445e5519b972..1ae99d43575a7 100644
--- a/clang/lib/AST/StmtProfile.cpp
+++ b/clang/lib/AST/StmtProfile.cpp
@@ -989,6 +989,11 @@ void StmtProfiler::VisitOMPReverseDirective(const OMPReverseDirective *S) {
   VisitOMPLoopTransformationDirective(S);
 }
 
+void StmtProfiler::VisitOMPInterchangeDirective(
+    const OMPInterchangeDirective *S) {
+  VisitOMPLoopTransformationDirective(S);
+}
+
 void StmtProfiler::VisitOMPForDirective(const OMPForDirective *S) {
   VisitOMPLoopDirective(S);
 }
diff --git a/clang/lib/Basic/OpenMPKinds.cpp b/clang/lib/Basic/OpenMPKinds.cpp
index 803808c38e2fe..ff5d5c8bdc981 100644
--- a/clang/lib/Basic/OpenMPKinds.cpp
+++ b/clang/lib/Basic/OpenMPKinds.cpp
@@ -684,7 +684,8 @@ bool clang::isOpenMPLoopBoundSharingDirective(OpenMPDirectiveKind Kind) {
 }
 
 bool clang::isOpenMPLoopTransformationDirective(OpenMPDirectiveKind DKind) {
-  return DKind == OMPD_tile || DKind == OMPD_unroll || DKind == OMPD_reverse;
+  return DKind == OMPD_tile || DKind == OMPD_unroll || DKind == OMPD_reverse ||
+         DKind == OMPD_interchange;
 }
 
 bool clang::isOpenMPCombinedParallelADirective(OpenMPDirectiveKind DKind) {
@@ -809,6 +810,7 @@ void clang::getOpenMPCaptureRegions(
   case OMPD_tile:
   case OMPD_unroll:
   case OMPD_reverse:
+  case OMPD_interchange:
     // loop transformations do not introduce captures.
     break;
   case OMPD_threadprivate:
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index 93c2f8900dd12..ba7c52cc6ab7b 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -225,6 +225,9 @@ void CodeGenFunction::EmitStmt(const Stmt *S, ArrayRef<const Attr *> Attrs) {
   case Stmt::OMPReverseDirectiveClass:
     EmitOMPReverseDirective(cast<OMPReverseDirective>(*S));
     break;
+  case Stmt::OMPInterchangeDirectiveClass:
+    EmitOMPInterchangeDirective(cast<OMPInterchangeDirective>(*S));
+    break;
   case Stmt::OMPForDirectiveClass:
     EmitOMPForDirective(cast<OMPForDirective>(*S));
     break;
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index ad6c044aa483b..7a37e452fb559 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -189,6 +189,9 @@ class OMPLoopScope : public CodeGenFunction::RunCleanupsScope {
       PreInits = Unroll->getPreInits();
     } else if (const auto *Reverse = dyn_cast<OMPReverseDirective>(&S)) {
       PreInits = Reverse->getPreInits();
+    } else if (const auto *Interchange =
+                   dyn_cast<OMPInterchangeDirective>(&S)) {
+      PreInits = Interchange->getPreInits();
     } else {
       llvm_unreachable("Unknown loop-based directive kind.");
     }
@@ -2770,6 +2773,13 @@ void CodeGenFunction::EmitOMPReverseDirective(const OMPReverseDirective &S) {
   EmitStmt(S.getTransformedStmt());
 }
 
+void CodeGenFunction::EmitOMPInterchangeDirective(
+    const OMPInterchangeDirective &S) {
+  // Emit the de-sugared statement.
+  OMPTransformDirectiveScopeRAII InterchangeScope(*this, &S);
+  EmitStmt(S.getTransformedStmt());
+}
+
 void CodeGenFunction::EmitOMPUnrollDirective(const OMPUnrollDirective &S) {
   bool UseOMPIRBuilder = CGM.getLangOpts().OpenMPIRBuilder;
 
diff --git a/clang/lib/CodeGen/CodeGenFunction.h b/clang/lib/CodeGen/CodeGenFunction.h
index ac738e1e82886..c2a8e65ca2d0a 100644
--- a/clang/lib/CodeGen/CodeGenFunction.h
+++ b/clang/lib/CodeGen/CodeGenFunction.h
@@ -3808,6 +3808,7 @@ class CodeGenFunction : public CodeGenTypeCache {
   void EmitOMPTileDirective(const OMPTileDirective &S);
   void EmitOMPUnrollDirective(const OMPUnrollDirective &S);
   void EmitOMPReverseDirective(const OMPReverseDirective &S);
+  void EmitOMPInterchangeDirective(const OMPInterchangeDirective &S);
   void EmitOMPForDirective(const OMPForDirective &S);
   void EmitOMPForSimdDirective(const OMPForSimdDirective &S);
   void EmitOMPSectionsDirective(const OMPSectionsDirective &S);
diff --git a/clang/lib/Parse/ParseOpenMP.cpp b/clang/lib/Parse/ParseOpenMP.cpp
index 57fcf6ce520ac..0e3e604203c86 100644
--- a/clang/lib/Parse/ParseOpenMP.cpp
+++ b/clang/lib/Parse/ParseOpenMP.cpp
@@ -2385,6 +2385,7 @@ Parser::DeclGroupPtrTy Parser::ParseOpenMPDeclarativeDirectiveWithExtDecl(
   case OMPD_tile:
   case OMPD_unroll:
   case OMPD_reverse:
+  case OMPD_interchange:
   case OMPD_task:
   case OMPD_taskyield:
   case OMPD_barrier:
@@ -2804,6 +2805,7 @@ StmtResult Parser::ParseOpenMPDeclarativeOrExecutableDirective(
   case OMPD_tile:
   case OMPD_unroll:
   case OMPD_reverse:
+  case OMPD_interchange:
   case OMPD_for:
   case OMPD_for_simd:
   case OMPD_sections:
diff --git a/clang/lib/Sema/SemaExceptionSpec.cpp b/clang/lib/Sema/SemaExceptionSpec.cpp
index 4de7183cde281..5991f496d3a0f 100644
--- a/clang/lib/Sema/SemaExceptionSpec.cpp
+++ b/clang/lib/Sema/SemaExceptionSpec.cpp
@@ -1487,6 +1487,7 @@ CanThrowResult Sema::canThrow(const Stmt *S) {
   case Stmt::OMPTileDirectiveClass:
   case Stmt::OMPUnrollDirectiveClass:
   case Stmt::OMPReverseDirectiveClass:
+  case Stmt::OMPInterchangeDirectiveClass:
   case Stmt::OMPSingleDirectiveClass:
   case Stmt::OMPTargetDataDirectiveClass:
   case Stmt::OMPTargetDirectiveClass:
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index c2fd4de933ae4..ef141003b7d61 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -39,6 +39,7 @@
 #include "llvm/ADT/IndexedMap.h"
 #include "llvm/ADT/PointerEmbeddedInt.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/Sequence.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/Frontend/OpenMP/OMPAssume.h"
@@ -4335,6 +4336,7 @@ void SemaOpenMP::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind,
   case OMPD_tile:
   case OMPD_unroll:
   case OMPD_reverse:
+  case OMPD_interchange:
     break;
   case OMPD_loop:
     // TODO: 'loop' may require additional parameters depending on the binding.
@@ -6552,6 +6554,10 @@ StmtResult SemaOpenMP::ActOnOpenMPExecutableDirective(
            "reverse directive does not support any clauses");
     Res = ActOnOpenMPReverseDirective(AStmt, StartLoc, EndLoc);
     break;
+  case OMPD_interchange:
+    Res = ActOnOpenMPInterchangeDirective(ClausesWithImplicit, AStmt, StartLoc,
+                                          EndLoc);
+    break;
   case OMPD_for:
     Res = ActOnOpenMPForDirective(ClausesWithImplicit, AStmt, StartLoc, EndLoc,
                                   VarsWithInheritedDSA);
@@ -15139,6 +15145,8 @@ bool SemaOpenMP::checkTransformableLoopNest(
           DependentPreInits = Dir->getPreInits();
         else if (auto *Dir = dyn_cast<OMPReverseDirective>(Transform))
           DependentPreInits = Dir->getPreInits();
+        else if (auto *Dir = dyn_cast<OMPInterchangeDirective>(Transform))
+          DependentPreInits = Dir->getPreInits();
         else
           llvm_unreachable("Unhandled loop transformation");
 
@@ -15937,6 +15945,160 @@ StmtResult SemaOpenMP::ActOnOpenMPReverseDirective(Stmt *AStmt,
                                      buildPreInits(Context, PreInits));
 }
 
+StmtResult SemaOpenMP::ActOnOpenMPInterchangeDirective(
+    ArrayRef<OMPClause *> Clauses, Stmt *AStmt, SourceLocation StartLoc,
+    SourceLocation EndLoc) {
+  ASTContext &Context = getASTContext();
+  DeclContext *CurContext = SemaRef.CurContext;
+  Scope *CurScope = SemaRef.getCurScope();
+
+  // Empty statement should only be possible if there already was an error.
+  if (!AStmt)
+    return StmtError();
+
+  // interchange without permutation clause swaps two loops.
+  constexpr size_t NumLoops = 2;
+
+  // Verify and diagnose loop nest.
+  SmallVector<OMPLoopBasedDirective::HelperExprs, 4> LoopHelpers(NumLoops);
+  Stmt *Body = nullptr;
+  SmallVector<SmallVector<Stmt *, 0>, 2> OriginalInits;
+  if (!checkTransformableLoopNest(OMPD_interchange, AStmt, NumLoops,
+                                  LoopHelpers, Body, OriginalInits))
+    return StmtError();
+
+  // Delay interchange to when template is completely instantiated.
+  if (CurContext->isDependentContext())
+    return OMPInterchangeDirective::Create(Context, StartLoc, EndLoc, Clauses,
+                                           NumLoops, AStmt, nullptr, nullptr);
+
+  assert(LoopHelpers.size() == NumLoops &&
+         "Expecting loop iteration space dimensionaly to match number of "
+         "affected loops");
+  assert(OriginalInits.size() == NumLoops &&
+         "Expecting loop iteration space dimensionaly to match number of "
+         "affected loops");
+
+  // Decode the permutation clause.
+  constexpr uint64_t Permutation[] = {1, 0};
+
+  // Find the affected loops.
+  SmallVector<Stmt *> LoopStmts(NumLoops, nullptr);
+  collectLoopStmts(AStmt, LoopStmts);
+
+  // Collect pre-init statements on the order before the permuation.
+  SmallVector<Stmt *> PreInits;
+  for (auto I : llvm::seq<int>(NumLoops)) {
+    OMPLoopBasedDirective::HelperExprs &LoopHelper = LoopHelpers[I];
+
+    assert(LoopHelper.Counters.size() == 1 &&
+           "Single-dimensional loop iteration space expected");
+    auto *OrigCntVar = cast<DeclRefExpr>(LoopHelper.Counters.front());
+
+    std::string OrigVarName = OrigCntVar->getNameInfo().getAsString();
+    addLoopPreInits(Context, LoopHelper, LoopStmts[I], OriginalInits[I],
+                    PreInits);
+  }
+
+  SmallVector<VarDecl *> PermutedIndVars;
+  PermutedIndVars.resize(NumLoops);
+  CaptureVars CopyTransformer(SemaRef);
+
+  // Create the permuted loops from the inside to the outside of the
+  // interchanged loop nest. Body of the innermost new loop is the original
+  // innermost body.
+  Stmt *Inner = Body;
+  for (auto TargetIdx : llvm::reverse(llvm::seq<int>(NumLoops))) {
+    // Get the original loop that belongs to this new position.
+    uint64_t SourceIdx = Permutation[TargetIdx];
+    OMPLoopBasedDirective::HelperExprs &SourceHelper = LoopHelpers[SourceIdx];
+    Stmt *SourceLoopStmt = LoopStmts[SourceIdx];
+    assert(SourceHelper.Counters.size() == 1 &&
+           "Single-dimensional loop iteration space expected");
+    auto *OrigCntVar = cast<DeclRefExpr>(SourceHelper.Counters.front());
+
+    // Normalized loop counter variable: From 0 to n-1, always an integer type.
+    DeclRefExpr *IterVarRef = cast<DeclRefExpr>(SourceHelper.IterationVarRef);
+    QualType IVTy = IterVarRef->getType();
+    assert(IVTy->isIntegerType() &&
+           "Expected the logical iteration counter to be an integer");
+
+    std::string OrigVarName = OrigCntVar->getNameInfo().getAsString();
+    SourceLocation OrigVarLoc = IterVarRef->getExprLoc();
+
+    // Make a copy of the NumIterations expression for each use: By the AST
+    // constraints, every expression object in a DeclContext must be unique.
+    auto MakeNumIterations = [&CopyTransformer, &SourceHelper]() -> Expr * {
+      return AssertSuccess(
+          CopyTransformer.TransformExpr(SourceHelper.NumIterations));
+    };
+
+    // Iteration variable for the permuted loop. Reuse the one from
+    // checkOpenMPLoop which will also be used to update the original loop
+    // variable.
+    std::string PermutedCntName =
+        (Twine(".permuted_") + llvm::utostr(TargetIdx) + ".iv." + OrigVarName)
+            .str();
+    auto *PermutedCntDecl = cast<VarDecl>(IterVarRef->getDecl());
+    PermutedCntDecl->setDeclName(
+        &SemaRef.PP.getIdentifierTable().get(PermutedCntName));
+    PermutedIndVars[TargetIdx] = PermutedCntDecl;
+    auto MakePermutedRef = [this, PermutedCntDecl, IVTy, OrigVarLoc]() {
+      return buildDeclRefExpr(SemaRef, PermutedCntDecl, IVTy, OrigVarLoc);
+    };
+
+    // For init-statement:
+    // \code{c}
+    //   auto .permuted_{target}.iv = 0
+    // \endcode
+    ExprResult Zero = SemaRef.ActOnIntegerConstant(OrigVarLoc, 0);
+    if (!Zero.isUsable())
+      return StmtError();
+    SemaRef.AddInitializerToDecl(PermutedCntDecl, Zero.get(),
+                                 /*DirectInit=*/false);
+    StmtResult InitStmt = new (Context)
+        DeclStmt(DeclGroupRef(PermutedCntDecl), OrigCntVar->getBeginLoc(),
+                 OrigCntVar->getEndLoc());
+    if (!InitStmt.isUsable())
+      return StmtError();
+
+    // For cond-expression:
+    // \code{c}
+    //   .permuted_{target}.iv < NumIterations
+    // \endcode
+    ExprResult CondExpr =
+        SemaRef.BuildBinOp(CurScope, SourceHelper.Cond->getExprLoc(), BO_LT,
+                           MakePermutedRef(), MakeNumIterations());
+    if (!CondExpr.isUsable())
+      return StmtError();
+
+    // For incr-statement:
+    // \code{c}
+    //   ++.tile.iv
+    // \endcode
+    ExprResult IncrStmt = SemaRef.BuildUnaryOp(
+        CurScope, SourceHelper.Inc->getExprLoc(), UO_PreInc, MakePermutedRef());
+    if (!IncrStmt.isUsable())
+      return StmtError();
+
+    SmallVector<Stmt *, 4> BodyParts(SourceHelper.Updates.begin(),
+                                     SourceHelper.Updates.end());
+    if (auto *SourceCXXFor = dyn_cast<CXXForRangeStmt>(SourceLoopStmt))
+      BodyParts.push_back(SourceCXXFor->getLoopVarStmt());
+    BodyParts.push_back(Inner);
+    Inner = CompoundStmt::Create(Context, BodyParts, FPOptionsOverride(),
+                                 Inner->getBeginLoc(), Inner->getEndLoc());
+    Inner = new (Context) ForStmt(
+        Context, InitStmt.get(), CondExpr.get(), nullptr, IncrStmt.get(), Inner,
+        SourceHelper.Init->getBeginLoc(), SourceHelper.Init->getBeginLoc(),
+        SourceHelper.Inc->getEndLoc());
+  }
+
+  return OMPInterchangeDirective::Create(Context, StartLoc, EndLoc, Clauses,
+                                         NumLoops, AStmt, Inner,
+                                         buildPreInits(Context, PreInits));
+}
+
 OMPClause *SemaOpenMP::ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind,
                                                    Expr *Expr,
                                                    SourceLocation StartLoc,
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index c1a08da885507..8da35f522d8ff 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -9184,6 +9184,17 @@ TreeTransform<Derived>::TransformOMPReverseDirective(OMPReverseDirective *D) {
   return Res;
 }
 
+template <typename Derived>
+StmtResult TreeTransform<Derived>::TransformOMPInterchangeDirective(
+    OMPInterchangeDirective *D) {
+  DeclarationNameInfo DirName;
+  getDerived().getSema().OpenMP().StartOpenMPDSABlock(
+      D->getDirectiveKind(), DirName, nullptr, D->getBeginLoc());
+  StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
+  getDerived().getSema().OpenMP().EndOpenMPDSABlock(Res.get());
+  return Res;
+}
+
 template <typename Derived>
 StmtResult
 TreeTransform<Derived>::TransformOMPForDirective(OMPForDirective *D) {
diff --git a/clang/lib/Serialization/ASTReaderStmt.cpp b/clang/lib/Serialization/ASTReaderStmt.cpp
index ff72679f993e2..92e14473e7b51 100644
--- a/clang/lib/Serialization/ASTReaderStmt.cpp
+++ b/clang/lib/Serialization/ASTReaderStmt.cpp
@@ -2432,6 +2432,10 @@ void ASTStmtReader::VisitOMPReverseDirective(OMPReverseDirective *D) {
   VisitOMPLoopTransformationDirective(D);
 }
 
+void ASTStmtReader::VisitOMPInterchangeDirective(OMPInterchangeDirective *D) {
+  VisitOMPLoopTransformationDirective(D);
+}
+
 void ASTStmtReader::VisitOMPForDirective(OMPForDirective *D) {
   VisitOMPLoopDirective(D);
   D->setHasCancel(Record.readBool());
@@ -3448,6 +3452,13 @@ Stmt *ASTReader::ReadStmtFromStream(ModuleFile &F) {
       break;
     }
 
+    case STMT_OMP_INTERCHANGE_DIRECTIVE: {
+      unsigned NumLoops = Record[ASTStmtReader::NumStmtFields];
+      unsigned NumClauses = Record[ASTStmtReader::NumStmtFields + 1];
+      S = OMPInterchangeDirective::CreateEmpty(Context, NumClauses, NumLoops);
+      break;
+    }
+
     case STMT_OMP_FOR_DIRECTIVE: {
       unsigned CollapsedNum = Record[ASTStmtReader::NumStmtFields];
       unsigned NumClauses = Record[ASTStmtReader::NumStmtFields + 1];
diff --git a/clang/lib/Serialization/ASTWriterStmt.cpp b/clang/lib/Serialization/ASTWriterStmt.cpp
index ffb09ece981e6..0ac8e1967e17c 100644
--- a/clang/lib/Serialization/ASTWriterStmt.cpp
+++ b/clang/lib/Serialization/ASTWriterStmt.cpp
@@ -2417,6 +2417,11 @@ void ASTStmtWriter::VisitOMPReverseDirective(OMPReverseDirective *D) {
   Code = serialization::STMT_OMP_REVERSE_DIRECTIVE;
 }
 
+void ASTStmtWriter::VisitOMPInterchangeDirective(OMPInterchangeDirective *D) {
+  VisitOMPLoopTransformationDirective(D);
+  Code = serialization::STMT_OMP_INTERCHANGE_DIRECTIVE;
+}
+
 void ASTStmtWriter::VisitOMPForDirective(OMPForDirective *D) {
   VisitOMPLoopDirective(D);
   Record.writeBool(D->hasCancel());
diff --git a/clang/test/OpenMP/interchange_ast_print.cpp b/clang/test/OpenMP/interchange_ast_print.cpp
new file mode 100644
index 0000000000000..f8bf075cd300f
--- /dev/null
+++ b/clang/test/OpenMP/interchange_ast_print.cpp
@@ -0,0 +1,135 @@
+// Check no warnings/errors
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++20 -fopenmp -fopenmp-version=60 -fsyntax-only -verify %s
+// expected-no-diagnostics
+
+// Check AST and unparsing
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++20 -fopenmp -fopenmp-version=60 -ast-dump  %s | FileCheck %s --check-prefix=DUMP
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++20 -fopenmp -fopenmp-version=60 -ast-print %s | FileCheck %s --check-prefix=PRINT
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++20 -fopenmp -fopenmp-version=60 -emit-pch -o %t %s
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++20 -fopenmp -fopenmp-version=60 -include-pch %t -ast-dump-all %s | FileCheck %s --check-prefix=DUMP
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++20 -fopenmp -fopenmp-version=60 -include-pch %t -ast-print    %s | FileCheck %s --check-prefix=PRINT
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...);
+
+// PRINT-LABEL: void foo1(
+// DUMP-LABEL:  FunctionDecl {{.*}} foo1
+void foo1() {
+  // PRINT: #pragma omp interchange
+  // DUMP:  OMPInterchangeDirective
+  #pragma omp interchange
+  // PRINT: for (int i = 7; i < 17; i += 3)
+  // DUMP-NEXT: ForStmt
+  for (int i = 7; i < 17; i += 3)
+    // PRINT: for (int j = 7; j < 17; j += 3)
+    // DUMP:  ForStmt
+    for (int j = 7; j < 17; j += 3)
+      // PRINT: body(i, j);
+      // DUMP:  CallExpr
+      body(i, j);
+}
+
+
+
+
+// PRINT-LABEL: void foo3(
+// DUMP-LABEL:  FunctionDecl {{.*}} foo3
+void foo3() {
+  // PRINT: #pragma omp for collapse(3)
+  // DUMP:      OMPForDirective
+  // DUMP-NEXT:   OMPCollapseClause
+  // DUMP-NEXT:     ConstantExpr
+  // DUMP-NEXT:       value: Int 3
+  // DUMP-NEXT:     IntegerLiteral {{.*}} 3
+  // DUMP-NEXT:     CapturedStmt
+  // DUMP-NEXT:       CapturedDecl
+  #pragma omp for collapse(3)
+  // PRINT: #pragma omp interchange
+  // DUMP:  OMPInterchangeDirective
+  #pragma omp interchange
+  // PRINT: for (int i = 7; i < 17; i += 1)
+  // DUMP-NEXT: ForStmt
+  for (int i = 7; i < 17; i += 1)
+    // PRINT: for (int j = 7; j < 17; j += 1)
+    // DUMP:  ForStmt
+    for (int j = 7; j < 17; j += 1)
+      // PRINT: for (int k = 7; k < 17; k += 1)
+      // DUMP:  ForStmt
+      for (int k = 7; k < 17; k += 1)
+        // PRINT: body(i, j, k);
+        // DUMP:  CallExpr
+        body(i, j, k);
+}
+
+
+// PRINT-LABEL: void foo6(
+// DUMP-LABEL: FunctionTemplateDecl {{.*}} foo6
+template<int Tile>
+void foo6() {
+    // PRINT:     #pragma omp interchange
+    // DUMP:      OMPInterchangeDirective
+    #pragma omp interchange
+      // PRINT-NEXT: for (int i = 0; i < 11; i += 2)
+      // DUMP-NEXT:  ForStmt
+      for (int i = 0; i < 11; i += 2)
+        // PRINT-NEXT: #pragma omp tile sizes(Tile)
+        // DUMP:       OMPTileDirective
+        #pragma omp tile sizes(Tile)
+        // PRINT-NEXT: for (int j = 0; j < 13; j += 2)
+        // DUMP:       ForStmt
+        for (int j = 0; j < 13; j += 2)
+          // PRINT-NEXT: body(i, j);
+          // DUMP:       CallExpr
+          body(i, j);
+}
+
+// Also test instantiating the template.
+void tfoo6() {
+  foo6<32>();
+}
+
+
+// PRINT-LABEL: void foo7(
+// DUMP-LABEL: FunctionDecl {{.*}} foo7
+void foo7() {
+  double arr[128];
+  // PRINT: #pragma omp interchange
+  // DUMP:  OMPInterchangeDirective
+  #pragma omp interchange
+  // PRINT-NEXT: for (double c = 42; auto &&v : arr)
+  // DUMP-NEXT:  CXXForRangeStmt
+  for (double c = 42; auto &&v : arr)
+    // PRINT-NEXT: for (int i = 0; i < 42; i += 2)
+    // DUMP:       ForStmt
+    for (int i = 0; i < 42; i += 2)
+      // PRINT-NEXT: body(c, v, i);
+      // DUMP:       CallExpr
+      body(c, v, i);
+}
+
+
+// PRINT-LABEL: void foo8(
+// DUMP-LABEL: FunctionDecl {{.*}} foo8
+void foo8() {
+  double arr[128];
+  // PRINT: #pragma omp interchange
+  // DUMP:  OMPInterchangeDirective
+  #pragma omp interchange
+  // PRINT-NEXT: for (int i = 0; i < 42; i += 2)
+  // DUMP-NEXT:  ForStmt
+  for (int i = 0; i < 42; i += 2)
+    // PRINT-NEXT: for (double c = 42; auto &&v : arr)
+    // DUMP:       CXXForRangeStmt
+    for (double c = 42; auto &&v : arr)
+      // PRINT-NEXT: body(i, c, v);
+      // DUMP:       CallExpr
+      body(i, c, v);
+}
+
+#endif
+
diff --git a/clang/test/OpenMP/interchange_codegen.cpp b/clang/test/OpenMP/interchange_codegen.cpp
new file mode 100644
index 0000000000000..9c1782183cf98
--- /dev/null
+++ b/clang/test/OpenMP/interchange_codegen.cpp
@@ -0,0 +1,1990 @@
+// 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 _
+
+// expected-no-diagnostics
+
+// Check code generation
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -std=c++20 -fclang-abi-compat=latest -fopenmp -fopenmp-version=60 -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
+
+// Check same results after serialization round-trip
+// FIXME: They should be exactly the same but currently differ in function order
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -std=c++20 -fclang-abi-compat=latest -fopenmp -fopenmp-version=60 -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -std=c++20 -fclang-abi-compat=latest -fopenmp -fopenmp-version=60 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK2
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...) {}
+
+
+
+
+extern "C" void foo2(int start1, int start2, int end1, int end2, int step1, int step2) {
+#pragma omp interchange
+  for (int i = start1; i < end1; i += step1)
+    for (int j = start2; j < end2; j += step2)
+      body(i, j);
+}
+
+
+extern "C" void foo3() {
+#pragma omp for
+#pragma omp interchange
+  for (int i = 7; i < 17; i += 3)
+    for (int j = 7; j < 17; j += 3)
+      body(i, j);
+}
+
+
+extern "C" void foo4() {
+#pragma omp for collapse(2)
+  for (int k = 7; k < 17; k += 3)
+#pragma omp interchange
+    for (int i = 7; i < 17; i += 3)
+      for (int j = 7; j < 17; j += 3)
+        body(i, j);
+}
+
+
+extern "C" void foo6() {
+#pragma omp for collapse(4)
+  for (int i = 7; i < 17; i += 3)
+#pragma omp interchange
+    for (int j = 7; j < 17; j += 3)
+      for (int k = 7; k < 17; k += 3)
+        for (int l = 7; l < 17; l += 3)
+          body(i, j, k, l);
+}
+
+
+extern "C" void foo9() {
+  double arr[128];
+  #pragma omp interchange
+  for (double c = 42; auto && v : arr)
+    for (int i = 0; i < 42; i += 2)
+      body(c, v, i);
+}
+
+
+extern "C" void foo10() {
+  double A[128], B[16];
+  #pragma omp for collapse(4)
+  for (int i = 0; i < 128; ++i)
+    #pragma omp interchange
+    for (double c = 42; auto aa : A)
+      for (double d = 42; auto &bb : B)
+        for (int j = 0; j < 128; ++j)
+          body(i, c, aa, d, bb, j);
+}
+
+#endif /* HEADER */
+
+// CHECK1-LABEL: define {{[^@]+}}@body
+// CHECK1-SAME: (...) #[[ATTR0:[0-9]+]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@foo2
+// CHECK1-SAME: (i32 noundef [[START1:%.*]], i32 noundef [[START2:%.*]], i32 noundef [[END1:%.*]], i32 noundef [[END2:%.*]], i32 noundef [[STEP1:%.*]], i32 noundef [[STEP2:%.*]]) #[[ATTR0]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[START1_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[START2_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[END1_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[END2_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[STEP1_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[STEP2_ADDR:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTNEW_STEP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTNEW_STEP7:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_8:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_0_IV_J:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_1_IV_I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    store i32 [[START1]], ptr [[START1_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[START2]], ptr [[START2_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[END1]], ptr [[END1_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[END2]], ptr [[END2_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[STEP1]], ptr [[STEP1_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[STEP2]], ptr [[STEP2_ADDR]], align 4
+// CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[START1_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP0]], ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[START1_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[END1_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[STEP1_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[DOTNEW_STEP]], align 4
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], [[TMP5]]
+// CHECK1-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTNEW_STEP]], align 4
+// CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], [[TMP6]]
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTNEW_STEP]], align 4
+// CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], [[TMP7]]
+// CHECK1-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
+// CHECK1-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[START2_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP8]], ptr [[J]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[START2_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP9]], ptr [[DOTCAPTURE_EXPR_5]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[END2_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP10]], ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[STEP2_ADDR]], align 4
+// CHECK1-NEXT:    store i32 [[TMP11]], ptr [[DOTNEW_STEP7]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_5]], align 4
+// CHECK1-NEXT:    [[SUB9:%.*]] = sub i32 [[TMP12]], [[TMP13]]
+// CHECK1-NEXT:    [[SUB10:%.*]] = sub i32 [[SUB9]], 1
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTNEW_STEP7]], align 4
+// CHECK1-NEXT:    [[ADD11:%.*]] = add i32 [[SUB10]], [[TMP14]]
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTNEW_STEP7]], align 4
+// CHECK1-NEXT:    [[DIV12:%.*]] = udiv i32 [[ADD11]], [[TMP15]]
+// CHECK1-NEXT:    [[SUB13:%.*]] = sub i32 [[DIV12]], 1
+// CHECK1-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_8]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK1:       for.cond:
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_8]], align 4
+// CHECK1-NEXT:    [[ADD14:%.*]] = add i32 [[TMP17]], 1
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp ult i32 [[TMP16]], [[ADD14]]
+// CHECK1-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END24:%.*]]
+// CHECK1:       for.body:
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_5]], align 4
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTNEW_STEP7]], align 4
+// CHECK1-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], [[TMP20]]
+// CHECK1-NEXT:    [[ADD15:%.*]] = add i32 [[TMP18]], [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD15]], ptr [[J]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND16:%.*]]
+// CHECK1:       for.cond16:
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK1-NEXT:    [[ADD17:%.*]] = add i32 [[TMP22]], 1
+// CHECK1-NEXT:    [[CMP18:%.*]] = icmp ult i32 [[TMP21]], [[ADD17]]
+// CHECK1-NEXT:    br i1 [[CMP18]], label [[FOR_BODY19:%.*]], label [[FOR_END:%.*]]
+// CHECK1:       for.body19:
+// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTNEW_STEP]], align 4
+// CHECK1-NEXT:    [[MUL20:%.*]] = mul i32 [[TMP24]], [[TMP25]]
+// CHECK1-NEXT:    [[ADD21:%.*]] = add i32 [[TMP23]], [[MUL20]]
+// CHECK1-NEXT:    store i32 [[ADD21]], ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP27:%.*]] = load i32, ptr [[J]], align 4
+// CHECK1-NEXT:    call void (...) @body(i32 noundef [[TMP26]], i32 noundef [[TMP27]])
+// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK1:       for.inc:
+// CHECK1-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[INC:%.*]] = add i32 [[TMP28]], 1
+// CHECK1-NEXT:    store i32 [[INC]], ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND16]], !llvm.loop [[LOOP3:![0-9]+]]
+// CHECK1:       for.end:
+// CHECK1-NEXT:    br label [[FOR_INC22:%.*]]
+// CHECK1:       for.inc22:
+// CHECK1-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    [[INC23:%.*]] = add i32 [[TMP29]], 1
+// CHECK1-NEXT:    store i32 [[INC23]], ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
+// CHECK1:       for.end24:
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@foo3
+// CHECK1-SAME: () #[[ATTR0]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[J:%.*]] = 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:    [[DOTPERMUTED_0_IV_J:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_1_IV_I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2:[0-9]+]])
+// CHECK1-NEXT:    store i32 7, ptr [[I]], align 4
+// CHECK1-NEXT:    store i32 7, ptr [[J]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 3, 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:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 3
+// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK1:       cond.true:
+// CHECK1-NEXT:    br label [[COND_END:%.*]]
+// CHECK1:       cond.false:
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    br label [[COND_END]]
+// CHECK1:       cond.end:
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1:       omp.inner.for.cond:
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
+// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1:       omp.inner.for.body:
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    [[MUL2:%.*]] = mul nsw i32 [[TMP7]], 3
+// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 7, [[MUL2]]
+// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[J]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK1:       for.cond:
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP8]], 4
+// CHECK1-NEXT:    br i1 [[CMP4]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// CHECK1:       for.body:
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP9]], 3
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 7, [[MUL5]]
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J]], align 4
+// CHECK1-NEXT:    call void (...) @body(i32 noundef [[TMP10]], i32 noundef [[TMP11]])
+// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK1:       for.inc:
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK1-NEXT:    store i32 [[INC]], ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
+// CHECK1:       for.end:
+// 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:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1:       omp.inner.for.end:
+// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK1:       omp.loop.exit:
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3:[0-9]+]], i32 [[TMP0]])
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@foo4
+// CHECK1-SAME: () #[[ATTR0]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[J:%.*]] = 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:    [[K:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_0_IV_J:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_1_IV_I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
+// CHECK1-NEXT:    store i32 7, ptr [[I]], align 4
+// CHECK1-NEXT:    store i32 7, ptr [[J]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 15, 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:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 15
+// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK1:       cond.true:
+// CHECK1-NEXT:    br label [[COND_END:%.*]]
+// CHECK1:       cond.false:
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    br label [[COND_END]]
+// CHECK1:       cond.end:
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 15, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1:       omp.inner.for.cond:
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
+// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1:       omp.inner.for.body:
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP6]], 4
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 3
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 7, [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[K]], align 4
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP8]], 4
+// CHECK1-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 4
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], [[MUL4]]
+// CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK1-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[TMP9]], 3
+// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 7, [[MUL7]]
+// CHECK1-NEXT:    store i32 [[ADD8]], ptr [[J]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK1:       for.cond:
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP10]], 4
+// CHECK1-NEXT:    br i1 [[CMP9]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// CHECK1:       for.body:
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP11]], 3
+// CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 7, [[MUL10]]
+// CHECK1-NEXT:    store i32 [[ADD11]], ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[J]], align 4
+// CHECK1-NEXT:    call void (...) @body(i32 noundef [[TMP12]], i32 noundef [[TMP13]])
+// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK1:       for.inc:
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP14]], 1
+// CHECK1-NEXT:    store i32 [[INC]], ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
+// CHECK1:       for.end:
+// 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:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK1-NEXT:    store i32 [[ADD12]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1:       omp.inner.for.end:
+// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK1:       omp.loop.exit:
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@foo6
+// CHECK1-SAME: () #[[ATTR0]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[K:%.*]] = 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:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_0_IV_K:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_1_IV_J:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[L:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
+// CHECK1-NEXT:    store i32 7, ptr [[J]], align 4
+// CHECK1-NEXT:    store i32 7, ptr [[K]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 255, 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:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 255
+// CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK1:       cond.true:
+// CHECK1-NEXT:    br label [[COND_END:%.*]]
+// CHECK1:       cond.false:
+// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    br label [[COND_END]]
+// CHECK1:       cond.end:
+// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 255, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK1-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1:       omp.inner.for.cond:
+// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK1-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
+// CHECK1-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1:       omp.inner.for.body:
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP6]], 64
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 3
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 7, [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV5:%.*]] = sdiv i32 [[TMP8]], 64
+// CHECK1-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[DIV5]], 64
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], [[MUL6]]
+// CHECK1-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB]], 16
+// CHECK1-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
+// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
+// CHECK1-NEXT:    store i32 [[ADD9]], ptr [[DOTPERMUTED_0_IV_K]], align 4
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV10:%.*]] = sdiv i32 [[TMP10]], 64
+// CHECK1-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 64
+// CHECK1-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP9]], [[MUL11]]
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV13:%.*]] = sdiv i32 [[TMP12]], 64
+// CHECK1-NEXT:    [[MUL14:%.*]] = mul nsw i32 [[DIV13]], 64
+// CHECK1-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP11]], [[MUL14]]
+// CHECK1-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 16
+// CHECK1-NEXT:    [[MUL17:%.*]] = mul nsw i32 [[DIV16]], 16
+// CHECK1-NEXT:    [[SUB18:%.*]] = sub nsw i32 [[SUB12]], [[MUL17]]
+// CHECK1-NEXT:    [[DIV19:%.*]] = sdiv i32 [[SUB18]], 4
+// CHECK1-NEXT:    [[MUL20:%.*]] = mul nsw i32 [[DIV19]], 1
+// CHECK1-NEXT:    [[ADD21:%.*]] = add nsw i32 0, [[MUL20]]
+// CHECK1-NEXT:    store i32 [[ADD21]], ptr [[DOTPERMUTED_1_IV_J]], align 4
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV22:%.*]] = sdiv i32 [[TMP14]], 64
+// CHECK1-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 64
+// CHECK1-NEXT:    [[SUB24:%.*]] = sub nsw i32 [[TMP13]], [[MUL23]]
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV25:%.*]] = sdiv i32 [[TMP16]], 64
+// CHECK1-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[DIV25]], 64
+// CHECK1-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP15]], [[MUL26]]
+// CHECK1-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 16
+// CHECK1-NEXT:    [[MUL29:%.*]] = mul nsw i32 [[DIV28]], 16
+// CHECK1-NEXT:    [[SUB30:%.*]] = sub nsw i32 [[SUB24]], [[MUL29]]
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV31:%.*]] = sdiv i32 [[TMP18]], 64
+// CHECK1-NEXT:    [[MUL32:%.*]] = mul nsw i32 [[DIV31]], 64
+// CHECK1-NEXT:    [[SUB33:%.*]] = sub nsw i32 [[TMP17]], [[MUL32]]
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[DIV34:%.*]] = sdiv i32 [[TMP20]], 64
+// CHECK1-NEXT:    [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 64
+// CHECK1-NEXT:    [[SUB36:%.*]] = sub nsw i32 [[TMP19]], [[MUL35]]
+// CHECK1-NEXT:    [[DIV37:%.*]] = sdiv i32 [[SUB36]], 16
+// CHECK1-NEXT:    [[MUL38:%.*]] = mul nsw i32 [[DIV37]], 16
+// CHECK1-NEXT:    [[SUB39:%.*]] = sub nsw i32 [[SUB33]], [[MUL38]]
+// CHECK1-NEXT:    [[DIV40:%.*]] = sdiv i32 [[SUB39]], 4
+// CHECK1-NEXT:    [[MUL41:%.*]] = mul nsw i32 [[DIV40]], 4
+// CHECK1-NEXT:    [[SUB42:%.*]] = sub nsw i32 [[SUB30]], [[MUL41]]
+// CHECK1-NEXT:    [[MUL43:%.*]] = mul nsw i32 [[SUB42]], 3
+// CHECK1-NEXT:    [[ADD44:%.*]] = add nsw i32 7, [[MUL43]]
+// CHECK1-NEXT:    store i32 [[ADD44]], ptr [[L]], align 4
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_K]], align 4
+// CHECK1-NEXT:    [[MUL45:%.*]] = mul nsw i32 [[TMP21]], 3
+// CHECK1-NEXT:    [[ADD46:%.*]] = add nsw i32 7, [[MUL45]]
+// CHECK1-NEXT:    store i32 [[ADD46]], ptr [[K]], align 4
+// CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_J]], align 4
+// CHECK1-NEXT:    [[MUL47:%.*]] = mul nsw i32 [[TMP22]], 3
+// CHECK1-NEXT:    [[ADD48:%.*]] = add nsw i32 7, [[MUL47]]
+// CHECK1-NEXT:    store i32 [[ADD48]], ptr [[J]], align 4
+// CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    [[TMP24:%.*]] = load i32, ptr [[J]], align 4
+// CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[K]], align 4
+// CHECK1-NEXT:    [[TMP26:%.*]] = load i32, ptr [[L]], align 4
+// CHECK1-NEXT:    call void (...) @body(i32 noundef [[TMP23]], i32 noundef [[TMP24]], i32 noundef [[TMP25]], i32 noundef [[TMP26]])
+// 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:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    [[ADD49:%.*]] = add nsw i32 [[TMP27]], 1
+// CHECK1-NEXT:    store i32 [[ADD49]], ptr [[DOTOMP_IV]], align 4
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1:       omp.inner.for.end:
+// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK1:       omp.loop.exit:
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@foo9
+// CHECK1-SAME: () #[[ATTR0]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[ARR:%.*]] = alloca [128 x double], align 16
+// CHECK1-NEXT:    [[C:%.*]] = alloca double, align 8
+// CHECK1-NEXT:    [[__RANGE2:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[__END2:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[__BEGIN2:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_0_IV_I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_1_IV___BEGIN2:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[V:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    store double 4.200000e+01, ptr [[C]], align 8
+// CHECK1-NEXT:    store ptr [[ARR]], ptr [[__RANGE2]], align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[__RANGE2]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP0]], i64 0, i64 0
+// CHECK1-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds double, ptr [[ARRAYDECAY]], i64 128
+// CHECK1-NEXT:    store ptr [[ADD_PTR]], ptr [[__END2]], align 8
+// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[__RANGE2]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY1:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP1]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[ARRAYDECAY1]], ptr [[__BEGIN2]], align 8
+// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[__RANGE2]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY2:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP2]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[ARRAYDECAY2]], ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[__END2]], align 8
+// CHECK1-NEXT:    store ptr [[TMP3]], ptr [[DOTCAPTURE_EXPR_3]], align 8
+// CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_3]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK1-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[TMP4]] to i64
+// CHECK1-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[TMP5]] to i64
+// CHECK1-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
+// CHECK1-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 8
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
+// CHECK1-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[DIV]], 1
+// CHECK1-NEXT:    store i64 [[SUB5]], ptr [[DOTCAPTURE_EXPR_4]], align 8
+// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK1-NEXT:    store i32 0, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK1:       for.cond:
+// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], 21
+// CHECK1-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END15:%.*]]
+// CHECK1:       for.body:
+// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 2
+// CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK1-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
+// CHECK1-NEXT:    store i64 0, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK1-NEXT:    br label [[FOR_COND7:%.*]]
+// CHECK1:       for.cond7:
+// CHECK1-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK1-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_4]], align 8
+// CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i64 [[TMP9]], 1
+// CHECK1-NEXT:    [[CMP9:%.*]] = icmp slt i64 [[TMP8]], [[ADD8]]
+// CHECK1-NEXT:    br i1 [[CMP9]], label [[FOR_BODY10:%.*]], label [[FOR_END:%.*]]
+// CHECK1:       for.body10:
+// CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK1-NEXT:    [[MUL11:%.*]] = mul nsw i64 [[TMP11]], 1
+// CHECK1-NEXT:    [[ADD_PTR12:%.*]] = getelementptr inbounds double, ptr [[TMP10]], i64 [[MUL11]]
+// CHECK1-NEXT:    store ptr [[ADD_PTR12]], ptr [[__BEGIN2]], align 8
+// CHECK1-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[__BEGIN2]], align 8
+// CHECK1-NEXT:    store ptr [[TMP12]], ptr [[V]], align 8
+// CHECK1-NEXT:    [[TMP13:%.*]] = load double, ptr [[C]], align 8
+// CHECK1-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[V]], align 8
+// CHECK1-NEXT:    [[TMP15:%.*]] = load double, ptr [[TMP14]], align 8
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4
+// CHECK1-NEXT:    call void (...) @body(double noundef [[TMP13]], double noundef [[TMP15]], i32 noundef [[TMP16]])
+// CHECK1-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK1:       for.inc:
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK1-NEXT:    [[INC:%.*]] = add nsw i64 [[TMP17]], 1
+// CHECK1-NEXT:    store i64 [[INC]], ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK1-NEXT:    br label [[FOR_COND7]], !llvm.loop [[LOOP8:![0-9]+]]
+// CHECK1:       for.end:
+// CHECK1-NEXT:    br label [[FOR_INC13:%.*]]
+// CHECK1:       for.inc13:
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK1-NEXT:    [[INC14:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK1-NEXT:    store i32 [[INC14]], ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK1-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
+// CHECK1:       for.end15:
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK1-LABEL: define {{[^@]+}}@foo10
+// CHECK1-SAME: () #[[ATTR0]] {
+// CHECK1-NEXT:  entry:
+// CHECK1-NEXT:    [[A:%.*]] = alloca [128 x double], align 16
+// CHECK1-NEXT:    [[B:%.*]] = alloca [16 x double], align 16
+// CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[C:%.*]] = alloca double, align 8
+// CHECK1-NEXT:    [[__RANGE3:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[__END3:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[__BEGIN3:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_7:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[D:%.*]] = alloca double, align 8
+// CHECK1-NEXT:    [[__RANGE4:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[__END4:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[__BEGIN4:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_14:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_15:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTCAPTURE_EXPR_28:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_0_IV___BEGIN4:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTPERMUTED_1_IV___BEGIN3:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[I37:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[DOTPERMUTED_0_IV___BEGIN438:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[DOTPERMUTED_1_IV___BEGIN339:%.*]] = alloca i64, align 8
+// CHECK1-NEXT:    [[J40:%.*]] = alloca i32, align 4
+// CHECK1-NEXT:    [[BB:%.*]] = alloca ptr, align 8
+// CHECK1-NEXT:    [[AA:%.*]] = alloca double, align 8
+// CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
+// CHECK1-NEXT:    store double 4.200000e+01, ptr [[C]], align 8
+// CHECK1-NEXT:    store ptr [[A]], ptr [[__RANGE3]], align 8
+// CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[__RANGE3]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP1]], i64 0, i64 0
+// CHECK1-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds double, ptr [[ARRAYDECAY]], i64 128
+// CHECK1-NEXT:    store ptr [[ADD_PTR]], ptr [[__END3]], align 8
+// CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[__RANGE3]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY4:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP2]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[ARRAYDECAY4]], ptr [[__BEGIN3]], align 8
+// CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[__RANGE3]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY5:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP3]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[ARRAYDECAY5]], ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[__END3]], align 8
+// CHECK1-NEXT:    store ptr [[TMP4]], ptr [[DOTCAPTURE_EXPR_6]], align 8
+// CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_6]], align 8
+// CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK1-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[TMP5]] to i64
+// CHECK1-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[TMP6]] to i64
+// CHECK1-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
+// CHECK1-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 8
+// CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
+// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
+// CHECK1-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
+// CHECK1-NEXT:    [[SUB8:%.*]] = sub nsw i64 [[DIV]], 1
+// CHECK1-NEXT:    store i64 [[SUB8]], ptr [[DOTCAPTURE_EXPR_7]], align 8
+// CHECK1-NEXT:    store double 4.200000e+01, ptr [[D]], align 8
+// CHECK1-NEXT:    store ptr [[B]], ptr [[__RANGE4]], align 8
+// CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[__RANGE4]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY9:%.*]] = getelementptr inbounds [16 x double], ptr [[TMP7]], i64 0, i64 0
+// CHECK1-NEXT:    [[ADD_PTR10:%.*]] = getelementptr inbounds double, ptr [[ARRAYDECAY9]], i64 16
+// CHECK1-NEXT:    store ptr [[ADD_PTR10]], ptr [[__END4]], align 8
+// CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[__RANGE4]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY11:%.*]] = getelementptr inbounds [16 x double], ptr [[TMP8]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[ARRAYDECAY11]], ptr [[__BEGIN4]], align 8
+// CHECK1-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[__RANGE4]], align 8
+// CHECK1-NEXT:    [[ARRAYDECAY13:%.*]] = getelementptr inbounds [16 x double], ptr [[TMP9]], i64 0, i64 0
+// CHECK1-NEXT:    store ptr [[ARRAYDECAY13]], ptr [[DOTCAPTURE_EXPR_12]], align 8
+// CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[__END4]], align 8
+// CHECK1-NEXT:    store ptr [[TMP10]], ptr [[DOTCAPTURE_EXPR_14]], align 8
+// CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_14]], align 8
+// CHECK1-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_12]], align 8
+// CHECK1-NEXT:    [[SUB_PTR_LHS_CAST16:%.*]] = ptrtoint ptr [[TMP11]] to i64
+// CHECK1-NEXT:    [[SUB_PTR_RHS_CAST17:%.*]] = ptrtoint ptr [[TMP12]] to i64
+// CHECK1-NEXT:    [[SUB_PTR_SUB18:%.*]] = sub i64 [[SUB_PTR_LHS_CAST16]], [[SUB_PTR_RHS_CAST17]]
+// CHECK1-NEXT:    [[SUB_PTR_DIV19:%.*]] = sdiv exact i64 [[SUB_PTR_SUB18]], 8
+// CHECK1-NEXT:    [[SUB20:%.*]] = sub nsw i64 [[SUB_PTR_DIV19]], 1
+// CHECK1-NEXT:    [[ADD21:%.*]] = add nsw i64 [[SUB20]], 1
+// CHECK1-NEXT:    [[DIV22:%.*]] = sdiv i64 [[ADD21]], 1
+// CHECK1-NEXT:    [[SUB23:%.*]] = sub nsw i64 [[DIV22]], 1
+// CHECK1-NEXT:    store i64 [[SUB23]], ptr [[DOTCAPTURE_EXPR_15]], align 8
+// CHECK1-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_15]], align 8
+// CHECK1-NEXT:    [[ADD25:%.*]] = add nsw i64 [[TMP13]], 1
+// CHECK1-NEXT:    store i64 [[ADD25]], ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_7]], align 8
+// CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i64 [[TMP14]], 1
+// CHECK1-NEXT:    store i64 [[ADD27]], ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB29:%.*]] = sub nsw i64 [[TMP15]], 0
+// CHECK1-NEXT:    [[DIV30:%.*]] = sdiv i64 [[SUB29]], 1
+// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i64 128, [[DIV30]]
+// CHECK1-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB31:%.*]] = sub nsw i64 [[TMP16]], 0
+// CHECK1-NEXT:    [[DIV32:%.*]] = sdiv i64 [[SUB31]], 1
+// CHECK1-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[MUL]], [[DIV32]]
+// CHECK1-NEXT:    [[MUL34:%.*]] = mul nsw i64 [[MUL33]], 128
+// CHECK1-NEXT:    [[SUB35:%.*]] = sub nsw i64 [[MUL34]], 1
+// CHECK1-NEXT:    store i64 [[SUB35]], ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK1-NEXT:    store i64 0, ptr [[DOTPERMUTED_0_IV___BEGIN4]], align 8
+// CHECK1-NEXT:    store i64 0, ptr [[DOTPERMUTED_1_IV___BEGIN3]], align 8
+// CHECK1-NEXT:    store i32 0, ptr [[J]], align 4
+// CHECK1-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i64 0, [[TMP17]]
+// CHECK1-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK1:       land.lhs.true:
+// CHECK1-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[CMP36:%.*]] = icmp slt i64 0, [[TMP18]]
+// CHECK1-NEXT:    br i1 [[CMP36]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// CHECK1:       omp.precond.then:
+// CHECK1-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
+// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK1-NEXT:    store i64 [[TMP19]], ptr [[DOTOMP_UB]], align 8
+// CHECK1-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK1-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// CHECK1-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// CHECK1-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK1-NEXT:    [[CMP41:%.*]] = icmp sgt i64 [[TMP20]], [[TMP21]]
+// CHECK1-NEXT:    br i1 [[CMP41]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK1:       cond.true:
+// CHECK1-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK1-NEXT:    br label [[COND_END:%.*]]
+// CHECK1:       cond.false:
+// CHECK1-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// CHECK1-NEXT:    br label [[COND_END]]
+// CHECK1:       cond.end:
+// CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP22]], [[COND_TRUE]] ], [ [[TMP23]], [[COND_FALSE]] ]
+// CHECK1-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
+// CHECK1-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
+// CHECK1-NEXT:    store i64 [[TMP24]], ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK1:       omp.inner.for.cond:
+// CHECK1-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// CHECK1-NEXT:    [[CMP42:%.*]] = icmp sle i64 [[TMP25]], [[TMP26]]
+// CHECK1-NEXT:    br i1 [[CMP42]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK1:       omp.inner.for.body:
+// CHECK1-NEXT:    [[TMP27:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP28:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB43:%.*]] = sub nsw i64 [[TMP28]], 0
+// CHECK1-NEXT:    [[DIV44:%.*]] = sdiv i64 [[SUB43]], 1
+// CHECK1-NEXT:    [[MUL45:%.*]] = mul nsw i64 1, [[DIV44]]
+// CHECK1-NEXT:    [[TMP29:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB46:%.*]] = sub nsw i64 [[TMP29]], 0
+// CHECK1-NEXT:    [[DIV47:%.*]] = sdiv i64 [[SUB46]], 1
+// CHECK1-NEXT:    [[MUL48:%.*]] = mul nsw i64 [[MUL45]], [[DIV47]]
+// CHECK1-NEXT:    [[MUL49:%.*]] = mul nsw i64 [[MUL48]], 128
+// CHECK1-NEXT:    [[DIV50:%.*]] = sdiv i64 [[TMP27]], [[MUL49]]
+// CHECK1-NEXT:    [[MUL51:%.*]] = mul nsw i64 [[DIV50]], 1
+// CHECK1-NEXT:    [[ADD52:%.*]] = add nsw i64 0, [[MUL51]]
+// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[ADD52]] to i32
+// CHECK1-NEXT:    store i32 [[CONV]], ptr [[I37]], align 4
+// CHECK1-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP31:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP32:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB53:%.*]] = sub nsw i64 [[TMP32]], 0
+// CHECK1-NEXT:    [[DIV54:%.*]] = sdiv i64 [[SUB53]], 1
+// CHECK1-NEXT:    [[MUL55:%.*]] = mul nsw i64 1, [[DIV54]]
+// CHECK1-NEXT:    [[TMP33:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB56:%.*]] = sub nsw i64 [[TMP33]], 0
+// CHECK1-NEXT:    [[DIV57:%.*]] = sdiv i64 [[SUB56]], 1
+// CHECK1-NEXT:    [[MUL58:%.*]] = mul nsw i64 [[MUL55]], [[DIV57]]
+// CHECK1-NEXT:    [[MUL59:%.*]] = mul nsw i64 [[MUL58]], 128
+// CHECK1-NEXT:    [[DIV60:%.*]] = sdiv i64 [[TMP31]], [[MUL59]]
+// CHECK1-NEXT:    [[TMP34:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB61:%.*]] = sub nsw i64 [[TMP34]], 0
+// CHECK1-NEXT:    [[DIV62:%.*]] = sdiv i64 [[SUB61]], 1
+// CHECK1-NEXT:    [[MUL63:%.*]] = mul nsw i64 1, [[DIV62]]
+// CHECK1-NEXT:    [[TMP35:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB64:%.*]] = sub nsw i64 [[TMP35]], 0
+// CHECK1-NEXT:    [[DIV65:%.*]] = sdiv i64 [[SUB64]], 1
+// CHECK1-NEXT:    [[MUL66:%.*]] = mul nsw i64 [[MUL63]], [[DIV65]]
+// CHECK1-NEXT:    [[MUL67:%.*]] = mul nsw i64 [[MUL66]], 128
+// CHECK1-NEXT:    [[MUL68:%.*]] = mul nsw i64 [[DIV60]], [[MUL67]]
+// CHECK1-NEXT:    [[SUB69:%.*]] = sub nsw i64 [[TMP30]], [[MUL68]]
+// CHECK1-NEXT:    [[TMP36:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB70:%.*]] = sub nsw i64 [[TMP36]], 0
+// CHECK1-NEXT:    [[DIV71:%.*]] = sdiv i64 [[SUB70]], 1
+// CHECK1-NEXT:    [[MUL72:%.*]] = mul nsw i64 1, [[DIV71]]
+// CHECK1-NEXT:    [[MUL73:%.*]] = mul nsw i64 [[MUL72]], 128
+// CHECK1-NEXT:    [[DIV74:%.*]] = sdiv i64 [[SUB69]], [[MUL73]]
+// CHECK1-NEXT:    [[MUL75:%.*]] = mul nsw i64 [[DIV74]], 1
+// CHECK1-NEXT:    [[ADD76:%.*]] = add nsw i64 0, [[MUL75]]
+// CHECK1-NEXT:    store i64 [[ADD76]], ptr [[DOTPERMUTED_0_IV___BEGIN438]], align 8
+// CHECK1-NEXT:    [[TMP37:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP38:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP39:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB77:%.*]] = sub nsw i64 [[TMP39]], 0
+// CHECK1-NEXT:    [[DIV78:%.*]] = sdiv i64 [[SUB77]], 1
+// CHECK1-NEXT:    [[MUL79:%.*]] = mul nsw i64 1, [[DIV78]]
+// CHECK1-NEXT:    [[TMP40:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB80:%.*]] = sub nsw i64 [[TMP40]], 0
+// CHECK1-NEXT:    [[DIV81:%.*]] = sdiv i64 [[SUB80]], 1
+// CHECK1-NEXT:    [[MUL82:%.*]] = mul nsw i64 [[MUL79]], [[DIV81]]
+// CHECK1-NEXT:    [[MUL83:%.*]] = mul nsw i64 [[MUL82]], 128
+// CHECK1-NEXT:    [[DIV84:%.*]] = sdiv i64 [[TMP38]], [[MUL83]]
+// CHECK1-NEXT:    [[TMP41:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB85:%.*]] = sub nsw i64 [[TMP41]], 0
+// CHECK1-NEXT:    [[DIV86:%.*]] = sdiv i64 [[SUB85]], 1
+// CHECK1-NEXT:    [[MUL87:%.*]] = mul nsw i64 1, [[DIV86]]
+// CHECK1-NEXT:    [[TMP42:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB88:%.*]] = sub nsw i64 [[TMP42]], 0
+// CHECK1-NEXT:    [[DIV89:%.*]] = sdiv i64 [[SUB88]], 1
+// CHECK1-NEXT:    [[MUL90:%.*]] = mul nsw i64 [[MUL87]], [[DIV89]]
+// CHECK1-NEXT:    [[MUL91:%.*]] = mul nsw i64 [[MUL90]], 128
+// CHECK1-NEXT:    [[MUL92:%.*]] = mul nsw i64 [[DIV84]], [[MUL91]]
+// CHECK1-NEXT:    [[SUB93:%.*]] = sub nsw i64 [[TMP37]], [[MUL92]]
+// CHECK1-NEXT:    [[TMP43:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP44:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP45:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB94:%.*]] = sub nsw i64 [[TMP45]], 0
+// CHECK1-NEXT:    [[DIV95:%.*]] = sdiv i64 [[SUB94]], 1
+// CHECK1-NEXT:    [[MUL96:%.*]] = mul nsw i64 1, [[DIV95]]
+// CHECK1-NEXT:    [[TMP46:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB97:%.*]] = sub nsw i64 [[TMP46]], 0
+// CHECK1-NEXT:    [[DIV98:%.*]] = sdiv i64 [[SUB97]], 1
+// CHECK1-NEXT:    [[MUL99:%.*]] = mul nsw i64 [[MUL96]], [[DIV98]]
+// CHECK1-NEXT:    [[MUL100:%.*]] = mul nsw i64 [[MUL99]], 128
+// CHECK1-NEXT:    [[DIV101:%.*]] = sdiv i64 [[TMP44]], [[MUL100]]
+// CHECK1-NEXT:    [[TMP47:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB102:%.*]] = sub nsw i64 [[TMP47]], 0
+// CHECK1-NEXT:    [[DIV103:%.*]] = sdiv i64 [[SUB102]], 1
+// CHECK1-NEXT:    [[MUL104:%.*]] = mul nsw i64 1, [[DIV103]]
+// CHECK1-NEXT:    [[TMP48:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB105:%.*]] = sub nsw i64 [[TMP48]], 0
+// CHECK1-NEXT:    [[DIV106:%.*]] = sdiv i64 [[SUB105]], 1
+// CHECK1-NEXT:    [[MUL107:%.*]] = mul nsw i64 [[MUL104]], [[DIV106]]
+// CHECK1-NEXT:    [[MUL108:%.*]] = mul nsw i64 [[MUL107]], 128
+// CHECK1-NEXT:    [[MUL109:%.*]] = mul nsw i64 [[DIV101]], [[MUL108]]
+// CHECK1-NEXT:    [[SUB110:%.*]] = sub nsw i64 [[TMP43]], [[MUL109]]
+// CHECK1-NEXT:    [[TMP49:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB111:%.*]] = sub nsw i64 [[TMP49]], 0
+// CHECK1-NEXT:    [[DIV112:%.*]] = sdiv i64 [[SUB111]], 1
+// CHECK1-NEXT:    [[MUL113:%.*]] = mul nsw i64 1, [[DIV112]]
+// CHECK1-NEXT:    [[MUL114:%.*]] = mul nsw i64 [[MUL113]], 128
+// CHECK1-NEXT:    [[DIV115:%.*]] = sdiv i64 [[SUB110]], [[MUL114]]
+// CHECK1-NEXT:    [[TMP50:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB116:%.*]] = sub nsw i64 [[TMP50]], 0
+// CHECK1-NEXT:    [[DIV117:%.*]] = sdiv i64 [[SUB116]], 1
+// CHECK1-NEXT:    [[MUL118:%.*]] = mul nsw i64 1, [[DIV117]]
+// CHECK1-NEXT:    [[MUL119:%.*]] = mul nsw i64 [[MUL118]], 128
+// CHECK1-NEXT:    [[MUL120:%.*]] = mul nsw i64 [[DIV115]], [[MUL119]]
+// CHECK1-NEXT:    [[SUB121:%.*]] = sub nsw i64 [[SUB93]], [[MUL120]]
+// CHECK1-NEXT:    [[DIV122:%.*]] = sdiv i64 [[SUB121]], 128
+// CHECK1-NEXT:    [[MUL123:%.*]] = mul nsw i64 [[DIV122]], 1
+// CHECK1-NEXT:    [[ADD124:%.*]] = add nsw i64 0, [[MUL123]]
+// CHECK1-NEXT:    store i64 [[ADD124]], ptr [[DOTPERMUTED_1_IV___BEGIN339]], align 8
+// CHECK1-NEXT:    [[TMP51:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP52:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP53:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB125:%.*]] = sub nsw i64 [[TMP53]], 0
+// CHECK1-NEXT:    [[DIV126:%.*]] = sdiv i64 [[SUB125]], 1
+// CHECK1-NEXT:    [[MUL127:%.*]] = mul nsw i64 1, [[DIV126]]
+// CHECK1-NEXT:    [[TMP54:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB128:%.*]] = sub nsw i64 [[TMP54]], 0
+// CHECK1-NEXT:    [[DIV129:%.*]] = sdiv i64 [[SUB128]], 1
+// CHECK1-NEXT:    [[MUL130:%.*]] = mul nsw i64 [[MUL127]], [[DIV129]]
+// CHECK1-NEXT:    [[MUL131:%.*]] = mul nsw i64 [[MUL130]], 128
+// CHECK1-NEXT:    [[DIV132:%.*]] = sdiv i64 [[TMP52]], [[MUL131]]
+// CHECK1-NEXT:    [[TMP55:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB133:%.*]] = sub nsw i64 [[TMP55]], 0
+// CHECK1-NEXT:    [[DIV134:%.*]] = sdiv i64 [[SUB133]], 1
+// CHECK1-NEXT:    [[MUL135:%.*]] = mul nsw i64 1, [[DIV134]]
+// CHECK1-NEXT:    [[TMP56:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB136:%.*]] = sub nsw i64 [[TMP56]], 0
+// CHECK1-NEXT:    [[DIV137:%.*]] = sdiv i64 [[SUB136]], 1
+// CHECK1-NEXT:    [[MUL138:%.*]] = mul nsw i64 [[MUL135]], [[DIV137]]
+// CHECK1-NEXT:    [[MUL139:%.*]] = mul nsw i64 [[MUL138]], 128
+// CHECK1-NEXT:    [[MUL140:%.*]] = mul nsw i64 [[DIV132]], [[MUL139]]
+// CHECK1-NEXT:    [[SUB141:%.*]] = sub nsw i64 [[TMP51]], [[MUL140]]
+// CHECK1-NEXT:    [[TMP57:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP58:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP59:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB142:%.*]] = sub nsw i64 [[TMP59]], 0
+// CHECK1-NEXT:    [[DIV143:%.*]] = sdiv i64 [[SUB142]], 1
+// CHECK1-NEXT:    [[MUL144:%.*]] = mul nsw i64 1, [[DIV143]]
+// CHECK1-NEXT:    [[TMP60:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB145:%.*]] = sub nsw i64 [[TMP60]], 0
+// CHECK1-NEXT:    [[DIV146:%.*]] = sdiv i64 [[SUB145]], 1
+// CHECK1-NEXT:    [[MUL147:%.*]] = mul nsw i64 [[MUL144]], [[DIV146]]
+// CHECK1-NEXT:    [[MUL148:%.*]] = mul nsw i64 [[MUL147]], 128
+// CHECK1-NEXT:    [[DIV149:%.*]] = sdiv i64 [[TMP58]], [[MUL148]]
+// CHECK1-NEXT:    [[TMP61:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB150:%.*]] = sub nsw i64 [[TMP61]], 0
+// CHECK1-NEXT:    [[DIV151:%.*]] = sdiv i64 [[SUB150]], 1
+// CHECK1-NEXT:    [[MUL152:%.*]] = mul nsw i64 1, [[DIV151]]
+// CHECK1-NEXT:    [[TMP62:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB153:%.*]] = sub nsw i64 [[TMP62]], 0
+// CHECK1-NEXT:    [[DIV154:%.*]] = sdiv i64 [[SUB153]], 1
+// CHECK1-NEXT:    [[MUL155:%.*]] = mul nsw i64 [[MUL152]], [[DIV154]]
+// CHECK1-NEXT:    [[MUL156:%.*]] = mul nsw i64 [[MUL155]], 128
+// CHECK1-NEXT:    [[MUL157:%.*]] = mul nsw i64 [[DIV149]], [[MUL156]]
+// CHECK1-NEXT:    [[SUB158:%.*]] = sub nsw i64 [[TMP57]], [[MUL157]]
+// CHECK1-NEXT:    [[TMP63:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB159:%.*]] = sub nsw i64 [[TMP63]], 0
+// CHECK1-NEXT:    [[DIV160:%.*]] = sdiv i64 [[SUB159]], 1
+// CHECK1-NEXT:    [[MUL161:%.*]] = mul nsw i64 1, [[DIV160]]
+// CHECK1-NEXT:    [[MUL162:%.*]] = mul nsw i64 [[MUL161]], 128
+// CHECK1-NEXT:    [[DIV163:%.*]] = sdiv i64 [[SUB158]], [[MUL162]]
+// CHECK1-NEXT:    [[TMP64:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB164:%.*]] = sub nsw i64 [[TMP64]], 0
+// CHECK1-NEXT:    [[DIV165:%.*]] = sdiv i64 [[SUB164]], 1
+// CHECK1-NEXT:    [[MUL166:%.*]] = mul nsw i64 1, [[DIV165]]
+// CHECK1-NEXT:    [[MUL167:%.*]] = mul nsw i64 [[MUL166]], 128
+// CHECK1-NEXT:    [[MUL168:%.*]] = mul nsw i64 [[DIV163]], [[MUL167]]
+// CHECK1-NEXT:    [[SUB169:%.*]] = sub nsw i64 [[SUB141]], [[MUL168]]
+// CHECK1-NEXT:    [[TMP65:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP66:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP67:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB170:%.*]] = sub nsw i64 [[TMP67]], 0
+// CHECK1-NEXT:    [[DIV171:%.*]] = sdiv i64 [[SUB170]], 1
+// CHECK1-NEXT:    [[MUL172:%.*]] = mul nsw i64 1, [[DIV171]]
+// CHECK1-NEXT:    [[TMP68:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB173:%.*]] = sub nsw i64 [[TMP68]], 0
+// CHECK1-NEXT:    [[DIV174:%.*]] = sdiv i64 [[SUB173]], 1
+// CHECK1-NEXT:    [[MUL175:%.*]] = mul nsw i64 [[MUL172]], [[DIV174]]
+// CHECK1-NEXT:    [[MUL176:%.*]] = mul nsw i64 [[MUL175]], 128
+// CHECK1-NEXT:    [[DIV177:%.*]] = sdiv i64 [[TMP66]], [[MUL176]]
+// CHECK1-NEXT:    [[TMP69:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB178:%.*]] = sub nsw i64 [[TMP69]], 0
+// CHECK1-NEXT:    [[DIV179:%.*]] = sdiv i64 [[SUB178]], 1
+// CHECK1-NEXT:    [[MUL180:%.*]] = mul nsw i64 1, [[DIV179]]
+// CHECK1-NEXT:    [[TMP70:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB181:%.*]] = sub nsw i64 [[TMP70]], 0
+// CHECK1-NEXT:    [[DIV182:%.*]] = sdiv i64 [[SUB181]], 1
+// CHECK1-NEXT:    [[MUL183:%.*]] = mul nsw i64 [[MUL180]], [[DIV182]]
+// CHECK1-NEXT:    [[MUL184:%.*]] = mul nsw i64 [[MUL183]], 128
+// CHECK1-NEXT:    [[MUL185:%.*]] = mul nsw i64 [[DIV177]], [[MUL184]]
+// CHECK1-NEXT:    [[SUB186:%.*]] = sub nsw i64 [[TMP65]], [[MUL185]]
+// CHECK1-NEXT:    [[TMP71:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP72:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[TMP73:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB187:%.*]] = sub nsw i64 [[TMP73]], 0
+// CHECK1-NEXT:    [[DIV188:%.*]] = sdiv i64 [[SUB187]], 1
+// CHECK1-NEXT:    [[MUL189:%.*]] = mul nsw i64 1, [[DIV188]]
+// CHECK1-NEXT:    [[TMP74:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB190:%.*]] = sub nsw i64 [[TMP74]], 0
+// CHECK1-NEXT:    [[DIV191:%.*]] = sdiv i64 [[SUB190]], 1
+// CHECK1-NEXT:    [[MUL192:%.*]] = mul nsw i64 [[MUL189]], [[DIV191]]
+// CHECK1-NEXT:    [[MUL193:%.*]] = mul nsw i64 [[MUL192]], 128
+// CHECK1-NEXT:    [[DIV194:%.*]] = sdiv i64 [[TMP72]], [[MUL193]]
+// CHECK1-NEXT:    [[TMP75:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK1-NEXT:    [[SUB195:%.*]] = sub nsw i64 [[TMP75]], 0
+// CHECK1-NEXT:    [[DIV196:%.*]] = sdiv i64 [[SUB195]], 1
+// CHECK1-NEXT:    [[MUL197:%.*]] = mul nsw i64 1, [[DIV196]]
+// CHECK1-NEXT:    [[TMP76:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB198:%.*]] = sub nsw i64 [[TMP76]], 0
+// CHECK1-NEXT:    [[DIV199:%.*]] = sdiv i64 [[SUB198]], 1
+// CHECK1-NEXT:    [[MUL200:%.*]] = mul nsw i64 [[MUL197]], [[DIV199]]
+// CHECK1-NEXT:    [[MUL201:%.*]] = mul nsw i64 [[MUL200]], 128
+// CHECK1-NEXT:    [[MUL202:%.*]] = mul nsw i64 [[DIV194]], [[MUL201]]
+// CHECK1-NEXT:    [[SUB203:%.*]] = sub nsw i64 [[TMP71]], [[MUL202]]
+// CHECK1-NEXT:    [[TMP77:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB204:%.*]] = sub nsw i64 [[TMP77]], 0
+// CHECK1-NEXT:    [[DIV205:%.*]] = sdiv i64 [[SUB204]], 1
+// CHECK1-NEXT:    [[MUL206:%.*]] = mul nsw i64 1, [[DIV205]]
+// CHECK1-NEXT:    [[MUL207:%.*]] = mul nsw i64 [[MUL206]], 128
+// CHECK1-NEXT:    [[DIV208:%.*]] = sdiv i64 [[SUB203]], [[MUL207]]
+// CHECK1-NEXT:    [[TMP78:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK1-NEXT:    [[SUB209:%.*]] = sub nsw i64 [[TMP78]], 0
+// CHECK1-NEXT:    [[DIV210:%.*]] = sdiv i64 [[SUB209]], 1
+// CHECK1-NEXT:    [[MUL211:%.*]] = mul nsw i64 1, [[DIV210]]
+// CHECK1-NEXT:    [[MUL212:%.*]] = mul nsw i64 [[MUL211]], 128
+// CHECK1-NEXT:    [[MUL213:%.*]] = mul nsw i64 [[DIV208]], [[MUL212]]
+// CHECK1-NEXT:    [[SUB214:%.*]] = sub nsw i64 [[SUB186]], [[MUL213]]
+// CHECK1-NEXT:    [[DIV215:%.*]] = sdiv i64 [[SUB214]], 128
+// CHECK1-NEXT:    [[MUL216:%.*]] = mul nsw i64 [[DIV215]], 128
+// CHECK1-NEXT:    [[SUB217:%.*]] = sub nsw i64 [[SUB169]], [[MUL216]]
+// CHECK1-NEXT:    [[MUL218:%.*]] = mul nsw i64 [[SUB217]], 1
+// CHECK1-NEXT:    [[ADD219:%.*]] = add nsw i64 0, [[MUL218]]
+// CHECK1-NEXT:    [[CONV220:%.*]] = trunc i64 [[ADD219]] to i32
+// CHECK1-NEXT:    store i32 [[CONV220]], ptr [[J40]], align 4
+// CHECK1-NEXT:    [[TMP79:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_12]], align 8
+// CHECK1-NEXT:    [[TMP80:%.*]] = load i64, ptr [[DOTPERMUTED_0_IV___BEGIN438]], align 8
+// CHECK1-NEXT:    [[MUL221:%.*]] = mul nsw i64 [[TMP80]], 1
+// CHECK1-NEXT:    [[ADD_PTR222:%.*]] = getelementptr inbounds double, ptr [[TMP79]], i64 [[MUL221]]
+// CHECK1-NEXT:    store ptr [[ADD_PTR222]], ptr [[__BEGIN4]], align 8
+// CHECK1-NEXT:    [[TMP81:%.*]] = load ptr, ptr [[__BEGIN4]], align 8
+// CHECK1-NEXT:    store ptr [[TMP81]], ptr [[BB]], align 8
+// CHECK1-NEXT:    [[TMP82:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK1-NEXT:    [[TMP83:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN339]], align 8
+// CHECK1-NEXT:    [[MUL223:%.*]] = mul nsw i64 [[TMP83]], 1
+// CHECK1-NEXT:    [[ADD_PTR224:%.*]] = getelementptr inbounds double, ptr [[TMP82]], i64 [[MUL223]]
+// CHECK1-NEXT:    store ptr [[ADD_PTR224]], ptr [[__BEGIN3]], align 8
+// CHECK1-NEXT:    [[TMP84:%.*]] = load ptr, ptr [[__BEGIN3]], align 8
+// CHECK1-NEXT:    [[TMP85:%.*]] = load double, ptr [[TMP84]], align 8
+// CHECK1-NEXT:    store double [[TMP85]], ptr [[AA]], align 8
+// CHECK1-NEXT:    [[TMP86:%.*]] = load i32, ptr [[I37]], align 4
+// CHECK1-NEXT:    [[TMP87:%.*]] = load double, ptr [[C]], align 8
+// CHECK1-NEXT:    [[TMP88:%.*]] = load double, ptr [[AA]], align 8
+// CHECK1-NEXT:    [[TMP89:%.*]] = load double, ptr [[D]], align 8
+// CHECK1-NEXT:    [[TMP90:%.*]] = load ptr, ptr [[BB]], align 8
+// CHECK1-NEXT:    [[TMP91:%.*]] = load double, ptr [[TMP90]], align 8
+// CHECK1-NEXT:    [[TMP92:%.*]] = load i32, ptr [[J40]], align 4
+// CHECK1-NEXT:    call void (...) @body(i32 noundef [[TMP86]], double noundef [[TMP87]], double noundef [[TMP88]], double noundef [[TMP89]], double noundef [[TMP91]], i32 noundef [[TMP92]])
+// 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:    [[TMP93:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    [[ADD225:%.*]] = add nsw i64 [[TMP93]], 1
+// CHECK1-NEXT:    store i64 [[ADD225]], ptr [[DOTOMP_IV]], align 8
+// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK1:       omp.inner.for.end:
+// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK1:       omp.loop.exit:
+// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK1:       omp.precond.end:
+// CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3]], i32 [[TMP0]])
+// CHECK1-NEXT:    ret void
+//
+//
+// CHECK2-LABEL: define {{[^@]+}}@body
+// CHECK2-SAME: (...) #[[ATTR0:[0-9]+]] {
+// CHECK2-NEXT:  entry:
+// CHECK2-NEXT:    ret void
+//
+//
+// CHECK2-LABEL: define {{[^@]+}}@foo10
+// CHECK2-SAME: () #[[ATTR0]] {
+// CHECK2-NEXT:  entry:
+// CHECK2-NEXT:    [[A:%.*]] = alloca [128 x double], align 16
+// CHECK2-NEXT:    [[B:%.*]] = alloca [16 x double], align 16
+// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[C:%.*]] = alloca double, align 8
+// CHECK2-NEXT:    [[__RANGE3:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[__END3:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[__BEGIN3:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_7:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[D:%.*]] = alloca double, align 8
+// CHECK2-NEXT:    [[__RANGE4:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[__END4:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[__BEGIN4:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_12:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_14:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_15:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_28:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_0_IV___BEGIN4:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTPERMUTED_1_IV___BEGIN3:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[I37:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_0_IV___BEGIN438:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[DOTPERMUTED_1_IV___BEGIN339:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[J40:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[BB:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[AA:%.*]] = alloca double, align 8
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2:[0-9]+]])
+// CHECK2-NEXT:    store double 4.200000e+01, ptr [[C]], align 8
+// CHECK2-NEXT:    store ptr [[A]], ptr [[__RANGE3]], align 8
+// CHECK2-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[__RANGE3]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP1]], i64 0, i64 0
+// CHECK2-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds double, ptr [[ARRAYDECAY]], i64 128
+// CHECK2-NEXT:    store ptr [[ADD_PTR]], ptr [[__END3]], align 8
+// CHECK2-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[__RANGE3]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY4:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP2]], i64 0, i64 0
+// CHECK2-NEXT:    store ptr [[ARRAYDECAY4]], ptr [[__BEGIN3]], align 8
+// CHECK2-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[__RANGE3]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY5:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP3]], i64 0, i64 0
+// CHECK2-NEXT:    store ptr [[ARRAYDECAY5]], ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK2-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[__END3]], align 8
+// CHECK2-NEXT:    store ptr [[TMP4]], ptr [[DOTCAPTURE_EXPR_6]], align 8
+// CHECK2-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_6]], align 8
+// CHECK2-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK2-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[TMP5]] to i64
+// CHECK2-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[TMP6]] to i64
+// CHECK2-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
+// CHECK2-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 8
+// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
+// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
+// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
+// CHECK2-NEXT:    [[SUB8:%.*]] = sub nsw i64 [[DIV]], 1
+// CHECK2-NEXT:    store i64 [[SUB8]], ptr [[DOTCAPTURE_EXPR_7]], align 8
+// CHECK2-NEXT:    store double 4.200000e+01, ptr [[D]], align 8
+// CHECK2-NEXT:    store ptr [[B]], ptr [[__RANGE4]], align 8
+// CHECK2-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[__RANGE4]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY9:%.*]] = getelementptr inbounds [16 x double], ptr [[TMP7]], i64 0, i64 0
+// CHECK2-NEXT:    [[ADD_PTR10:%.*]] = getelementptr inbounds double, ptr [[ARRAYDECAY9]], i64 16
+// CHECK2-NEXT:    store ptr [[ADD_PTR10]], ptr [[__END4]], align 8
+// CHECK2-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[__RANGE4]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY11:%.*]] = getelementptr inbounds [16 x double], ptr [[TMP8]], i64 0, i64 0
+// CHECK2-NEXT:    store ptr [[ARRAYDECAY11]], ptr [[__BEGIN4]], align 8
+// CHECK2-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[__RANGE4]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY13:%.*]] = getelementptr inbounds [16 x double], ptr [[TMP9]], i64 0, i64 0
+// CHECK2-NEXT:    store ptr [[ARRAYDECAY13]], ptr [[DOTCAPTURE_EXPR_12]], align 8
+// CHECK2-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[__END4]], align 8
+// CHECK2-NEXT:    store ptr [[TMP10]], ptr [[DOTCAPTURE_EXPR_14]], align 8
+// CHECK2-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_14]], align 8
+// CHECK2-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_12]], align 8
+// CHECK2-NEXT:    [[SUB_PTR_LHS_CAST16:%.*]] = ptrtoint ptr [[TMP11]] to i64
+// CHECK2-NEXT:    [[SUB_PTR_RHS_CAST17:%.*]] = ptrtoint ptr [[TMP12]] to i64
+// CHECK2-NEXT:    [[SUB_PTR_SUB18:%.*]] = sub i64 [[SUB_PTR_LHS_CAST16]], [[SUB_PTR_RHS_CAST17]]
+// CHECK2-NEXT:    [[SUB_PTR_DIV19:%.*]] = sdiv exact i64 [[SUB_PTR_SUB18]], 8
+// CHECK2-NEXT:    [[SUB20:%.*]] = sub nsw i64 [[SUB_PTR_DIV19]], 1
+// CHECK2-NEXT:    [[ADD21:%.*]] = add nsw i64 [[SUB20]], 1
+// CHECK2-NEXT:    [[DIV22:%.*]] = sdiv i64 [[ADD21]], 1
+// CHECK2-NEXT:    [[SUB23:%.*]] = sub nsw i64 [[DIV22]], 1
+// CHECK2-NEXT:    store i64 [[SUB23]], ptr [[DOTCAPTURE_EXPR_15]], align 8
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_15]], align 8
+// CHECK2-NEXT:    [[ADD25:%.*]] = add nsw i64 [[TMP13]], 1
+// CHECK2-NEXT:    store i64 [[ADD25]], ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_7]], align 8
+// CHECK2-NEXT:    [[ADD27:%.*]] = add nsw i64 [[TMP14]], 1
+// CHECK2-NEXT:    store i64 [[ADD27]], ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB29:%.*]] = sub nsw i64 [[TMP15]], 0
+// CHECK2-NEXT:    [[DIV30:%.*]] = sdiv i64 [[SUB29]], 1
+// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i64 128, [[DIV30]]
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB31:%.*]] = sub nsw i64 [[TMP16]], 0
+// CHECK2-NEXT:    [[DIV32:%.*]] = sdiv i64 [[SUB31]], 1
+// CHECK2-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[MUL]], [[DIV32]]
+// CHECK2-NEXT:    [[MUL34:%.*]] = mul nsw i64 [[MUL33]], 128
+// CHECK2-NEXT:    [[SUB35:%.*]] = sub nsw i64 [[MUL34]], 1
+// CHECK2-NEXT:    store i64 [[SUB35]], ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK2-NEXT:    store i64 0, ptr [[DOTPERMUTED_0_IV___BEGIN4]], align 8
+// CHECK2-NEXT:    store i64 0, ptr [[DOTPERMUTED_1_IV___BEGIN3]], align 8
+// CHECK2-NEXT:    store i32 0, ptr [[J]], align 4
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i64 0, [[TMP17]]
+// CHECK2-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
+// CHECK2:       land.lhs.true:
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[CMP36:%.*]] = icmp slt i64 0, [[TMP18]]
+// CHECK2-NEXT:    br i1 [[CMP36]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
+// CHECK2:       omp.precond.then:
+// CHECK2-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
+// CHECK2-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK2-NEXT:    store i64 [[TMP19]], ptr [[DOTOMP_UB]], align 8
+// CHECK2-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
+// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
+// CHECK2-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// CHECK2-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK2-NEXT:    [[CMP41:%.*]] = icmp sgt i64 [[TMP20]], [[TMP21]]
+// CHECK2-NEXT:    br i1 [[CMP41]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK2:       cond.true:
+// CHECK2-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_28]], align 8
+// CHECK2-NEXT:    br label [[COND_END:%.*]]
+// CHECK2:       cond.false:
+// CHECK2-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// CHECK2-NEXT:    br label [[COND_END]]
+// CHECK2:       cond.end:
+// CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP22]], [[COND_TRUE]] ], [ [[TMP23]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
+// CHECK2-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
+// CHECK2-NEXT:    store i64 [[TMP24]], ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2:       omp.inner.for.cond:
+// CHECK2-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
+// CHECK2-NEXT:    [[CMP42:%.*]] = icmp sle i64 [[TMP25]], [[TMP26]]
+// CHECK2-NEXT:    br i1 [[CMP42]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2:       omp.inner.for.body:
+// CHECK2-NEXT:    [[TMP27:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP28:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB43:%.*]] = sub nsw i64 [[TMP28]], 0
+// CHECK2-NEXT:    [[DIV44:%.*]] = sdiv i64 [[SUB43]], 1
+// CHECK2-NEXT:    [[MUL45:%.*]] = mul nsw i64 1, [[DIV44]]
+// CHECK2-NEXT:    [[TMP29:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB46:%.*]] = sub nsw i64 [[TMP29]], 0
+// CHECK2-NEXT:    [[DIV47:%.*]] = sdiv i64 [[SUB46]], 1
+// CHECK2-NEXT:    [[MUL48:%.*]] = mul nsw i64 [[MUL45]], [[DIV47]]
+// CHECK2-NEXT:    [[MUL49:%.*]] = mul nsw i64 [[MUL48]], 128
+// CHECK2-NEXT:    [[DIV50:%.*]] = sdiv i64 [[TMP27]], [[MUL49]]
+// CHECK2-NEXT:    [[MUL51:%.*]] = mul nsw i64 [[DIV50]], 1
+// CHECK2-NEXT:    [[ADD52:%.*]] = add nsw i64 0, [[MUL51]]
+// CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[ADD52]] to i32
+// CHECK2-NEXT:    store i32 [[CONV]], ptr [[I37]], align 4
+// CHECK2-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP31:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP32:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB53:%.*]] = sub nsw i64 [[TMP32]], 0
+// CHECK2-NEXT:    [[DIV54:%.*]] = sdiv i64 [[SUB53]], 1
+// CHECK2-NEXT:    [[MUL55:%.*]] = mul nsw i64 1, [[DIV54]]
+// CHECK2-NEXT:    [[TMP33:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB56:%.*]] = sub nsw i64 [[TMP33]], 0
+// CHECK2-NEXT:    [[DIV57:%.*]] = sdiv i64 [[SUB56]], 1
+// CHECK2-NEXT:    [[MUL58:%.*]] = mul nsw i64 [[MUL55]], [[DIV57]]
+// CHECK2-NEXT:    [[MUL59:%.*]] = mul nsw i64 [[MUL58]], 128
+// CHECK2-NEXT:    [[DIV60:%.*]] = sdiv i64 [[TMP31]], [[MUL59]]
+// CHECK2-NEXT:    [[TMP34:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB61:%.*]] = sub nsw i64 [[TMP34]], 0
+// CHECK2-NEXT:    [[DIV62:%.*]] = sdiv i64 [[SUB61]], 1
+// CHECK2-NEXT:    [[MUL63:%.*]] = mul nsw i64 1, [[DIV62]]
+// CHECK2-NEXT:    [[TMP35:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB64:%.*]] = sub nsw i64 [[TMP35]], 0
+// CHECK2-NEXT:    [[DIV65:%.*]] = sdiv i64 [[SUB64]], 1
+// CHECK2-NEXT:    [[MUL66:%.*]] = mul nsw i64 [[MUL63]], [[DIV65]]
+// CHECK2-NEXT:    [[MUL67:%.*]] = mul nsw i64 [[MUL66]], 128
+// CHECK2-NEXT:    [[MUL68:%.*]] = mul nsw i64 [[DIV60]], [[MUL67]]
+// CHECK2-NEXT:    [[SUB69:%.*]] = sub nsw i64 [[TMP30]], [[MUL68]]
+// CHECK2-NEXT:    [[TMP36:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB70:%.*]] = sub nsw i64 [[TMP36]], 0
+// CHECK2-NEXT:    [[DIV71:%.*]] = sdiv i64 [[SUB70]], 1
+// CHECK2-NEXT:    [[MUL72:%.*]] = mul nsw i64 1, [[DIV71]]
+// CHECK2-NEXT:    [[MUL73:%.*]] = mul nsw i64 [[MUL72]], 128
+// CHECK2-NEXT:    [[DIV74:%.*]] = sdiv i64 [[SUB69]], [[MUL73]]
+// CHECK2-NEXT:    [[MUL75:%.*]] = mul nsw i64 [[DIV74]], 1
+// CHECK2-NEXT:    [[ADD76:%.*]] = add nsw i64 0, [[MUL75]]
+// CHECK2-NEXT:    store i64 [[ADD76]], ptr [[DOTPERMUTED_0_IV___BEGIN438]], align 8
+// CHECK2-NEXT:    [[TMP37:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP38:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP39:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB77:%.*]] = sub nsw i64 [[TMP39]], 0
+// CHECK2-NEXT:    [[DIV78:%.*]] = sdiv i64 [[SUB77]], 1
+// CHECK2-NEXT:    [[MUL79:%.*]] = mul nsw i64 1, [[DIV78]]
+// CHECK2-NEXT:    [[TMP40:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB80:%.*]] = sub nsw i64 [[TMP40]], 0
+// CHECK2-NEXT:    [[DIV81:%.*]] = sdiv i64 [[SUB80]], 1
+// CHECK2-NEXT:    [[MUL82:%.*]] = mul nsw i64 [[MUL79]], [[DIV81]]
+// CHECK2-NEXT:    [[MUL83:%.*]] = mul nsw i64 [[MUL82]], 128
+// CHECK2-NEXT:    [[DIV84:%.*]] = sdiv i64 [[TMP38]], [[MUL83]]
+// CHECK2-NEXT:    [[TMP41:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB85:%.*]] = sub nsw i64 [[TMP41]], 0
+// CHECK2-NEXT:    [[DIV86:%.*]] = sdiv i64 [[SUB85]], 1
+// CHECK2-NEXT:    [[MUL87:%.*]] = mul nsw i64 1, [[DIV86]]
+// CHECK2-NEXT:    [[TMP42:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB88:%.*]] = sub nsw i64 [[TMP42]], 0
+// CHECK2-NEXT:    [[DIV89:%.*]] = sdiv i64 [[SUB88]], 1
+// CHECK2-NEXT:    [[MUL90:%.*]] = mul nsw i64 [[MUL87]], [[DIV89]]
+// CHECK2-NEXT:    [[MUL91:%.*]] = mul nsw i64 [[MUL90]], 128
+// CHECK2-NEXT:    [[MUL92:%.*]] = mul nsw i64 [[DIV84]], [[MUL91]]
+// CHECK2-NEXT:    [[SUB93:%.*]] = sub nsw i64 [[TMP37]], [[MUL92]]
+// CHECK2-NEXT:    [[TMP43:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP44:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP45:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB94:%.*]] = sub nsw i64 [[TMP45]], 0
+// CHECK2-NEXT:    [[DIV95:%.*]] = sdiv i64 [[SUB94]], 1
+// CHECK2-NEXT:    [[MUL96:%.*]] = mul nsw i64 1, [[DIV95]]
+// CHECK2-NEXT:    [[TMP46:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB97:%.*]] = sub nsw i64 [[TMP46]], 0
+// CHECK2-NEXT:    [[DIV98:%.*]] = sdiv i64 [[SUB97]], 1
+// CHECK2-NEXT:    [[MUL99:%.*]] = mul nsw i64 [[MUL96]], [[DIV98]]
+// CHECK2-NEXT:    [[MUL100:%.*]] = mul nsw i64 [[MUL99]], 128
+// CHECK2-NEXT:    [[DIV101:%.*]] = sdiv i64 [[TMP44]], [[MUL100]]
+// CHECK2-NEXT:    [[TMP47:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB102:%.*]] = sub nsw i64 [[TMP47]], 0
+// CHECK2-NEXT:    [[DIV103:%.*]] = sdiv i64 [[SUB102]], 1
+// CHECK2-NEXT:    [[MUL104:%.*]] = mul nsw i64 1, [[DIV103]]
+// CHECK2-NEXT:    [[TMP48:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB105:%.*]] = sub nsw i64 [[TMP48]], 0
+// CHECK2-NEXT:    [[DIV106:%.*]] = sdiv i64 [[SUB105]], 1
+// CHECK2-NEXT:    [[MUL107:%.*]] = mul nsw i64 [[MUL104]], [[DIV106]]
+// CHECK2-NEXT:    [[MUL108:%.*]] = mul nsw i64 [[MUL107]], 128
+// CHECK2-NEXT:    [[MUL109:%.*]] = mul nsw i64 [[DIV101]], [[MUL108]]
+// CHECK2-NEXT:    [[SUB110:%.*]] = sub nsw i64 [[TMP43]], [[MUL109]]
+// CHECK2-NEXT:    [[TMP49:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB111:%.*]] = sub nsw i64 [[TMP49]], 0
+// CHECK2-NEXT:    [[DIV112:%.*]] = sdiv i64 [[SUB111]], 1
+// CHECK2-NEXT:    [[MUL113:%.*]] = mul nsw i64 1, [[DIV112]]
+// CHECK2-NEXT:    [[MUL114:%.*]] = mul nsw i64 [[MUL113]], 128
+// CHECK2-NEXT:    [[DIV115:%.*]] = sdiv i64 [[SUB110]], [[MUL114]]
+// CHECK2-NEXT:    [[TMP50:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB116:%.*]] = sub nsw i64 [[TMP50]], 0
+// CHECK2-NEXT:    [[DIV117:%.*]] = sdiv i64 [[SUB116]], 1
+// CHECK2-NEXT:    [[MUL118:%.*]] = mul nsw i64 1, [[DIV117]]
+// CHECK2-NEXT:    [[MUL119:%.*]] = mul nsw i64 [[MUL118]], 128
+// CHECK2-NEXT:    [[MUL120:%.*]] = mul nsw i64 [[DIV115]], [[MUL119]]
+// CHECK2-NEXT:    [[SUB121:%.*]] = sub nsw i64 [[SUB93]], [[MUL120]]
+// CHECK2-NEXT:    [[DIV122:%.*]] = sdiv i64 [[SUB121]], 128
+// CHECK2-NEXT:    [[MUL123:%.*]] = mul nsw i64 [[DIV122]], 1
+// CHECK2-NEXT:    [[ADD124:%.*]] = add nsw i64 0, [[MUL123]]
+// CHECK2-NEXT:    store i64 [[ADD124]], ptr [[DOTPERMUTED_1_IV___BEGIN339]], align 8
+// CHECK2-NEXT:    [[TMP51:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP52:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP53:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB125:%.*]] = sub nsw i64 [[TMP53]], 0
+// CHECK2-NEXT:    [[DIV126:%.*]] = sdiv i64 [[SUB125]], 1
+// CHECK2-NEXT:    [[MUL127:%.*]] = mul nsw i64 1, [[DIV126]]
+// CHECK2-NEXT:    [[TMP54:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB128:%.*]] = sub nsw i64 [[TMP54]], 0
+// CHECK2-NEXT:    [[DIV129:%.*]] = sdiv i64 [[SUB128]], 1
+// CHECK2-NEXT:    [[MUL130:%.*]] = mul nsw i64 [[MUL127]], [[DIV129]]
+// CHECK2-NEXT:    [[MUL131:%.*]] = mul nsw i64 [[MUL130]], 128
+// CHECK2-NEXT:    [[DIV132:%.*]] = sdiv i64 [[TMP52]], [[MUL131]]
+// CHECK2-NEXT:    [[TMP55:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB133:%.*]] = sub nsw i64 [[TMP55]], 0
+// CHECK2-NEXT:    [[DIV134:%.*]] = sdiv i64 [[SUB133]], 1
+// CHECK2-NEXT:    [[MUL135:%.*]] = mul nsw i64 1, [[DIV134]]
+// CHECK2-NEXT:    [[TMP56:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB136:%.*]] = sub nsw i64 [[TMP56]], 0
+// CHECK2-NEXT:    [[DIV137:%.*]] = sdiv i64 [[SUB136]], 1
+// CHECK2-NEXT:    [[MUL138:%.*]] = mul nsw i64 [[MUL135]], [[DIV137]]
+// CHECK2-NEXT:    [[MUL139:%.*]] = mul nsw i64 [[MUL138]], 128
+// CHECK2-NEXT:    [[MUL140:%.*]] = mul nsw i64 [[DIV132]], [[MUL139]]
+// CHECK2-NEXT:    [[SUB141:%.*]] = sub nsw i64 [[TMP51]], [[MUL140]]
+// CHECK2-NEXT:    [[TMP57:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP58:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP59:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB142:%.*]] = sub nsw i64 [[TMP59]], 0
+// CHECK2-NEXT:    [[DIV143:%.*]] = sdiv i64 [[SUB142]], 1
+// CHECK2-NEXT:    [[MUL144:%.*]] = mul nsw i64 1, [[DIV143]]
+// CHECK2-NEXT:    [[TMP60:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB145:%.*]] = sub nsw i64 [[TMP60]], 0
+// CHECK2-NEXT:    [[DIV146:%.*]] = sdiv i64 [[SUB145]], 1
+// CHECK2-NEXT:    [[MUL147:%.*]] = mul nsw i64 [[MUL144]], [[DIV146]]
+// CHECK2-NEXT:    [[MUL148:%.*]] = mul nsw i64 [[MUL147]], 128
+// CHECK2-NEXT:    [[DIV149:%.*]] = sdiv i64 [[TMP58]], [[MUL148]]
+// CHECK2-NEXT:    [[TMP61:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB150:%.*]] = sub nsw i64 [[TMP61]], 0
+// CHECK2-NEXT:    [[DIV151:%.*]] = sdiv i64 [[SUB150]], 1
+// CHECK2-NEXT:    [[MUL152:%.*]] = mul nsw i64 1, [[DIV151]]
+// CHECK2-NEXT:    [[TMP62:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB153:%.*]] = sub nsw i64 [[TMP62]], 0
+// CHECK2-NEXT:    [[DIV154:%.*]] = sdiv i64 [[SUB153]], 1
+// CHECK2-NEXT:    [[MUL155:%.*]] = mul nsw i64 [[MUL152]], [[DIV154]]
+// CHECK2-NEXT:    [[MUL156:%.*]] = mul nsw i64 [[MUL155]], 128
+// CHECK2-NEXT:    [[MUL157:%.*]] = mul nsw i64 [[DIV149]], [[MUL156]]
+// CHECK2-NEXT:    [[SUB158:%.*]] = sub nsw i64 [[TMP57]], [[MUL157]]
+// CHECK2-NEXT:    [[TMP63:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB159:%.*]] = sub nsw i64 [[TMP63]], 0
+// CHECK2-NEXT:    [[DIV160:%.*]] = sdiv i64 [[SUB159]], 1
+// CHECK2-NEXT:    [[MUL161:%.*]] = mul nsw i64 1, [[DIV160]]
+// CHECK2-NEXT:    [[MUL162:%.*]] = mul nsw i64 [[MUL161]], 128
+// CHECK2-NEXT:    [[DIV163:%.*]] = sdiv i64 [[SUB158]], [[MUL162]]
+// CHECK2-NEXT:    [[TMP64:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB164:%.*]] = sub nsw i64 [[TMP64]], 0
+// CHECK2-NEXT:    [[DIV165:%.*]] = sdiv i64 [[SUB164]], 1
+// CHECK2-NEXT:    [[MUL166:%.*]] = mul nsw i64 1, [[DIV165]]
+// CHECK2-NEXT:    [[MUL167:%.*]] = mul nsw i64 [[MUL166]], 128
+// CHECK2-NEXT:    [[MUL168:%.*]] = mul nsw i64 [[DIV163]], [[MUL167]]
+// CHECK2-NEXT:    [[SUB169:%.*]] = sub nsw i64 [[SUB141]], [[MUL168]]
+// CHECK2-NEXT:    [[TMP65:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP66:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP67:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB170:%.*]] = sub nsw i64 [[TMP67]], 0
+// CHECK2-NEXT:    [[DIV171:%.*]] = sdiv i64 [[SUB170]], 1
+// CHECK2-NEXT:    [[MUL172:%.*]] = mul nsw i64 1, [[DIV171]]
+// CHECK2-NEXT:    [[TMP68:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB173:%.*]] = sub nsw i64 [[TMP68]], 0
+// CHECK2-NEXT:    [[DIV174:%.*]] = sdiv i64 [[SUB173]], 1
+// CHECK2-NEXT:    [[MUL175:%.*]] = mul nsw i64 [[MUL172]], [[DIV174]]
+// CHECK2-NEXT:    [[MUL176:%.*]] = mul nsw i64 [[MUL175]], 128
+// CHECK2-NEXT:    [[DIV177:%.*]] = sdiv i64 [[TMP66]], [[MUL176]]
+// CHECK2-NEXT:    [[TMP69:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB178:%.*]] = sub nsw i64 [[TMP69]], 0
+// CHECK2-NEXT:    [[DIV179:%.*]] = sdiv i64 [[SUB178]], 1
+// CHECK2-NEXT:    [[MUL180:%.*]] = mul nsw i64 1, [[DIV179]]
+// CHECK2-NEXT:    [[TMP70:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB181:%.*]] = sub nsw i64 [[TMP70]], 0
+// CHECK2-NEXT:    [[DIV182:%.*]] = sdiv i64 [[SUB181]], 1
+// CHECK2-NEXT:    [[MUL183:%.*]] = mul nsw i64 [[MUL180]], [[DIV182]]
+// CHECK2-NEXT:    [[MUL184:%.*]] = mul nsw i64 [[MUL183]], 128
+// CHECK2-NEXT:    [[MUL185:%.*]] = mul nsw i64 [[DIV177]], [[MUL184]]
+// CHECK2-NEXT:    [[SUB186:%.*]] = sub nsw i64 [[TMP65]], [[MUL185]]
+// CHECK2-NEXT:    [[TMP71:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP72:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[TMP73:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB187:%.*]] = sub nsw i64 [[TMP73]], 0
+// CHECK2-NEXT:    [[DIV188:%.*]] = sdiv i64 [[SUB187]], 1
+// CHECK2-NEXT:    [[MUL189:%.*]] = mul nsw i64 1, [[DIV188]]
+// CHECK2-NEXT:    [[TMP74:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB190:%.*]] = sub nsw i64 [[TMP74]], 0
+// CHECK2-NEXT:    [[DIV191:%.*]] = sdiv i64 [[SUB190]], 1
+// CHECK2-NEXT:    [[MUL192:%.*]] = mul nsw i64 [[MUL189]], [[DIV191]]
+// CHECK2-NEXT:    [[MUL193:%.*]] = mul nsw i64 [[MUL192]], 128
+// CHECK2-NEXT:    [[DIV194:%.*]] = sdiv i64 [[TMP72]], [[MUL193]]
+// CHECK2-NEXT:    [[TMP75:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_24]], align 8
+// CHECK2-NEXT:    [[SUB195:%.*]] = sub nsw i64 [[TMP75]], 0
+// CHECK2-NEXT:    [[DIV196:%.*]] = sdiv i64 [[SUB195]], 1
+// CHECK2-NEXT:    [[MUL197:%.*]] = mul nsw i64 1, [[DIV196]]
+// CHECK2-NEXT:    [[TMP76:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB198:%.*]] = sub nsw i64 [[TMP76]], 0
+// CHECK2-NEXT:    [[DIV199:%.*]] = sdiv i64 [[SUB198]], 1
+// CHECK2-NEXT:    [[MUL200:%.*]] = mul nsw i64 [[MUL197]], [[DIV199]]
+// CHECK2-NEXT:    [[MUL201:%.*]] = mul nsw i64 [[MUL200]], 128
+// CHECK2-NEXT:    [[MUL202:%.*]] = mul nsw i64 [[DIV194]], [[MUL201]]
+// CHECK2-NEXT:    [[SUB203:%.*]] = sub nsw i64 [[TMP71]], [[MUL202]]
+// CHECK2-NEXT:    [[TMP77:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB204:%.*]] = sub nsw i64 [[TMP77]], 0
+// CHECK2-NEXT:    [[DIV205:%.*]] = sdiv i64 [[SUB204]], 1
+// CHECK2-NEXT:    [[MUL206:%.*]] = mul nsw i64 1, [[DIV205]]
+// CHECK2-NEXT:    [[MUL207:%.*]] = mul nsw i64 [[MUL206]], 128
+// CHECK2-NEXT:    [[DIV208:%.*]] = sdiv i64 [[SUB203]], [[MUL207]]
+// CHECK2-NEXT:    [[TMP78:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_26]], align 8
+// CHECK2-NEXT:    [[SUB209:%.*]] = sub nsw i64 [[TMP78]], 0
+// CHECK2-NEXT:    [[DIV210:%.*]] = sdiv i64 [[SUB209]], 1
+// CHECK2-NEXT:    [[MUL211:%.*]] = mul nsw i64 1, [[DIV210]]
+// CHECK2-NEXT:    [[MUL212:%.*]] = mul nsw i64 [[MUL211]], 128
+// CHECK2-NEXT:    [[MUL213:%.*]] = mul nsw i64 [[DIV208]], [[MUL212]]
+// CHECK2-NEXT:    [[SUB214:%.*]] = sub nsw i64 [[SUB186]], [[MUL213]]
+// CHECK2-NEXT:    [[DIV215:%.*]] = sdiv i64 [[SUB214]], 128
+// CHECK2-NEXT:    [[MUL216:%.*]] = mul nsw i64 [[DIV215]], 128
+// CHECK2-NEXT:    [[SUB217:%.*]] = sub nsw i64 [[SUB169]], [[MUL216]]
+// CHECK2-NEXT:    [[MUL218:%.*]] = mul nsw i64 [[SUB217]], 1
+// CHECK2-NEXT:    [[ADD219:%.*]] = add nsw i64 0, [[MUL218]]
+// CHECK2-NEXT:    [[CONV220:%.*]] = trunc i64 [[ADD219]] to i32
+// CHECK2-NEXT:    store i32 [[CONV220]], ptr [[J40]], align 4
+// CHECK2-NEXT:    [[TMP79:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_12]], align 8
+// CHECK2-NEXT:    [[TMP80:%.*]] = load i64, ptr [[DOTPERMUTED_0_IV___BEGIN438]], align 8
+// CHECK2-NEXT:    [[MUL221:%.*]] = mul nsw i64 [[TMP80]], 1
+// CHECK2-NEXT:    [[ADD_PTR222:%.*]] = getelementptr inbounds double, ptr [[TMP79]], i64 [[MUL221]]
+// CHECK2-NEXT:    store ptr [[ADD_PTR222]], ptr [[__BEGIN4]], align 8
+// CHECK2-NEXT:    [[TMP81:%.*]] = load ptr, ptr [[__BEGIN4]], align 8
+// CHECK2-NEXT:    store ptr [[TMP81]], ptr [[BB]], align 8
+// CHECK2-NEXT:    [[TMP82:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK2-NEXT:    [[TMP83:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN339]], align 8
+// CHECK2-NEXT:    [[MUL223:%.*]] = mul nsw i64 [[TMP83]], 1
+// CHECK2-NEXT:    [[ADD_PTR224:%.*]] = getelementptr inbounds double, ptr [[TMP82]], i64 [[MUL223]]
+// CHECK2-NEXT:    store ptr [[ADD_PTR224]], ptr [[__BEGIN3]], align 8
+// CHECK2-NEXT:    [[TMP84:%.*]] = load ptr, ptr [[__BEGIN3]], align 8
+// CHECK2-NEXT:    [[TMP85:%.*]] = load double, ptr [[TMP84]], align 8
+// CHECK2-NEXT:    store double [[TMP85]], ptr [[AA]], align 8
+// CHECK2-NEXT:    [[TMP86:%.*]] = load i32, ptr [[I37]], align 4
+// CHECK2-NEXT:    [[TMP87:%.*]] = load double, ptr [[C]], align 8
+// CHECK2-NEXT:    [[TMP88:%.*]] = load double, ptr [[AA]], align 8
+// CHECK2-NEXT:    [[TMP89:%.*]] = load double, ptr [[D]], align 8
+// CHECK2-NEXT:    [[TMP90:%.*]] = load ptr, ptr [[BB]], align 8
+// CHECK2-NEXT:    [[TMP91:%.*]] = load double, ptr [[TMP90]], align 8
+// CHECK2-NEXT:    [[TMP92:%.*]] = load i32, ptr [[J40]], align 4
+// CHECK2-NEXT:    call void (...) @body(i32 noundef [[TMP86]], double noundef [[TMP87]], double noundef [[TMP88]], double noundef [[TMP89]], double noundef [[TMP91]], i32 noundef [[TMP92]])
+// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2:       omp.body.continue:
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2:       omp.inner.for.inc:
+// CHECK2-NEXT:    [[TMP93:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    [[ADD225:%.*]] = add nsw i64 [[TMP93]], 1
+// CHECK2-NEXT:    store i64 [[ADD225]], ptr [[DOTOMP_IV]], align 8
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2:       omp.inner.for.end:
+// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK2:       omp.loop.exit:
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
+// CHECK2:       omp.precond.end:
+// CHECK2-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3:[0-9]+]], i32 [[TMP0]])
+// CHECK2-NEXT:    ret void
+//
+//
+// CHECK2-LABEL: define {{[^@]+}}@foo2
+// CHECK2-SAME: (i32 noundef [[START1:%.*]], i32 noundef [[START2:%.*]], i32 noundef [[END1:%.*]], i32 noundef [[END2:%.*]], i32 noundef [[STEP1:%.*]], i32 noundef [[STEP2:%.*]]) #[[ATTR0]] {
+// CHECK2-NEXT:  entry:
+// CHECK2-NEXT:    [[START1_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[START2_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[END1_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[END2_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[STEP1_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[STEP2_ADDR:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTNEW_STEP:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTNEW_STEP7:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_8:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_0_IV_J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_1_IV_I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    store i32 [[START1]], ptr [[START1_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[START2]], ptr [[START2_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[END1]], ptr [[END1_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[END2]], ptr [[END2_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[STEP1]], ptr [[STEP1_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[STEP2]], ptr [[STEP2_ADDR]], align 4
+// CHECK2-NEXT:    [[TMP0:%.*]] = load i32, ptr [[START1_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP0]], ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[START1_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP1]], ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[END1_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[STEP1_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP3]], ptr [[DOTNEW_STEP]], align 4
+// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
+// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], [[TMP5]]
+// CHECK2-NEXT:    [[SUB3:%.*]] = sub i32 [[SUB]], 1
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTNEW_STEP]], align 4
+// CHECK2-NEXT:    [[ADD:%.*]] = add i32 [[SUB3]], [[TMP6]]
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTNEW_STEP]], align 4
+// CHECK2-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], [[TMP7]]
+// CHECK2-NEXT:    [[SUB4:%.*]] = sub i32 [[DIV]], 1
+// CHECK2-NEXT:    store i32 [[SUB4]], ptr [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[START2_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP8]], ptr [[J]], align 4
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[START2_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP9]], ptr [[DOTCAPTURE_EXPR_5]], align 4
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[END2_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP10]], ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[STEP2_ADDR]], align 4
+// CHECK2-NEXT:    store i32 [[TMP11]], ptr [[DOTNEW_STEP7]], align 4
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_6]], align 4
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_5]], align 4
+// CHECK2-NEXT:    [[SUB9:%.*]] = sub i32 [[TMP12]], [[TMP13]]
+// CHECK2-NEXT:    [[SUB10:%.*]] = sub i32 [[SUB9]], 1
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTNEW_STEP7]], align 4
+// CHECK2-NEXT:    [[ADD11:%.*]] = add i32 [[SUB10]], [[TMP14]]
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTNEW_STEP7]], align 4
+// CHECK2-NEXT:    [[DIV12:%.*]] = udiv i32 [[ADD11]], [[TMP15]]
+// CHECK2-NEXT:    [[SUB13:%.*]] = sub i32 [[DIV12]], 1
+// CHECK2-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_8]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK2:       for.cond:
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_8]], align 4
+// CHECK2-NEXT:    [[ADD14:%.*]] = add i32 [[TMP17]], 1
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp ult i32 [[TMP16]], [[ADD14]]
+// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END24:%.*]]
+// CHECK2:       for.body:
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_5]], align 4
+// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTNEW_STEP7]], align 4
+// CHECK2-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], [[TMP20]]
+// CHECK2-NEXT:    [[ADD15:%.*]] = add i32 [[TMP18]], [[MUL]]
+// CHECK2-NEXT:    store i32 [[ADD15]], ptr [[J]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND16:%.*]]
+// CHECK2:       for.cond16:
+// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
+// CHECK2-NEXT:    [[ADD17:%.*]] = add i32 [[TMP22]], 1
+// CHECK2-NEXT:    [[CMP18:%.*]] = icmp ult i32 [[TMP21]], [[ADD17]]
+// CHECK2-NEXT:    br i1 [[CMP18]], label [[FOR_BODY19:%.*]], label [[FOR_END:%.*]]
+// CHECK2:       for.body19:
+// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
+// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTNEW_STEP]], align 4
+// CHECK2-NEXT:    [[MUL20:%.*]] = mul i32 [[TMP24]], [[TMP25]]
+// CHECK2-NEXT:    [[ADD21:%.*]] = add i32 [[TMP23]], [[MUL20]]
+// CHECK2-NEXT:    store i32 [[ADD21]], ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, ptr [[J]], align 4
+// CHECK2-NEXT:    call void (...) @body(i32 noundef [[TMP26]], i32 noundef [[TMP27]])
+// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK2:       for.inc:
+// CHECK2-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[INC:%.*]] = add i32 [[TMP28]], 1
+// CHECK2-NEXT:    store i32 [[INC]], ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND16]], !llvm.loop [[LOOP3:![0-9]+]]
+// CHECK2:       for.end:
+// CHECK2-NEXT:    br label [[FOR_INC22:%.*]]
+// CHECK2:       for.inc22:
+// CHECK2-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    [[INC23:%.*]] = add i32 [[TMP29]], 1
+// CHECK2-NEXT:    store i32 [[INC23]], ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
+// CHECK2:       for.end24:
+// CHECK2-NEXT:    ret void
+//
+//
+// CHECK2-LABEL: define {{[^@]+}}@foo3
+// CHECK2-SAME: () #[[ATTR0]] {
+// CHECK2-NEXT:  entry:
+// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_0_IV_J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_1_IV_I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
+// CHECK2-NEXT:    store i32 7, ptr [[I]], align 4
+// CHECK2-NEXT:    store i32 7, ptr [[J]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    store i32 3, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 3
+// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK2:       cond.true:
+// CHECK2-NEXT:    br label [[COND_END:%.*]]
+// CHECK2:       cond.false:
+// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    br label [[COND_END]]
+// CHECK2:       cond.end:
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2:       omp.inner.for.cond:
+// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
+// CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2:       omp.inner.for.body:
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
+// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK2-NEXT:    store i32 [[ADD]], ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    [[MUL2:%.*]] = mul nsw i32 [[TMP7]], 3
+// CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 7, [[MUL2]]
+// CHECK2-NEXT:    store i32 [[ADD3]], ptr [[J]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK2:       for.cond:
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp slt i32 [[TMP8]], 4
+// CHECK2-NEXT:    br i1 [[CMP4]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// CHECK2:       for.body:
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP9]], 3
+// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 7, [[MUL5]]
+// CHECK2-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[J]], align 4
+// CHECK2-NEXT:    call void (...) @body(i32 noundef [[TMP10]], i32 noundef [[TMP11]])
+// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK2:       for.inc:
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP12]], 1
+// CHECK2-NEXT:    store i32 [[INC]], ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
+// CHECK2:       for.end:
+// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2:       omp.body.continue:
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2:       omp.inner.for.inc:
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP13]], 1
+// CHECK2-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2:       omp.inner.for.end:
+// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK2:       omp.loop.exit:
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    ret void
+//
+//
+// CHECK2-LABEL: define {{[^@]+}}@foo4
+// CHECK2-SAME: () #[[ATTR0]] {
+// CHECK2-NEXT:  entry:
+// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[K:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_0_IV_J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_1_IV_I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
+// CHECK2-NEXT:    store i32 7, ptr [[I]], align 4
+// CHECK2-NEXT:    store i32 7, ptr [[J]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    store i32 15, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 15
+// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK2:       cond.true:
+// CHECK2-NEXT:    br label [[COND_END:%.*]]
+// CHECK2:       cond.false:
+// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    br label [[COND_END]]
+// CHECK2:       cond.end:
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 15, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2:       omp.inner.for.cond:
+// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
+// CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2:       omp.inner.for.body:
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP6]], 4
+// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 3
+// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 7, [[MUL]]
+// CHECK2-NEXT:    store i32 [[ADD]], ptr [[K]], align 4
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP8]], 4
+// CHECK2-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 4
+// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], [[MUL4]]
+// CHECK2-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
+// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
+// CHECK2-NEXT:    store i32 [[ADD6]], ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_J]], align 4
+// CHECK2-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[TMP9]], 3
+// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 7, [[MUL7]]
+// CHECK2-NEXT:    store i32 [[ADD8]], ptr [[J]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK2:       for.cond:
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP10]], 4
+// CHECK2-NEXT:    br i1 [[CMP9]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
+// CHECK2:       for.body:
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP11]], 3
+// CHECK2-NEXT:    [[ADD11:%.*]] = add nsw i32 7, [[MUL10]]
+// CHECK2-NEXT:    store i32 [[ADD11]], ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[J]], align 4
+// CHECK2-NEXT:    call void (...) @body(i32 noundef [[TMP12]], i32 noundef [[TMP13]])
+// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK2:       for.inc:
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP14]], 1
+// CHECK2-NEXT:    store i32 [[INC]], ptr [[DOTPERMUTED_1_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
+// CHECK2:       for.end:
+// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2:       omp.body.continue:
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2:       omp.inner.for.inc:
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP15]], 1
+// CHECK2-NEXT:    store i32 [[ADD12]], ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2:       omp.inner.for.end:
+// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK2:       omp.loop.exit:
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    ret void
+//
+//
+// CHECK2-LABEL: define {{[^@]+}}@foo6
+// CHECK2-SAME: () #[[ATTR0]] {
+// CHECK2-NEXT:  entry:
+// CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[K:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_0_IV_K:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_1_IV_J:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[L:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB2]])
+// CHECK2-NEXT:    store i32 7, ptr [[J]], align 4
+// CHECK2-NEXT:    store i32 7, ptr [[K]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    store i32 255, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
+// CHECK2-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP0]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
+// CHECK2-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 255
+// CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// CHECK2:       cond.true:
+// CHECK2-NEXT:    br label [[COND_END:%.*]]
+// CHECK2:       cond.false:
+// CHECK2-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    br label [[COND_END]]
+// CHECK2:       cond.end:
+// CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 255, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
+// CHECK2-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
+// CHECK2-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// CHECK2:       omp.inner.for.cond:
+// CHECK2-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
+// CHECK2-NEXT:    [[CMP4:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
+// CHECK2-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// CHECK2:       omp.inner.for.body:
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP6]], 64
+// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 3
+// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 7, [[MUL]]
+// CHECK2-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV5:%.*]] = sdiv i32 [[TMP8]], 64
+// CHECK2-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[DIV5]], 64
+// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], [[MUL6]]
+// CHECK2-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB]], 16
+// CHECK2-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
+// CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
+// CHECK2-NEXT:    store i32 [[ADD9]], ptr [[DOTPERMUTED_0_IV_K]], align 4
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV10:%.*]] = sdiv i32 [[TMP10]], 64
+// CHECK2-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 64
+// CHECK2-NEXT:    [[SUB12:%.*]] = sub nsw i32 [[TMP9]], [[MUL11]]
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV13:%.*]] = sdiv i32 [[TMP12]], 64
+// CHECK2-NEXT:    [[MUL14:%.*]] = mul nsw i32 [[DIV13]], 64
+// CHECK2-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP11]], [[MUL14]]
+// CHECK2-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 16
+// CHECK2-NEXT:    [[MUL17:%.*]] = mul nsw i32 [[DIV16]], 16
+// CHECK2-NEXT:    [[SUB18:%.*]] = sub nsw i32 [[SUB12]], [[MUL17]]
+// CHECK2-NEXT:    [[DIV19:%.*]] = sdiv i32 [[SUB18]], 4
+// CHECK2-NEXT:    [[MUL20:%.*]] = mul nsw i32 [[DIV19]], 1
+// CHECK2-NEXT:    [[ADD21:%.*]] = add nsw i32 0, [[MUL20]]
+// CHECK2-NEXT:    store i32 [[ADD21]], ptr [[DOTPERMUTED_1_IV_J]], align 4
+// CHECK2-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV22:%.*]] = sdiv i32 [[TMP14]], 64
+// CHECK2-NEXT:    [[MUL23:%.*]] = mul nsw i32 [[DIV22]], 64
+// CHECK2-NEXT:    [[SUB24:%.*]] = sub nsw i32 [[TMP13]], [[MUL23]]
+// CHECK2-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV25:%.*]] = sdiv i32 [[TMP16]], 64
+// CHECK2-NEXT:    [[MUL26:%.*]] = mul nsw i32 [[DIV25]], 64
+// CHECK2-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP15]], [[MUL26]]
+// CHECK2-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 16
+// CHECK2-NEXT:    [[MUL29:%.*]] = mul nsw i32 [[DIV28]], 16
+// CHECK2-NEXT:    [[SUB30:%.*]] = sub nsw i32 [[SUB24]], [[MUL29]]
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV31:%.*]] = sdiv i32 [[TMP18]], 64
+// CHECK2-NEXT:    [[MUL32:%.*]] = mul nsw i32 [[DIV31]], 64
+// CHECK2-NEXT:    [[SUB33:%.*]] = sub nsw i32 [[TMP17]], [[MUL32]]
+// CHECK2-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[DIV34:%.*]] = sdiv i32 [[TMP20]], 64
+// CHECK2-NEXT:    [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 64
+// CHECK2-NEXT:    [[SUB36:%.*]] = sub nsw i32 [[TMP19]], [[MUL35]]
+// CHECK2-NEXT:    [[DIV37:%.*]] = sdiv i32 [[SUB36]], 16
+// CHECK2-NEXT:    [[MUL38:%.*]] = mul nsw i32 [[DIV37]], 16
+// CHECK2-NEXT:    [[SUB39:%.*]] = sub nsw i32 [[SUB33]], [[MUL38]]
+// CHECK2-NEXT:    [[DIV40:%.*]] = sdiv i32 [[SUB39]], 4
+// CHECK2-NEXT:    [[MUL41:%.*]] = mul nsw i32 [[DIV40]], 4
+// CHECK2-NEXT:    [[SUB42:%.*]] = sub nsw i32 [[SUB30]], [[MUL41]]
+// CHECK2-NEXT:    [[MUL43:%.*]] = mul nsw i32 [[SUB42]], 3
+// CHECK2-NEXT:    [[ADD44:%.*]] = add nsw i32 7, [[MUL43]]
+// CHECK2-NEXT:    store i32 [[ADD44]], ptr [[L]], align 4
+// CHECK2-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_K]], align 4
+// CHECK2-NEXT:    [[MUL45:%.*]] = mul nsw i32 [[TMP21]], 3
+// CHECK2-NEXT:    [[ADD46:%.*]] = add nsw i32 7, [[MUL45]]
+// CHECK2-NEXT:    store i32 [[ADD46]], ptr [[K]], align 4
+// CHECK2-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTPERMUTED_1_IV_J]], align 4
+// CHECK2-NEXT:    [[MUL47:%.*]] = mul nsw i32 [[TMP22]], 3
+// CHECK2-NEXT:    [[ADD48:%.*]] = add nsw i32 7, [[MUL47]]
+// CHECK2-NEXT:    store i32 [[ADD48]], ptr [[J]], align 4
+// CHECK2-NEXT:    [[TMP23:%.*]] = load i32, ptr [[I]], align 4
+// CHECK2-NEXT:    [[TMP24:%.*]] = load i32, ptr [[J]], align 4
+// CHECK2-NEXT:    [[TMP25:%.*]] = load i32, ptr [[K]], align 4
+// CHECK2-NEXT:    [[TMP26:%.*]] = load i32, ptr [[L]], align 4
+// CHECK2-NEXT:    call void (...) @body(i32 noundef [[TMP23]], i32 noundef [[TMP24]], i32 noundef [[TMP25]], i32 noundef [[TMP26]])
+// CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// CHECK2:       omp.body.continue:
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// CHECK2:       omp.inner.for.inc:
+// CHECK2-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    [[ADD49:%.*]] = add nsw i32 [[TMP27]], 1
+// CHECK2-NEXT:    store i32 [[ADD49]], ptr [[DOTOMP_IV]], align 4
+// CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// CHECK2:       omp.inner.for.end:
+// CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// CHECK2:       omp.loop.exit:
+// CHECK2-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP0]])
+// CHECK2-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB3]], i32 [[TMP0]])
+// CHECK2-NEXT:    ret void
+//
+//
+// CHECK2-LABEL: define {{[^@]+}}@foo9
+// CHECK2-SAME: () #[[ATTR0]] {
+// CHECK2-NEXT:  entry:
+// CHECK2-NEXT:    [[ARR:%.*]] = alloca [128 x double], align 16
+// CHECK2-NEXT:    [[C:%.*]] = alloca double, align 8
+// CHECK2-NEXT:    [[__RANGE2:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[__END2:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[__BEGIN2:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_0_IV_I:%.*]] = alloca i32, align 4
+// CHECK2-NEXT:    [[DOTPERMUTED_1_IV___BEGIN2:%.*]] = alloca i64, align 8
+// CHECK2-NEXT:    [[V:%.*]] = alloca ptr, align 8
+// CHECK2-NEXT:    store double 4.200000e+01, ptr [[C]], align 8
+// CHECK2-NEXT:    store ptr [[ARR]], ptr [[__RANGE2]], align 8
+// CHECK2-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[__RANGE2]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP0]], i64 0, i64 0
+// CHECK2-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds double, ptr [[ARRAYDECAY]], i64 128
+// CHECK2-NEXT:    store ptr [[ADD_PTR]], ptr [[__END2]], align 8
+// CHECK2-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[__RANGE2]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY1:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP1]], i64 0, i64 0
+// CHECK2-NEXT:    store ptr [[ARRAYDECAY1]], ptr [[__BEGIN2]], align 8
+// CHECK2-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[__RANGE2]], align 8
+// CHECK2-NEXT:    [[ARRAYDECAY2:%.*]] = getelementptr inbounds [128 x double], ptr [[TMP2]], i64 0, i64 0
+// CHECK2-NEXT:    store ptr [[ARRAYDECAY2]], ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK2-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[__END2]], align 8
+// CHECK2-NEXT:    store ptr [[TMP3]], ptr [[DOTCAPTURE_EXPR_3]], align 8
+// CHECK2-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_3]], align 8
+// CHECK2-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK2-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint ptr [[TMP4]] to i64
+// CHECK2-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint ptr [[TMP5]] to i64
+// CHECK2-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
+// CHECK2-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 8
+// CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
+// CHECK2-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
+// CHECK2-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
+// CHECK2-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[DIV]], 1
+// CHECK2-NEXT:    store i64 [[SUB5]], ptr [[DOTCAPTURE_EXPR_4]], align 8
+// CHECK2-NEXT:    store i32 0, ptr [[I]], align 4
+// CHECK2-NEXT:    store i32 0, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND:%.*]]
+// CHECK2:       for.cond:
+// CHECK2-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], 21
+// CHECK2-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END15:%.*]]
+// CHECK2:       for.body:
+// CHECK2-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 2
+// CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL]]
+// CHECK2-NEXT:    store i32 [[ADD6]], ptr [[I]], align 4
+// CHECK2-NEXT:    store i64 0, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK2-NEXT:    br label [[FOR_COND7:%.*]]
+// CHECK2:       for.cond7:
+// CHECK2-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK2-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_4]], align 8
+// CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i64 [[TMP9]], 1
+// CHECK2-NEXT:    [[CMP9:%.*]] = icmp slt i64 [[TMP8]], [[ADD8]]
+// CHECK2-NEXT:    br i1 [[CMP9]], label [[FOR_BODY10:%.*]], label [[FOR_END:%.*]]
+// CHECK2:       for.body10:
+// CHECK2-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTCAPTURE_EXPR_]], align 8
+// CHECK2-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK2-NEXT:    [[MUL11:%.*]] = mul nsw i64 [[TMP11]], 1
+// CHECK2-NEXT:    [[ADD_PTR12:%.*]] = getelementptr inbounds double, ptr [[TMP10]], i64 [[MUL11]]
+// CHECK2-NEXT:    store ptr [[ADD_PTR12]], ptr [[__BEGIN2]], align 8
+// CHECK2-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[__BEGIN2]], align 8
+// CHECK2-NEXT:    store ptr [[TMP12]], ptr [[V]], align 8
+// CHECK2-NEXT:    [[TMP13:%.*]] = load double, ptr [[C]], align 8
+// CHECK2-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[V]], align 8
+// CHECK2-NEXT:    [[TMP15:%.*]] = load double, ptr [[TMP14]], align 8
+// CHECK2-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4
+// CHECK2-NEXT:    call void (...) @body(double noundef [[TMP13]], double noundef [[TMP15]], i32 noundef [[TMP16]])
+// CHECK2-NEXT:    br label [[FOR_INC:%.*]]
+// CHECK2:       for.inc:
+// CHECK2-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK2-NEXT:    [[INC:%.*]] = add nsw i64 [[TMP17]], 1
+// CHECK2-NEXT:    store i64 [[INC]], ptr [[DOTPERMUTED_1_IV___BEGIN2]], align 8
+// CHECK2-NEXT:    br label [[FOR_COND7]], !llvm.loop [[LOOP8:![0-9]+]]
+// CHECK2:       for.end:
+// CHECK2-NEXT:    br label [[FOR_INC13:%.*]]
+// CHECK2:       for.inc13:
+// CHECK2-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK2-NEXT:    [[INC14:%.*]] = add nsw i32 [[TMP18]], 1
+// CHECK2-NEXT:    store i32 [[INC14]], ptr [[DOTPERMUTED_0_IV_I]], align 4
+// CHECK2-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
+// CHECK2:       for.end15:
+// CHECK2-NEXT:    ret void
+//
diff --git a/clang/test/OpenMP/interchange_messages.cpp b/clang/test/OpenMP/interchange_messages.cpp
new file mode 100644
index 0000000000000..175c2f1efa744
--- /dev/null
+++ b/clang/test/OpenMP/interchange_messages.cpp
@@ -0,0 +1,77 @@
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++17 -fopenmp -fopenmp-version=60 -fsyntax-only -Wuninitialized -verify %s
+
+void func() {
+
+  // expected-warning at +1 {{extra tokens at the end of '#pragma omp interchange' are ignored}}
+  #pragma omp interchange foo
+  for (int i = 0; i < 7; ++i)
+    for (int j = 0; j < 13; ++j)
+      ;
+
+  // expected-error at +1 {{unexpected OpenMP clause 'collapse' in directive '#pragma omp interchange'}}
+  #pragma omp interchange collapse(2)
+  for (int i = 0; i < 7; ++i)
+    for (int j = 0; j < 13; ++j)
+      ;
+
+  {
+    // expected-error at +2 {{expected statement}}
+    #pragma omp interchange
+  }
+
+  // expected-error at +2 {{statement after '#pragma omp interchange' must be a for loop}}
+  #pragma omp interchange
+  int b = 0;
+
+  // expected-error at +3 {{statement after '#pragma omp interchange' must be a for loop}}
+  #pragma omp interchange
+  for (int i = 0; i < 7; ++i)
+    ;
+
+  // expected-error at +2 {{statement after '#pragma omp interchange' must be a for loop}}
+  #pragma omp interchange
+  for (int i = 0; i < 7; ++i) {
+    int k = 3;
+    for (int j = 0; j < 7; ++j)
+      ;
+  }
+
+  // expected-error at +3 {{expected loop invariant expression}}
+  #pragma omp interchange
+  for (int i = 0; i < 7; ++i)
+    for (int j = i; j < 7; ++j)
+      ;
+
+  // expected-error at +3 {{expected loop invariant expression}}
+  #pragma omp interchange
+  for (int i = 0; i < 7; ++i)
+    for (int j = 0; j < i; ++j)
+      ;
+
+  // expected-error at +3 {{expected loop invariant expression}}
+  #pragma omp interchange
+  for (int i = 0; i < 7; ++i)
+    for (int j = 0; j < i; ++j)
+      ;
+
+  // expected-error at +6 {{expected 3 for loops after '#pragma omp for', but found only 2}}
+  // expected-note at +1 {{as specified in 'collapse' clause}}
+  #pragma omp for collapse(3)
+  #pragma omp interchange 
+  for (int i = 0; i < 7; ++i)
+    for (int j = 0; j < 13; ++j)
+      ;
+
+  // expected-error at +2 {{statement after '#pragma omp interchange' must be a for loop}}
+  #pragma omp interchange
+  #pragma omp for
+  for (int i = 0; i < 7; ++i)
+    ;
+
+  // expected-error at +3 {{condition of OpenMP for loop must be a relational comparison ('<', '<=', '>', '>=', or '!=') of loop variable 'j'}}
+  #pragma omp interchange 
+  for (int i = 0; i < 7; ++i)
+    for (int j = 0; j/3<7; ++j)
+      ;
+}
+
diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
index 51784ed02b111..a6a002533936f 100644
--- a/clang/tools/libclang/CIndex.cpp
+++ b/clang/tools/libclang/CIndex.cpp
@@ -2180,6 +2180,7 @@ class EnqueueVisitor : public ConstStmtVisitor<EnqueueVisitor, void>,
   void VisitOMPTileDirective(const OMPTileDirective *D);
   void VisitOMPUnrollDirective(const OMPUnrollDirective *D);
   void VisitOMPReverseDirective(const OMPReverseDirective *D);
+  void VisitOMPInterchangeDirective(const OMPInterchangeDirective *D);
   void VisitOMPForDirective(const OMPForDirective *D);
   void VisitOMPForSimdDirective(const OMPForSimdDirective *D);
   void VisitOMPSectionsDirective(const OMPSectionsDirective *D);
@@ -3226,6 +3227,11 @@ void EnqueueVisitor::VisitOMPReverseDirective(const OMPReverseDirective *D) {
   VisitOMPLoopTransformationDirective(D);
 }
 
+void EnqueueVisitor::VisitOMPInterchangeDirective(
+    const OMPInterchangeDirective *D) {
+  VisitOMPLoopTransformationDirective(D);
+}
+
 void EnqueueVisitor::VisitOMPForDirective(const OMPForDirective *D) {
   VisitOMPLoopDirective(D);
 }
@@ -6086,6 +6092,8 @@ CXString clang_getCursorKindSpelling(enum CXCursorKind Kind) {
     return cxstring::createRef("OMPUnrollDirective");
   case CXCursor_OMPReverseDirective:
     return cxstring::createRef("OMPReverseDirective");
+  case CXCursor_OMPInterchangeDirective:
+    return cxstring::createRef("OMPInterchangeDirective");
   case CXCursor_OMPForDirective:
     return cxstring::createRef("OMPForDirective");
   case CXCursor_OMPForSimdDirective:
diff --git a/clang/tools/libclang/CXCursor.cpp b/clang/tools/libclang/CXCursor.cpp
index 275714d8fbcdd..9d485ce85b829 100644
--- a/clang/tools/libclang/CXCursor.cpp
+++ b/clang/tools/libclang/CXCursor.cpp
@@ -675,6 +675,9 @@ CXCursor cxcursor::MakeCXCursor(const Stmt *S, const Decl *Parent,
   case Stmt::OMPReverseDirectiveClass:
     K = CXCursor_OMPReverseDirective;
     break;
+  case Stmt::OMPInterchangeDirectiveClass:
+    K = CXCursor_OMPTileDirective;
+    break;
   case Stmt::OMPForDirectiveClass:
     K = CXCursor_OMPForDirective;
     break;
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index 0ad3d919103fe..3395f82ce1ba2 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -540,6 +540,9 @@ def OMP_Unroll : Directive<"unroll"> {
 def OMP_Reverse : Directive<"reverse"> {
   let association = AS_Loop;
 }
+def OMP_Interchange : Directive<"interchange"> {
+  let association = AS_Loop;
+}
 def OMP_For : Directive<"for"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
diff --git a/openmp/runtime/test/transform/interchange/foreach.cpp b/openmp/runtime/test/transform/interchange/foreach.cpp
new file mode 100644
index 0000000000000..17eaed6029d01
--- /dev/null
+++ b/openmp/runtime/test/transform/interchange/foreach.cpp
@@ -0,0 +1,216 @@
+// RUN: %libomp-cxx20-compile-and-run | FileCheck %s --match-full-lines
+
+#ifndef HEADER
+#define HEADER
+
+#include <cstdlib>
+#include <cstdarg>
+#include <cstdio>
+#include <vector>
+
+struct Reporter {
+  const char *name;
+
+  Reporter(const char *name) : name(name) { print("ctor"); }
+
+  Reporter() : name("<anon>") { print("ctor"); }
+
+  Reporter(const Reporter &that) : name(that.name) { print("copy ctor"); }
+
+  Reporter(Reporter &&that) : name(that.name) { print("move ctor"); }
+
+  ~Reporter() { print("dtor"); }
+
+  const Reporter &operator=(const Reporter &that) {
+    print("copy assign");
+    this->name = that.name;
+    return *this;
+  }
+
+  const Reporter &operator=(Reporter &&that) {
+    print("move assign");
+    this->name = that.name;
+    return *this;
+  }
+
+  struct Iterator {
+    const Reporter *owner;
+    int pos;
+
+    Iterator(const Reporter *owner, int pos) : owner(owner), pos(pos) {}
+
+    Iterator(const Iterator &that) : owner(that.owner), pos(that.pos) {
+      owner->print("iterator copy ctor");
+    }
+
+    Iterator(Iterator &&that) : owner(that.owner), pos(that.pos) {
+      owner->print("iterator move ctor");
+    }
+
+    ~Iterator() { owner->print("iterator dtor"); }
+
+    const Iterator &operator=(const Iterator &that) {
+      owner->print("iterator copy assign");
+      this->owner = that.owner;
+      this->pos = that.pos;
+      return *this;
+    }
+
+    const Iterator &operator=(Iterator &&that) {
+      owner->print("iterator move assign");
+      this->owner = that.owner;
+      this->pos = that.pos;
+      return *this;
+    }
+
+    bool operator==(const Iterator &that) const {
+      owner->print("iterator %d == %d", 2 - this->pos, 2 - that.pos);
+      return this->pos == that.pos;
+    }
+
+    Iterator &operator++() {
+      owner->print("iterator prefix ++");
+      pos -= 1;
+      return *this;
+    }
+
+    Iterator operator++(int) {
+      owner->print("iterator postfix ++");
+      auto result = *this;
+      pos -= 1;
+      return result;
+    }
+
+    int operator*() const {
+      int result = 2 - pos;
+      owner->print("iterator deref: %i", result);
+      return result;
+    }
+
+    size_t operator-(const Iterator &that) const {
+      int result = (2 - this->pos) - (2 - that.pos);
+      owner->print("iterator distance: %d", result);
+      return result;
+    }
+
+    Iterator operator+(int steps) const {
+      owner->print("iterator advance: %i += %i", 2 - this->pos, steps);
+      return Iterator(owner, pos - steps);
+    }
+
+    void print(const char *msg) const { owner->print(msg); }
+  };
+
+  Iterator begin() const {
+    print("begin()");
+    return Iterator(this, 2);
+  }
+
+  Iterator end() const {
+    print("end()");
+    return Iterator(this, -1);
+  }
+
+  void print(const char *msg, ...) const {
+    va_list args;
+    va_start(args, msg);
+    printf("[%s] ", name);
+    vprintf(msg, args);
+    printf("\n");
+    va_end(args);
+  }
+};
+
+int main() {
+  printf("do\n");
+#pragma omp interchange
+  for (Reporter c{"C"}; auto &&v : Reporter("A"))
+    for (Reporter d{"D"}; auto &&w : Reporter("B"))
+      printf("v=%d w=%d\n", v, w);
+  printf("done\n");
+  return EXIT_SUCCESS;
+}
+
+#endif /* HEADER */
+
+// CHECK:      do
+// CHECK-NEXT: [C] ctor
+// CHECK-NEXT: [A] ctor
+// CHECK-NEXT: [A] end()
+// CHECK-NEXT: [A] begin()
+// CHECK-NEXT: [A] begin()
+// CHECK-NEXT: [A] iterator distance: 3
+// CHECK-NEXT: [D] ctor
+// CHECK-NEXT: [B] ctor
+// CHECK-NEXT: [B] end()
+// CHECK-NEXT: [B] begin()
+// CHECK-NEXT: [B] begin()
+// CHECK-NEXT: [B] iterator distance: 3
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: v=0 w=0
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: v=1 w=0
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: v=2 w=0
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: v=0 w=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: v=1 w=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: v=2 w=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: v=0 w=2
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: v=1 w=2
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: v=2 w=2
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] dtor
+// CHECK-NEXT: [D] dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] dtor
+// CHECK-NEXT: [C] dtor
+// CHECK-NEXT: done
diff --git a/openmp/runtime/test/transform/interchange/intfor.c b/openmp/runtime/test/transform/interchange/intfor.c
new file mode 100644
index 0000000000000..b4842f0f82913
--- /dev/null
+++ b/openmp/runtime/test/transform/interchange/intfor.c
@@ -0,0 +1,38 @@
+// RUN: %libomp-compile-and-run | FileCheck %s --match-full-lines
+
+#ifndef HEADER
+#define HEADER
+
+#include <stdlib.h>
+#include <stdio.h>
+
+int main() {
+  printf("do\n");
+#pragma omp interchange
+  for (int i = 7; i < 17; i += 3)
+    for (int j = 8; j < 18; j += 3)
+      printf("i=%d j=%d\n", i, j);
+  printf("done\n");
+  return EXIT_SUCCESS;
+}
+
+#endif /* HEADER */
+
+// CHECK:      do
+// CHECK-NEXT: i=7 j=8
+// CHECK-NEXT: i=10 j=8
+// CHECK-NEXT: i=13 j=8
+// CHECK-NEXT: i=16 j=8
+// CHECK-NEXT: i=7 j=11
+// CHECK-NEXT: i=10 j=11
+// CHECK-NEXT: i=13 j=11
+// CHECK-NEXT: i=16 j=11
+// CHECK-NEXT: i=7 j=14
+// CHECK-NEXT: i=10 j=14
+// CHECK-NEXT: i=13 j=14
+// CHECK-NEXT: i=16 j=14
+// CHECK-NEXT: i=7 j=17
+// CHECK-NEXT: i=10 j=17
+// CHECK-NEXT: i=13 j=17
+// CHECK-NEXT: i=16 j=17
+// CHECK-NEXT: done
diff --git a/openmp/runtime/test/transform/interchange/iterfor.cpp b/openmp/runtime/test/transform/interchange/iterfor.cpp
new file mode 100644
index 0000000000000..51219a07402e3
--- /dev/null
+++ b/openmp/runtime/test/transform/interchange/iterfor.cpp
@@ -0,0 +1,222 @@
+// RUN: %libomp-cxx20-compile-and-run | FileCheck %s --match-full-lines
+
+#ifndef HEADER
+#define HEADER
+
+#include <cstdlib>
+#include <cstdarg>
+#include <cstdio>
+#include <vector>
+
+struct Reporter {
+  const char *name;
+
+  Reporter(const char *name) : name(name) { print("ctor"); }
+
+  Reporter() : name("<anon>") { print("ctor"); }
+
+  Reporter(const Reporter &that) : name(that.name) { print("copy ctor"); }
+
+  Reporter(Reporter &&that) : name(that.name) { print("move ctor"); }
+
+  ~Reporter() { print("dtor"); }
+
+  const Reporter &operator=(const Reporter &that) {
+    print("copy assign");
+    this->name = that.name;
+    return *this;
+  }
+
+  const Reporter &operator=(Reporter &&that) {
+    print("move assign");
+    this->name = that.name;
+    return *this;
+  }
+
+  struct Iterator {
+    const Reporter *owner;
+    int pos;
+
+    Iterator(const Reporter *owner, int pos) : owner(owner), pos(pos) {}
+
+    Iterator(const Iterator &that) : owner(that.owner), pos(that.pos) {
+      owner->print("iterator copy ctor");
+    }
+
+    Iterator(Iterator &&that) : owner(that.owner), pos(that.pos) {
+      owner->print("iterator move ctor");
+    }
+
+    ~Iterator() { owner->print("iterator dtor"); }
+
+    const Iterator &operator=(const Iterator &that) {
+      owner->print("iterator copy assign");
+      this->owner = that.owner;
+      this->pos = that.pos;
+      return *this;
+    }
+
+    const Iterator &operator=(Iterator &&that) {
+      owner->print("iterator move assign");
+      this->owner = that.owner;
+      this->pos = that.pos;
+      return *this;
+    }
+
+    bool operator==(const Iterator &that) const {
+      owner->print("iterator %d == %d", 2 - this->pos, 2 - that.pos);
+      return this->pos == that.pos;
+    }
+
+    bool operator!=(const Iterator &that) const {
+      owner->print("iterator %d != %d", 2 - this->pos, 2 - that.pos);
+      return this->pos == that.pos;
+    }
+
+    Iterator &operator++() {
+      owner->print("iterator prefix ++");
+      pos -= 1;
+      return *this;
+    }
+
+    Iterator operator++(int) {
+      owner->print("iterator postfix ++");
+      auto result = *this;
+      pos -= 1;
+      return result;
+    }
+
+    int operator*() const {
+      int result = 2 - pos;
+      owner->print("iterator deref: %i", result);
+      return result;
+    }
+
+    size_t operator-(const Iterator &that) const {
+      int result = (2 - this->pos) - (2 - that.pos);
+      owner->print("iterator distance: %d", result);
+      return result;
+    }
+
+    Iterator operator+(int steps) const {
+      owner->print("iterator advance: %i += %i", 2 - this->pos, steps);
+      return Iterator(owner, pos - steps);
+    }
+  };
+
+  Iterator begin() const {
+    print("begin()");
+    return Iterator(this, 2);
+  }
+
+  Iterator end() const {
+    print("end()");
+    return Iterator(this, -1);
+  }
+
+  void print(const char *msg, ...) const {
+    va_list args;
+    va_start(args, msg);
+    printf("[%s] ", name);
+    vprintf(msg, args);
+    printf("\n");
+    va_end(args);
+  }
+};
+
+int main() {
+  printf("do\n");
+  Reporter A("A"), B("B");
+#pragma omp interchange
+  for (auto it = A.begin(); it != A.end(); ++it)
+    for (auto jt = B.begin(); jt != B.end(); ++jt)
+      printf("i=%d j=%d\n", *it, *jt);
+  printf("done\n");
+  return EXIT_SUCCESS;
+}
+
+#endif /* HEADER */
+
+// CHECK:      do
+// CHECK-NEXT: [A] ctor
+// CHECK-NEXT: [B] ctor
+// CHECK-NEXT: [A] begin()
+// CHECK-NEXT: [A] begin()
+// CHECK-NEXT: [A] end()
+// CHECK-NEXT: [A] iterator distance: 3
+// CHECK-NEXT: [B] begin()
+// CHECK-NEXT: [B] begin()
+// CHECK-NEXT: [B] end()
+// CHECK-NEXT: [B] iterator distance: 3
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: i=0 j=0
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: i=1 j=0
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: i=2 j=0
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: i=0 j=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: i=1 j=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: i=2 j=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: i=0 j=2
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: i=1 j=2
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: i=2 j=2
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: done
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] dtor
+// CHECK-NEXT: [A] dtor
diff --git a/openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-foreach.cpp b/openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-foreach.cpp
new file mode 100644
index 0000000000000..54399d25a3f15
--- /dev/null
+++ b/openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-foreach.cpp
@@ -0,0 +1,340 @@
+// RUN: %libomp-cxx20-compile-and-run | FileCheck %s --match-full-lines
+
+#ifndef HEADER
+#define HEADER
+
+#include <cstdlib>
+#include <cstdarg>
+#include <cstdio>
+#include <vector>
+
+struct Reporter {
+  const char *name;
+
+  Reporter(const char *name) : name(name) { print("ctor"); }
+
+  Reporter() : name("<anon>") { print("ctor"); }
+
+  Reporter(const Reporter &that) : name(that.name) { print("copy ctor"); }
+
+  Reporter(Reporter &&that) : name(that.name) { print("move ctor"); }
+
+  ~Reporter() { print("dtor"); }
+
+  const Reporter &operator=(const Reporter &that) {
+    print("copy assign");
+    this->name = that.name;
+    return *this;
+  }
+
+  const Reporter &operator=(Reporter &&that) {
+    print("move assign");
+    this->name = that.name;
+    return *this;
+  }
+
+  struct Iterator {
+    const Reporter *owner;
+    int pos;
+
+    Iterator(const Reporter *owner, int pos) : owner(owner), pos(pos) {}
+
+    Iterator(const Iterator &that) : owner(that.owner), pos(that.pos) {
+      owner->print("iterator copy ctor");
+    }
+
+    Iterator(Iterator &&that) : owner(that.owner), pos(that.pos) {
+      owner->print("iterator move ctor");
+    }
+
+    ~Iterator() { owner->print("iterator dtor"); }
+
+    const Iterator &operator=(const Iterator &that) {
+      owner->print("iterator copy assign");
+      this->owner = that.owner;
+      this->pos = that.pos;
+      return *this;
+    }
+
+    const Iterator &operator=(Iterator &&that) {
+      owner->print("iterator move assign");
+      this->owner = that.owner;
+      this->pos = that.pos;
+      return *this;
+    }
+
+    bool operator==(const Iterator &that) const {
+      owner->print("iterator %d == %d", 2 - this->pos, 2 - that.pos);
+      return this->pos == that.pos;
+    }
+
+    Iterator &operator++() {
+      owner->print("iterator prefix ++");
+      pos -= 1;
+      return *this;
+    }
+
+    Iterator operator++(int) {
+      owner->print("iterator postfix ++");
+      auto result = *this;
+      pos -= 1;
+      return result;
+    }
+
+    int operator*() const {
+      int result = 2 - pos;
+      owner->print("iterator deref: %i", result);
+      return result;
+    }
+
+    size_t operator-(const Iterator &that) const {
+      int result = (2 - this->pos) - (2 - that.pos);
+      owner->print("iterator distance: %d", result);
+      return result;
+    }
+
+    Iterator operator+(int steps) const {
+      owner->print("iterator advance: %i += %i", 2 - this->pos, steps);
+      return Iterator(owner, pos - steps);
+    }
+  };
+
+  Iterator begin() const {
+    print("begin()");
+    return Iterator(this, 2);
+  }
+
+  Iterator end() const {
+    print("end()");
+    return Iterator(this, -1);
+  }
+
+  void print(const char *msg, ...) const {
+    va_list args;
+    va_start(args, msg);
+    printf("[%s] ", name);
+    vprintf(msg, args);
+    printf("\n");
+    va_end(args);
+  }
+};
+
+int main() {
+  printf("do\n");
+#pragma omp parallel for collapse(3) num_threads(1)
+  for (int i = 0; i < 2; ++i)
+#pragma omp interchange
+    for (Reporter c{"C"}; auto &&v : Reporter("A"))
+      for (Reporter d{"D"}; auto &&w : Reporter("B"))
+        for (int k = 0; k < 2; ++k)
+          printf("i=%d v=%d w=%d k=%d\n", i, v, w, k);
+  printf("done\n");
+  return EXIT_SUCCESS;
+}
+
+#endif /* HEADER */
+
+// CHECK:      do
+// CHECK-NEXT: [C] ctor
+// CHECK-NEXT: [A] ctor
+// CHECK-NEXT: [A] end()
+// CHECK-NEXT: [A] begin()
+// CHECK-NEXT: [A] begin()
+// CHECK-NEXT: [A] iterator distance: 3
+// CHECK-NEXT: [D] ctor
+// CHECK-NEXT: [B] ctor
+// CHECK-NEXT: [B] end()
+// CHECK-NEXT: [B] begin()
+// CHECK-NEXT: [B] begin()
+// CHECK-NEXT: [B] iterator distance: 3
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: i=0 v=0 w=0 k=0
+// CHECK-NEXT: i=0 v=0 w=0 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: i=0 v=1 w=0 k=0
+// CHECK-NEXT: i=0 v=1 w=0 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: i=0 v=2 w=0 k=0
+// CHECK-NEXT: i=0 v=2 w=0 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: i=0 v=0 w=1 k=0
+// CHECK-NEXT: i=0 v=0 w=1 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: i=0 v=1 w=1 k=0
+// CHECK-NEXT: i=0 v=1 w=1 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: i=0 v=2 w=1 k=0
+// CHECK-NEXT: i=0 v=2 w=1 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: i=0 v=0 w=2 k=0
+// CHECK-NEXT: i=0 v=0 w=2 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: i=0 v=1 w=2 k=0
+// CHECK-NEXT: i=0 v=1 w=2 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: i=0 v=2 w=2 k=0
+// CHECK-NEXT: i=0 v=2 w=2 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: i=1 v=0 w=0 k=0
+// CHECK-NEXT: i=1 v=0 w=0 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: i=1 v=1 w=0 k=0
+// CHECK-NEXT: i=1 v=1 w=0 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 0
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 0
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: i=1 v=2 w=0 k=0
+// CHECK-NEXT: i=1 v=2 w=0 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: i=1 v=0 w=1 k=0
+// CHECK-NEXT: i=1 v=0 w=1 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: i=1 v=1 w=1 k=0
+// CHECK-NEXT: i=1 v=1 w=1 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 1
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 1
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: i=1 v=2 w=1 k=0
+// CHECK-NEXT: i=1 v=2 w=1 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: [A] iterator advance: 0 += 0
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 0
+// CHECK-NEXT: i=1 v=0 w=2 k=0
+// CHECK-NEXT: i=1 v=0 w=2 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: [A] iterator advance: 0 += 1
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 1
+// CHECK-NEXT: i=1 v=1 w=2 k=0
+// CHECK-NEXT: i=1 v=1 w=2 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator advance: 0 += 2
+// CHECK-NEXT: [B] iterator move assign
+// CHECK-NEXT: [B] iterator deref: 2
+// CHECK-NEXT: [A] iterator advance: 0 += 2
+// CHECK-NEXT: [A] iterator move assign
+// CHECK-NEXT: [A] iterator deref: 2
+// CHECK-NEXT: i=1 v=2 w=2 k=0
+// CHECK-NEXT: i=1 v=2 w=2 k=1
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] iterator dtor
+// CHECK-NEXT: [B] dtor
+// CHECK-NEXT: [D] dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] iterator dtor
+// CHECK-NEXT: [A] dtor
+// CHECK-NEXT: [C] dtor
+// CHECK-NEXT: done
diff --git a/openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-intfor.cpp b/openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-intfor.cpp
new file mode 100644
index 0000000000000..a4520afdb0f9c
--- /dev/null
+++ b/openmp/runtime/test/transform/interchange/parallel-wsloop-collapse-intfor.cpp
@@ -0,0 +1,106 @@
+// RUN: %libomp-cxx-compile-and-run | FileCheck %s --match-full-lines
+
+#ifndef HEADER
+#define HEADER
+
+#include <cstdlib>
+#include <cstdio>
+
+int main() {
+  printf("do\n");
+#pragma omp parallel for collapse(4) num_threads(1)
+  for (int i = 0; i < 3; ++i)
+#pragma omp interchange
+    for (int j = 0; j < 3; ++j)
+      for (int k = 0; k < 3; ++k)
+        for (int l = 0; l < 3; ++l)
+          printf("i=%d j=%d k=%d l=%d\n", i, j, k, l);
+  printf("done\n");
+  return EXIT_SUCCESS;
+}
+
+#endif /* HEADER */
+
+// CHECK:      do
+// CHECK-NEXT: i=0 j=0 k=0 l=0
+// CHECK-NEXT: i=0 j=0 k=0 l=1
+// CHECK-NEXT: i=0 j=0 k=0 l=2
+// CHECK-NEXT: i=0 j=1 k=0 l=0
+// CHECK-NEXT: i=0 j=1 k=0 l=1
+// CHECK-NEXT: i=0 j=1 k=0 l=2
+// CHECK-NEXT: i=0 j=2 k=0 l=0
+// CHECK-NEXT: i=0 j=2 k=0 l=1
+// CHECK-NEXT: i=0 j=2 k=0 l=2
+// CHECK-NEXT: i=0 j=0 k=1 l=0
+// CHECK-NEXT: i=0 j=0 k=1 l=1
+// CHECK-NEXT: i=0 j=0 k=1 l=2
+// CHECK-NEXT: i=0 j=1 k=1 l=0
+// CHECK-NEXT: i=0 j=1 k=1 l=1
+// CHECK-NEXT: i=0 j=1 k=1 l=2
+// CHECK-NEXT: i=0 j=2 k=1 l=0
+// CHECK-NEXT: i=0 j=2 k=1 l=1
+// CHECK-NEXT: i=0 j=2 k=1 l=2
+// CHECK-NEXT: i=0 j=0 k=2 l=0
+// CHECK-NEXT: i=0 j=0 k=2 l=1
+// CHECK-NEXT: i=0 j=0 k=2 l=2
+// CHECK-NEXT: i=0 j=1 k=2 l=0
+// CHECK-NEXT: i=0 j=1 k=2 l=1
+// CHECK-NEXT: i=0 j=1 k=2 l=2
+// CHECK-NEXT: i=0 j=2 k=2 l=0
+// CHECK-NEXT: i=0 j=2 k=2 l=1
+// CHECK-NEXT: i=0 j=2 k=2 l=2
+// CHECK-NEXT: i=1 j=0 k=0 l=0
+// CHECK-NEXT: i=1 j=0 k=0 l=1
+// CHECK-NEXT: i=1 j=0 k=0 l=2
+// CHECK-NEXT: i=1 j=1 k=0 l=0
+// CHECK-NEXT: i=1 j=1 k=0 l=1
+// CHECK-NEXT: i=1 j=1 k=0 l=2
+// CHECK-NEXT: i=1 j=2 k=0 l=0
+// CHECK-NEXT: i=1 j=2 k=0 l=1
+// CHECK-NEXT: i=1 j=2 k=0 l=2
+// CHECK-NEXT: i=1 j=0 k=1 l=0
+// CHECK-NEXT: i=1 j=0 k=1 l=1
+// CHECK-NEXT: i=1 j=0 k=1 l=2
+// CHECK-NEXT: i=1 j=1 k=1 l=0
+// CHECK-NEXT: i=1 j=1 k=1 l=1
+// CHECK-NEXT: i=1 j=1 k=1 l=2
+// CHECK-NEXT: i=1 j=2 k=1 l=0
+// CHECK-NEXT: i=1 j=2 k=1 l=1
+// CHECK-NEXT: i=1 j=2 k=1 l=2
+// CHECK-NEXT: i=1 j=0 k=2 l=0
+// CHECK-NEXT: i=1 j=0 k=2 l=1
+// CHECK-NEXT: i=1 j=0 k=2 l=2
+// CHECK-NEXT: i=1 j=1 k=2 l=0
+// CHECK-NEXT: i=1 j=1 k=2 l=1
+// CHECK-NEXT: i=1 j=1 k=2 l=2
+// CHECK-NEXT: i=1 j=2 k=2 l=0
+// CHECK-NEXT: i=1 j=2 k=2 l=1
+// CHECK-NEXT: i=1 j=2 k=2 l=2
+// CHECK-NEXT: i=2 j=0 k=0 l=0
+// CHECK-NEXT: i=2 j=0 k=0 l=1
+// CHECK-NEXT: i=2 j=0 k=0 l=2
+// CHECK-NEXT: i=2 j=1 k=0 l=0
+// CHECK-NEXT: i=2 j=1 k=0 l=1
+// CHECK-NEXT: i=2 j=1 k=0 l=2
+// CHECK-NEXT: i=2 j=2 k=0 l=0
+// CHECK-NEXT: i=2 j=2 k=0 l=1
+// CHECK-NEXT: i=2 j=2 k=0 l=2
+// CHECK-NEXT: i=2 j=0 k=1 l=0
+// CHECK-NEXT: i=2 j=0 k=1 l=1
+// CHECK-NEXT: i=2 j=0 k=1 l=2
+// CHECK-NEXT: i=2 j=1 k=1 l=0
+// CHECK-NEXT: i=2 j=1 k=1 l=1
+// CHECK-NEXT: i=2 j=1 k=1 l=2
+// CHECK-NEXT: i=2 j=2 k=1 l=0
+// CHECK-NEXT: i=2 j=2 k=1 l=1
+// CHECK-NEXT: i=2 j=2 k=1 l=2
+// CHECK-NEXT: i=2 j=0 k=2 l=0
+// CHECK-NEXT: i=2 j=0 k=2 l=1
+// CHECK-NEXT: i=2 j=0 k=2 l=2
+// CHECK-NEXT: i=2 j=1 k=2 l=0
+// CHECK-NEXT: i=2 j=1 k=2 l=1
+// CHECK-NEXT: i=2 j=1 k=2 l=2
+// CHECK-NEXT: i=2 j=2 k=2 l=0
+// CHECK-NEXT: i=2 j=2 k=2 l=1
+// CHECK-NEXT: i=2 j=2 k=2 l=2
+// CHECK-NEXT: done



More information about the llvm-branch-commits mailing list