[OpenMP] Implement '#pragma omp unroll'.

[Meinersbur]

Summary

Implementation of the unroll directive introduced in OpenMP 5.1. Follows the approach from D76342 for the tile directive (i.e. AST-based, not using the OpenMPIRBuilder). Tries to use llvm.loop.unroll.* metadata where possible, but has to fall back to an AST representation of the outer loop if the partially unrolled generated loop is associated with another directive (because it needs to compute the number of iterations).

[ABataev]

clang/include/clang/AST/OpenMPClause.h

899

sizes->full

902

Same

903

No param

904

Wrong description

905

No param

912

No param

946

sizes->partial

948–952

Fix params descriptions

957–960

Description

971

Make it private, if possible

clang/include/clang/AST/StmtOpenMP.h

5070

Description

5094–5105

Fix description

5111–5115

Description

clang/lib/AST/StmtProfile.cpp

466

Unelated change?

474

const Expr *

clang/lib/CodeGen/CGStmtOpenMP.cpp

2589

getSingleClause or hasClausesOfKind

2596

getSingleClause

2605

const Expr *

2606–2609

I suppose it is compiled-time expression, right? If so, use FactorExpr->EvaluateKnownConstInt() or something similar.

clang/lib/Parse/ParseOpenMP.cpp

2973–2974

Better to remove the default: case completely

clang/lib/Sema/SemaOpenMP.cpp

12558

I think this function can be made static local as it is used only SemaOpenMP.cpp

12822

getSingleClause

12825

Add a text message to the assert.

12830

getSingleClause

12848

Need to remove

12855–12857

I think this should be checked before the first OMPUnrollDirective::Create call

12931

unroll?

14963

Need to pass factor expr and real source loc

clang/lib/Serialization/ASTReaderStmt.cpp

3194

Add a message for the assert

[Meinersbur]

• Ready version

[Meinersbur]

• Undo spurious change

[Meinersbur]

Removed [WIP] flag; ready to review.

[ABataev]

Some previous comments were not addressed yet

clang/include/clang/AST/OpenMPClause.h

972–974

Make it private

clang/lib/AST/StmtOpenMP.cpp

141

Enclose in braces

clang/lib/CodeGen/CGStmtOpenMP.cpp

2600–2604

I suppose it is compiled-time expression, right? If so, use FactorExpr->EvaluateKnownConstInt() or something similar.

[Meinersbur]

• Address review

[ABataev]

clang/include/clang/AST/OpenMPClause.h

907

Fix params descriptions

clang/include/clang/AST/StmtOpenMP.h

463–479

Better to commit it it separately as NFC.

clang/lib/AST/StmtOpenMP.cpp

381

auto *

clang/lib/CodeGen/CGStmtOpenMP.cpp

2592–2596

Can we have full and partial at the same time? If no, emit partial in else substatement for full clause.

clang/lib/Sema/SemaOpenMP.cpp

8982–8983

Enclose in braces

10137–10141

Would be good if this change is committed separately as NFC.

12602

assert message?

12818

Can you use VerifyPositiveIntegerConstantInClause function instead?

12952–12953

Better to use PartialClause->getFactor()->isIntegerConstantExpr(Ctx) or something similar.

12961

Why do you need it? Just use original PartialClause->getFactor()

12982

MakeOuterRef

12990

MakeInnerRef

12997

MakeNumIterations

[Meinersbur]

clang/include/clang/AST/StmtOpenMP.h

463–479

D103665

clang/lib/Sema/SemaOpenMP.cpp

10137–10141

D103666

12818

The reason for this new function is that VerifyPositiveIntegerConstantInClause also emits a note on why it is not a constant expression:

note: read of non-const variable '.capture_expr.' is not allowed in a constant expression

Printing internal variable names is more confusing than helpful to the user.

12961

The AST invariant that no ASTNode can be used multiple times (within the same DeclContext) must be preserved. This is a utility lambda (like the other MakeXyz) that create a new IntegerLiteral node for ever use.

See discussion here: https://reviews.llvm.org/D94973?id=322600#inline-905341

[Meinersbur]

• Address review

[ABataev]

clang/lib/Sema/SemaOpenMP.cpp

12818

Maybe expand VerifyPositiveIntegerConstantInClause to exclude this rather than add a new function?

[Meinersbur]

• Integreate isConstantExpression into VerifyPositiveIntegerConstantInClause

[ABataev]

clang/lib/CodeGen/CGStmtOpenMP.cpp

2600–2604

Not done.

[Meinersbur]

• Merge branch 'arcpatch-D99459' into pragma-omp-unroll

[Meinersbur]

clang/lib/CodeGen/CGStmtOpenMP.cpp

2600–2604

It was done in another branch that I forgot to merge. Sorry.

[ABataev]

LG

.gitignore

 #OS X specific files.
 .DS_store
 
 # Ignore the user specified CMake presets in subproject directories.
 /*/CMakeUserPresets.json
 
 # Nested build directory
 /build*
+*.FileCheck
 
 #==============================================================================#
 # Explicit files to ignore (only matches one).
 #==============================================================================#
 # Various tag programs
 /tags
 /TAGS
 /GPATH

clang/include/clang-c/Index.h

enum CXCursorKind {
   /** OpenMP dispatch directive.
    */
   CXCursor_OMPDispatchDirective = 291,
 
   /** OpenMP masked directive.
    */
   CXCursor_OMPMaskedDirective = 292,
 
-  CXCursor_LastStmt = CXCursor_OMPMaskedDirective,
+  /** OpenMP unroll directive.
+   */
+  CXCursor_OMPUnrollDirective = 293,
+
+  CXCursor_LastStmt = CXCursor_OMPUnrollDirective,
 
   /**
    * Cursor that represents the translation unit itself.
    *
    * The translation unit cursor exists primarily to act as the root
    * cursor for traversing the contents of a translation unit.
    */
   CXCursor_TranslationUnit = 300,

clang/include/clang/AST/OpenMPClause.h

const_child_range used_children() const {
     return const_child_range(const_child_iterator(), const_child_iterator());
   }
 
   static bool classof(const OMPClause *T) {
     return T->getClauseKind() == llvm::omp::OMPC_sizes;
   }
 };
 
+/// Representation of the 'full' clause of the '#pragma omp unroll' directive.
+///
+/// \code
+/// #pragma omp unroll full
+/// for (int i = 0; i < 64; ++i)
+/// \endcode
+class OMPFullClause final : public OMPClause {
+  friend class OMPClauseReader;
+
+  /// Build an empty clause.
+  explicit OMPFullClause() : OMPClause(llvm::omp::OMPC_full, {}, {}) {}
+
+public:
+  /// Build an AST node for a 'full' clause.
+  ///
+  /// \param C   Context of the AST.
+  /// \param Loc Location of the 'full' identifier.
+  static OMPFullClause *Create(const ASTContext &C, SourceLocation StartLoc,
+                               SourceLocation EndLoc);
+
+  /// Build an empty 'full' AST node for deserialization.
+  ///
+  /// \param C Context of the AST.
+  static OMPFullClause *CreateEmpty(const ASTContext &C);
+
+  child_range children() { return {child_iterator(), child_iterator()}; }
+  const_child_range children() const {
+    return {const_child_iterator(), const_child_iterator()};
+  }
+
+  child_range used_children() {
+    return child_range(child_iterator(), child_iterator());
+  }
+  const_child_range used_children() const {
+    return const_child_range(const_child_iterator(), const_child_iterator());
+  }
+
+  static bool classof(const OMPClause *T) {
+    return T->getClauseKind() == llvm::omp::OMPC_full;
+  }
+};
+
+/// Representation of the 'partial' clause of the '#pragma omp unroll'
+/// directive.
+///
+/// \code
+/// #pragma omp unroll partial(4)
+/// for (int i = start; i < end; ++i)
+/// \endcode
+class OMPPartialClause final : public OMPClause {
+  friend class OMPClauseReader;
+
+  /// Location of '('.
+  SourceLocation LParenLoc;
+
+  /// Optional argument to the clause (unroll factor).
+  Stmt *Factor;
+
+  /// Build an empty clause.
+  explicit OMPPartialClause() : OMPClause(llvm::omp::OMPC_partial, {}, {}) {}
+
+  /// Set the unroll factor.
+  void setFactor(Expr *E) { Factor = E; }
+
+public:
+  /// Build an AST node for a 'partial' clause.
+  ///
+  /// \param C         Context of the AST.
+  /// \param StartLoc  Location of the 'partial' identifier.
+  /// \param LParenLoc Location of '('.
+  /// \param EndLoc    Location of ')'.
+  /// \param Factor    Clause argument.
+  static OMPPartialClause *Create(const ASTContext &C, SourceLocation StartLoc,
+                                  SourceLocation LParenLoc,
+                                  SourceLocation EndLoc, Expr *Factor);
+
+  /// Build an empty 'partial' AST node for deserialization.
+  ///
+  /// \param C     Context of the AST.
+  static OMPPartialClause *CreateEmpty(const ASTContext &C);
+
+  /// Sets the location of '('.
+  void setLParenLoc(SourceLocation Loc) { LParenLoc = Loc; }
+
+  /// Returns the location of '('.
+  SourceLocation getLParenLoc() const { return LParenLoc; }
+
+  /// Returns the argument of the clause or nullptr if not set.
+  Expr *getFactor() const { return cast_or_null<Expr>(Factor); }
+
+  child_range children() { return child_range(&Factor, &Factor + 1); }
+  const_child_range children() const {
+    return const_child_range(&Factor, &Factor + 1);
+  }
+
+  child_range used_children() {
+    return child_range(child_iterator(), child_iterator());
+  }
+  const_child_range used_children() const {
+    return const_child_range(const_child_iterator(), const_child_iterator());
+  }
+
+  static bool classof(const OMPClause *T) {
+    return T->getClauseKind() == llvm::omp::OMPC_partial;
+  }
+};
+
 /// This represents 'collapse' clause in the '#pragma omp ...'
 /// directive.
 ///
 /// \code
 /// #pragma omp simd collapse(3)
 /// \endcode
 /// In this example directive '#pragma omp simd' has clause 'collapse'
 /// with single expression '3'.

clang/include/clang/AST/RecursiveASTVisitor.h

DEF_TRAVERSE_STMT(OMPParallelDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })
 
 DEF_TRAVERSE_STMT(OMPSimdDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })
 
 DEF_TRAVERSE_STMT(OMPTileDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })
 
+DEF_TRAVERSE_STMT(OMPUnrollDirective,
+                  { TRY_TO(TraverseOMPExecutableDirective(S)); })
+
 DEF_TRAVERSE_STMT(OMPForDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })
 
 DEF_TRAVERSE_STMT(OMPForSimdDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })
 
 DEF_TRAVERSE_STMT(OMPSectionsDirective,
                   { TRY_TO(TraverseOMPExecutableDirective(S)); })

 template <typename Derived>
 bool RecursiveASTVisitor<Derived>::VisitOMPSizesClause(OMPSizesClause *C) {
   for (Expr *E : C->getSizesRefs())
     TRY_TO(TraverseStmt(E));
   return true;
 }
 
 template <typename Derived>
+bool RecursiveASTVisitor<Derived>::VisitOMPFullClause(OMPFullClause *C) {
+  return true;
+}
+
+template <typename Derived>
+bool RecursiveASTVisitor<Derived>::VisitOMPPartialClause(OMPPartialClause *C) {
+  TRY_TO(TraverseStmt(C->getFactor()));
+  return true;
+}
+
+template <typename Derived>
 bool
 RecursiveASTVisitor<Derived>::VisitOMPCollapseClause(OMPCollapseClause *C) {
   TRY_TO(TraverseStmt(C->getNumForLoops()));
   return true;
 }
 
 template <typename Derived>
 bool RecursiveASTVisitor<Derived>::VisitOMPDefaultClause(OMPDefaultClause *) {

clang/include/clang/AST/StmtOpenMP.h

public:
   }
 
   /// Gets a single clause of the specified kind associated with the
   /// current directive iff there is only one clause of this kind (and assertion
   /// is fired if there is more than one clause is associated with the
   /// directive). Returns nullptr if no clause of this kind is associated with
   /// the directive.
   template <typename SpecificClause>
-  const SpecificClause *getSingleClause() const {
-    auto Clauses = getClausesOfKind<SpecificClause>();
+  static const SpecificClause *getSingleClause(ArrayRef<OMPClause *> Clauses) {
+    auto ClausesOfKind = getClausesOfKind<SpecificClause>(Clauses);
 
-    if (Clauses.begin() != Clauses.end()) {
-      assert(std::next(Clauses.begin()) == Clauses.end() &&
+    if (ClausesOfKind.begin() != ClausesOfKind.end()) {
+      assert(std::next(ClausesOfKind.begin()) == ClausesOfKind.end() &&
              "There are at least 2 clauses of the specified kind");
-      return *Clauses.begin();
+      return *ClausesOfKind.begin();
     }
     return nullptr;
   }
 
+  template <typename SpecificClause>
+  const SpecificClause *getSingleClause() const {
+    return getSingleClause<SpecificClause>(clauses());
+  }
+
   /// Returns true if the current directive has one or more clauses of a
   /// specific kind.
   template <typename SpecificClause>
   bool hasClausesOfKind() const {
     auto Clauses = getClausesOfKind<SpecificClause>();
     return Clauses.begin() != Clauses.end();
   }
 
   /// Returns starting location of directive kind.
public:
   /// Return preinits statement.
   Stmt *getPreInits() const { return Data->getChildren()[PreInitsOffset]; }
 
   static bool classof(const Stmt *T) {
     return T->getStmtClass() == OMPTileDirectiveClass;
   }
 };
 
+/// This represents the '#pragma omp unroll' loop transformation directive.
+///
+/// \code
+/// #pragma omp unroll
+/// for (int i = 0; i < 64; ++i)
+/// \endcode
+class OMPUnrollDirective final : public OMPLoopBasedDirective {
+  friend class ASTStmtReader;
+  friend class OMPExecutableDirective;
+
+  /// Default list of offsets.
+  enum {
+    PreInitsOffset = 0,
+    TransformedStmtOffset,
+  };
+
+  explicit OMPUnrollDirective(SourceLocation StartLoc, SourceLocation EndLoc)
+      : OMPLoopBasedDirective(OMPUnrollDirectiveClass, llvm::omp::OMPD_unroll,
+                              StartLoc, EndLoc, 1) {}
+
+  /// Set the pre-init statements.
+  void setPreInits(Stmt *PreInits) {
+    Data->getChildren()[PreInitsOffset] = PreInits;
+  }
+
+  /// Set the de-sugared statement.
+  void setTransformedStmt(Stmt *S) {
+    Data->getChildren()[TransformedStmtOffset] = S;
+  }
+
+public:
+  /// Create a new AST node representation for '#pragma omp unroll'.
+  ///
+  /// \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 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 OMPUnrollDirective *
+  Create(const ASTContext &C, SourceLocation StartLoc, SourceLocation EndLoc,
+         ArrayRef<OMPClause *> Clauses, Stmt *AssociatedStmt,
+         Stmt *TransformedStmt, Stmt *PreInits);
+
+  /// Build an empty '#pragma omp unroll' AST node for deserialization.
+  ///
+  /// \param C          Context of the AST.
+  /// \param NumClauses Number of clauses to allocate.
+  static OMPUnrollDirective *CreateEmpty(const ASTContext &C,
+                                         unsigned NumClauses);
+
+  /// Get the de-sugared associated loops after unrolling.
+  ///
+  /// This is only used if the unrolled loop becomes an associated loop of
+  /// another directive, otherwise the loop is emitted directly using loop
+  /// transformation metadata. When the unrolled loop cannot be used by another
+  /// directive (e.g. because of the full clause), the transformed stmt can also
+  /// be nullptr.
+  Stmt *getTransformedStmt() const {
+    return Data->getChildren()[TransformedStmtOffset];
+  }
+
+  /// Return the pre-init statements.
+  Stmt *getPreInits() const { return Data->getChildren()[PreInitsOffset]; }
+
+  static bool classof(const Stmt *T) {
+    return T->getStmtClass() == OMPUnrollDirectiveClass;
+  }
+};
+
 /// This represents '#pragma omp scan' directive.
 ///
 /// \code
 /// #pragma omp scan inclusive(a)
 /// \endcode
 /// In this example directive '#pragma omp scan' has clause 'inclusive' with
 /// list item 'a'.
 class OMPScanDirective final : public OMPExecutableDirective {

clang/include/clang/Basic/DiagnosticSemaKinds.td

def err_omp_interop_prefer_type : Error<
   "expression">;
 def err_omp_interop_bad_depend_clause : Error<
   "'depend' clause requires the 'targetsync' interop type">;
 def err_omp_interop_var_multiple_actions : Error<
   "interop variable %0 used in multiple action clauses">;
 def err_omp_dispatch_statement_call
     : Error<"statement after '#pragma omp dispatch' must be a direct call"
             " to a target function or an assignment to one">;
+def err_omp_unroll_full_variable_trip_count : Error<
+  "loop to be fully unrolled must have a constant trip count">;
+def note_omp_directive_here : Note<"'%0' directive found here">;
 } // end of OpenMP category
 
 let CategoryName = "Related Result Type Issue" in {
 // Objective-C related result type compatibility
 def warn_related_result_type_compatibility_class : Warning<
   "method is expected to return an instance of its class type "
   "%diff{$, but is declared to return $|"
   ", but is declared to return different type}0,1">;

clang/include/clang/Basic/StmtNodes.td

 // OpenMP Directives.
 def OMPCanonicalLoop : StmtNode<Stmt>;
 def OMPExecutableDirective : StmtNode<Stmt, 1>;
 def OMPLoopBasedDirective : StmtNode<OMPExecutableDirective, 1>;
 def OMPLoopDirective : StmtNode<OMPLoopBasedDirective, 1>;
 def OMPParallelDirective : StmtNode<OMPExecutableDirective>;
 def OMPSimdDirective : StmtNode<OMPLoopDirective>;
 def OMPTileDirective : StmtNode<OMPLoopBasedDirective>;
+def OMPUnrollDirective : StmtNode<OMPLoopBasedDirective>;
 def OMPForDirective : StmtNode<OMPLoopDirective>;
 def OMPForSimdDirective : StmtNode<OMPLoopDirective>;
 def OMPSectionsDirective : StmtNode<OMPExecutableDirective>;
 def OMPSectionDirective : StmtNode<OMPExecutableDirective>;
 def OMPSingleDirective : StmtNode<OMPExecutableDirective>;
 def OMPMasterDirective : StmtNode<OMPExecutableDirective>;
 def OMPCriticalDirective : StmtNode<OMPExecutableDirective>;
 def OMPParallelForDirective : StmtNode<OMPLoopDirective>;

clang/include/clang/Sema/Sema.h

 #include "clang/AST/ExprObjC.h"
 #include "clang/AST/ExprOpenMP.h"
 #include "clang/AST/ExternalASTSource.h"
 #include "clang/AST/LocInfoType.h"
 #include "clang/AST/MangleNumberingContext.h"
 #include "clang/AST/NSAPI.h"
 #include "clang/AST/PrettyPrinter.h"
 #include "clang/AST/StmtCXX.h"
+#include "clang/AST/StmtOpenMP.h"
 #include "clang/AST/TypeLoc.h"
 #include "clang/AST/TypeOrdering.h"
 #include "clang/Basic/BitmaskEnum.h"
 #include "clang/Basic/ExpressionTraits.h"
 #include "clang/Basic/Module.h"
 #include "clang/Basic/OpenCLOptions.h"
 #include "clang/Basic/OpenMPKinds.h"
 #include "clang/Basic/PragmaKinds.h"
private:
   /// \param Selector selects diagnostic message: 0 for type and 1 for
   ///        declaration.
   /// \return true if the type or declaration is disabled.
   template <typename T, typename DiagLocT, typename DiagInfoT, typename MapT>
   bool checkOpenCLDisabledTypeOrDecl(T D, DiagLocT DiagLoc, DiagInfoT DiagInfo,
                                      MapT &Map, unsigned Selector = 0,
                                      SourceRange SrcRange = SourceRange());
 
+  /// Analyzes and checks a loop nest for use by a loop transformation.
+  ///
+  /// \param Kind          The loop transformation directive kind.
+  /// \param NumLoops      How many nested loops the directive is expecting.
+  /// \param AStmt         Associated statement of the transformation directive.
+  /// \param LoopHelpers   [out] The loop analysis result.
+  /// \param Body          [out] The body code nested in \p NumLoops loop.
+  /// \param OriginalInits [out] Collection of statements and declarations that
+  ///                      must have been executed/declared before entering the
+  ///                      loop.
+  ///
+  /// \return Whether there was any error.
+  bool checkTransformableLoopNest(
+      OpenMPDirectiveKind Kind, Stmt *AStmt, int NumLoops,
+      SmallVectorImpl<OMPLoopBasedDirective::HelperExprs> &LoopHelpers,
+      Stmt *&Body,
+      SmallVectorImpl<std::vector<llvm::PointerUnion<Stmt *, Decl *>>>
+          &OriginalInits);
+
   /// Helper to keep information about the current `omp begin/end declare
   /// variant` nesting.
   struct OMPDeclareVariantScope {
     /// The associated OpenMP context selector.
     OMPTraitInfo *TI;
 
     /// The associated OpenMP context selector mangling.
     std::string NameSuffix;
public:
   ActOnOpenMPSimdDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
                            SourceLocation StartLoc, SourceLocation EndLoc,
                            VarsWithInheritedDSAType &VarsWithImplicitDSA);
   /// Called on well-formed '#pragma omp tile' after parsing of its clauses and
   /// the associated statement.
   StmtResult ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
                                       Stmt *AStmt, SourceLocation StartLoc,
                                       SourceLocation EndLoc);
+  /// Called on well-formed '#pragma omp unroll' after parsing of its clauses
+  /// and the associated statement.
+  StmtResult ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
+                                        Stmt *AStmt, SourceLocation StartLoc,
+                                        SourceLocation EndLoc);
   /// Called on well-formed '\#pragma omp for' after parsing
   /// of the associated statement.
   StmtResult
   ActOnOpenMPForDirective(ArrayRef<OMPClause *> Clauses, Stmt *AStmt,
                           SourceLocation StartLoc, SourceLocation EndLoc,
                           VarsWithInheritedDSAType &VarsWithImplicitDSA);
   /// Called on well-formed '\#pragma omp for simd' after parsing
   /// of the associated statement.
public:
   OMPClause *ActOnOpenMPSimdlenClause(Expr *Length, SourceLocation StartLoc,
                                       SourceLocation LParenLoc,
                                       SourceLocation EndLoc);
   /// Called on well-form 'sizes' clause.
   OMPClause *ActOnOpenMPSizesClause(ArrayRef<Expr *> SizeExprs,
                                     SourceLocation StartLoc,
                                     SourceLocation LParenLoc,
                                     SourceLocation EndLoc);
+  /// Called on well-form 'full' clauses.
+  OMPClause *ActOnOpenMPFullClause(SourceLocation StartLoc,
+                                   SourceLocation EndLoc);
+  /// Called on well-form 'partial' clauses.
+  OMPClause *ActOnOpenMPPartialClause(Expr *FactorExpr, SourceLocation StartLoc,
+                                      SourceLocation LParenLoc,
+                                      SourceLocation EndLoc);
   /// Called on well-formed 'collapse' clause.
   OMPClause *ActOnOpenMPCollapseClause(Expr *NumForLoops,
                                        SourceLocation StartLoc,
                                        SourceLocation LParenLoc,
                                        SourceLocation EndLoc);
   /// Called on well-formed 'ordered' clause.
   OMPClause *
   ActOnOpenMPOrderedClause(SourceLocation StartLoc, SourceLocation EndLoc,

clang/include/clang/Serialization/ASTBitCodes.h

enum StmtCode {
   STMT_SEH_FINALLY,                 // SEHFinallyStmt
   STMT_SEH_TRY,                     // SEHTryStmt
 
   // OpenMP directives
   STMT_OMP_CANONICAL_LOOP,
   STMT_OMP_PARALLEL_DIRECTIVE,
   STMT_OMP_SIMD_DIRECTIVE,
   STMT_OMP_TILE_DIRECTIVE,
+  STMT_OMP_UNROLL_DIRECTIVE,
   STMT_OMP_FOR_DIRECTIVE,
   STMT_OMP_FOR_SIMD_DIRECTIVE,
   STMT_OMP_SECTIONS_DIRECTIVE,
   STMT_OMP_SECTION_DIRECTIVE,
   STMT_OMP_SINGLE_DIRECTIVE,
   STMT_OMP_MASTER_DIRECTIVE,
   STMT_OMP_CRITICAL_DIRECTIVE,
   STMT_OMP_PARALLEL_FOR_DIRECTIVE,

clang/lib/AST/OpenMPClause.cpp

 }
 
 OMPSizesClause *OMPSizesClause::CreateEmpty(const ASTContext &C,
                                             unsigned NumSizes) {
   void *Mem = C.Allocate(totalSizeToAlloc<Expr *>(NumSizes));
   return new (Mem) OMPSizesClause(NumSizes);
 }
 
+OMPFullClause *OMPFullClause::Create(const ASTContext &C,
+                                     SourceLocation StartLoc,
+                                     SourceLocation EndLoc) {
+  OMPFullClause *Clause = CreateEmpty(C);
+  Clause->setLocStart(StartLoc);
+  Clause->setLocEnd(EndLoc);
+  return Clause;
+}
+
+OMPFullClause *OMPFullClause::CreateEmpty(const ASTContext &C) {
+  return new (C) OMPFullClause();
+}
+
+OMPPartialClause *OMPPartialClause::Create(const ASTContext &C,
+                                           SourceLocation StartLoc,
+                                           SourceLocation LParenLoc,
+                                           SourceLocation EndLoc,
+                                           Expr *Factor) {
+  OMPPartialClause *Clause = CreateEmpty(C);
+  Clause->setLocStart(StartLoc);
+  Clause->setLParenLoc(LParenLoc);
+  Clause->setLocEnd(EndLoc);
+  Clause->setFactor(Factor);
+  return Clause;
+}
+
+OMPPartialClause *OMPPartialClause::CreateEmpty(const ASTContext &C) {
+  return new (C) OMPPartialClause();
+}
+
 OMPAllocateClause *
 OMPAllocateClause::Create(const ASTContext &C, SourceLocation StartLoc,
                           SourceLocation LParenLoc, Expr *Allocator,
                           SourceLocation ColonLoc, SourceLocation EndLoc,
                           ArrayRef<Expr *> VL) {
   // Allocate space for private variables and initializer expressions.
   void *Mem = C.Allocate(totalSizeToAlloc<Expr *>(VL.size()));
   auto *Clause = new (Mem) OMPAllocateClause(StartLoc, LParenLoc, Allocator,
for (auto Size : Node->getSizesRefs()) {
     if (!First)
       OS << ", ";
     Size->printPretty(OS, nullptr, Policy, 0);
     First = false;
   }
   OS << ")";
 }
 
+void OMPClausePrinter::VisitOMPFullClause(OMPFullClause *Node) { OS << "full"; }
+
+void OMPClausePrinter::VisitOMPPartialClause(OMPPartialClause *Node) {
+  OS << "partial";
+
+  if (Expr *Factor = Node->getFactor()) {
+    OS << '(';
+    Factor->printPretty(OS, nullptr, Policy, 0);
+    OS << ')';
+  }
+}
+
 void OMPClausePrinter::VisitOMPAllocatorClause(OMPAllocatorClause *Node) {
   OS << "allocator(";
   Node->getAllocator()->printPretty(OS, nullptr, Policy, 0);
   OS << ")";
 }
 
 void OMPClausePrinter::VisitOMPCollapseClause(OMPCollapseClause *Node) {
   OS << "collapse(";

clang/lib/AST/StmtOpenMP.cpp

 
 bool OMPLoopBasedDirective::doForAllLoops(
     Stmt *CurStmt, bool TryImperfectlyNestedLoops, unsigned NumLoops,
     llvm::function_ref<bool(unsigned, Stmt *)> Callback,
     llvm::function_ref<void(OMPLoopBasedDirective *)>
         OnTransformationCallback) {
   CurStmt = CurStmt->IgnoreContainers();
   for (unsigned Cnt = 0; Cnt < NumLoops; ++Cnt) {
-    while (auto *Dir = dyn_cast<OMPTileDirective>(CurStmt)) {
+    while (true) {
+      auto *OrigStmt = CurStmt;
+      if (auto *Dir = dyn_cast<OMPTileDirective>(OrigStmt)) {
         OnTransformationCallback(Dir);
         CurStmt = Dir->getTransformedStmt();
+      } else if (auto *Dir = dyn_cast<OMPUnrollDirective>(OrigStmt)) {
+        OnTransformationCallback(Dir);
+        CurStmt = Dir->getTransformedStmt();
+      } else
+        break;
+
+      if (!CurStmt) {
+        // May happen if the loop transformation does not result in a generated
+        // loop (such as full unrolling).
+        CurStmt = OrigStmt;
+        break;
+      }
     }
     if (auto *CanonLoop = dyn_cast<OMPCanonicalLoop>(CurStmt))
       CurStmt = CanonLoop->getLoopStmt();
     if (Callback(Cnt, CurStmt))
       return false;
     // Move on to the next nested for loop, or to the loop body.
     // OpenMP [2.8.1, simd construct, Restrictions]
     // All loops associated with the construct must be perfectly nested; that

 OMPTileDirective *OMPTileDirective::CreateEmpty(const ASTContext &C,
                                                 unsigned NumClauses,
                                                 unsigned NumLoops) {
   return createEmptyDirective<OMPTileDirective>(
       C, NumClauses, /*HasAssociatedStmt=*/true, TransformedStmtOffset + 1,
       SourceLocation(), SourceLocation(), NumLoops);
 }
 
+OMPUnrollDirective *
+OMPUnrollDirective::Create(const ASTContext &C, SourceLocation StartLoc,
+                           SourceLocation EndLoc, ArrayRef<OMPClause *> Clauses,
+                           Stmt *AssociatedStmt, Stmt *TransformedStmt,
+                           Stmt *PreInits) {
+  OMPUnrollDirective *Dir = createDirective<OMPUnrollDirective>(
+      C, Clauses, AssociatedStmt, TransformedStmtOffset + 1, StartLoc, EndLoc);
+  Dir->setTransformedStmt(TransformedStmt);
+  Dir->setPreInits(PreInits);
+  return Dir;
+}
+
+OMPUnrollDirective *OMPUnrollDirective::CreateEmpty(const ASTContext &C,
+                                                    unsigned NumClauses) {
+  return createEmptyDirective<OMPUnrollDirective>(
+      C, NumClauses, /*HasAssociatedStmt=*/true, TransformedStmtOffset + 1,
+      SourceLocation(), SourceLocation());
+}
+
 OMPForSimdDirective *
 OMPForSimdDirective::Create(const ASTContext &C, SourceLocation StartLoc,
                             SourceLocation EndLoc, unsigned CollapsedNum,
                             ArrayRef<OMPClause *> Clauses, Stmt *AssociatedStmt,
                             const HelperExprs &Exprs) {
   auto *Dir = createDirective<OMPForSimdDirective>(
       C, Clauses, AssociatedStmt, numLoopChildren(CollapsedNum, OMPD_for_simd),
       StartLoc, EndLoc, CollapsedNum);

clang/lib/AST/StmtPrinter.cpp

void StmtPrinter::VisitOMPSimdDirective(OMPSimdDirective *Node) {
   PrintOMPExecutableDirective(Node);
 }
 
 void StmtPrinter::VisitOMPTileDirective(OMPTileDirective *Node) {
   Indent() << "#pragma omp tile";
   PrintOMPExecutableDirective(Node);
 }
 
+void StmtPrinter::VisitOMPUnrollDirective(OMPUnrollDirective *Node) {
+  Indent() << "#pragma omp unroll";
+  PrintOMPExecutableDirective(Node);
+}
+
 void StmtPrinter::VisitOMPForDirective(OMPForDirective *Node) {
   Indent() << "#pragma omp for";
   PrintOMPExecutableDirective(Node);
 }
 
 void StmtPrinter::VisitOMPForSimdDirective(OMPForSimdDirective *Node) {
   Indent() << "#pragma omp for simd";
   PrintOMPExecutableDirective(Node);

clang/lib/AST/StmtProfile.cpp

 }
 
 void OMPClauseProfiler::VisitOMPSimdlenClause(const OMPSimdlenClause *C) {
   if (C->getSimdlen())
     Profiler->VisitStmt(C->getSimdlen());
 }
 
 void OMPClauseProfiler::VisitOMPSizesClause(const OMPSizesClause *C) {
   for (auto E : C->getSizesRefs())
     if (E)
       Profiler->VisitExpr(E);
 }
 
+void OMPClauseProfiler::VisitOMPFullClause(const OMPFullClause *C) {}
+
+void OMPClauseProfiler::VisitOMPPartialClause(const OMPPartialClause *C) {
+  if (Expr *Factor = C->getFactor())
+    Profiler->VisitExpr(Factor);
+}
+
 void OMPClauseProfiler::VisitOMPAllocatorClause(const OMPAllocatorClause *C) {
   if (C->getAllocator())
     Profiler->VisitStmt(C->getAllocator());
 }
 
 void OMPClauseProfiler::VisitOMPCollapseClause(const OMPCollapseClause *C) {
   if (C->getNumForLoops())
     Profiler->VisitStmt(C->getNumForLoops());

 void StmtProfiler::VisitOMPSimdDirective(const OMPSimdDirective *S) {
   VisitOMPLoopDirective(S);
 }
 
 void StmtProfiler::VisitOMPTileDirective(const OMPTileDirective *S) {
   VisitOMPLoopBasedDirective(S);
 }
 
+void StmtProfiler::VisitOMPUnrollDirective(const OMPUnrollDirective *S) {
+  VisitOMPLoopBasedDirective(S);
+}
+
 void StmtProfiler::VisitOMPForDirective(const OMPForDirective *S) {
   VisitOMPLoopDirective(S);
 }
 
 void StmtProfiler::VisitOMPForSimdDirective(const OMPForSimdDirective *S) {
   VisitOMPLoopDirective(S);
 }
 

clang/lib/Basic/OpenMPKinds.cpp

return DKind == OMPD_simd || DKind == OMPD_for || DKind == OMPD_for_simd ||
          DKind == OMPD_target_parallel_for_simd || DKind == OMPD_target_simd ||
          DKind == OMPD_teams_distribute ||
          DKind == OMPD_teams_distribute_simd ||
          DKind == OMPD_teams_distribute_parallel_for_simd ||
          DKind == OMPD_teams_distribute_parallel_for ||
          DKind == OMPD_target_teams_distribute ||
          DKind == OMPD_target_teams_distribute_parallel_for ||
          DKind == OMPD_target_teams_distribute_parallel_for_simd ||
-         DKind == OMPD_target_teams_distribute_simd || DKind == OMPD_tile;
+         DKind == OMPD_target_teams_distribute_simd || DKind == OMPD_tile ||
+         DKind == OMPD_unroll;
 }
 
 bool clang::isOpenMPWorksharingDirective(OpenMPDirectiveKind DKind) {
   return DKind == OMPD_for || DKind == OMPD_for_simd ||
          DKind == OMPD_sections || DKind == OMPD_section ||
          DKind == OMPD_single || DKind == OMPD_parallel_for ||
          DKind == OMPD_parallel_for_simd || DKind == OMPD_parallel_sections ||
          DKind == OMPD_target_parallel_for ||
return Kind == OMPD_distribute_parallel_for ||
          Kind == OMPD_distribute_parallel_for_simd ||
          Kind == OMPD_teams_distribute_parallel_for_simd ||
          Kind == OMPD_teams_distribute_parallel_for ||
          Kind == OMPD_target_teams_distribute_parallel_for ||
          Kind == OMPD_target_teams_distribute_parallel_for_simd;
 }
 
 bool clang::isOpenMPLoopTransformationDirective(OpenMPDirectiveKind DKind) {
-  return DKind == OMPD_tile;
+  return DKind == OMPD_tile || DKind == OMPD_unroll;
 }
 
 void clang::getOpenMPCaptureRegions(
     SmallVectorImpl<OpenMPDirectiveKind> &CaptureRegions,
     OpenMPDirectiveKind DKind) {
   assert(unsigned(DKind) < llvm::omp::Directive_enumSize);
   switch (DKind) {
   case OMPD_parallel:
void clang::getOpenMPCaptureRegions(
   case OMPD_ordered:
   case OMPD_atomic:
   case OMPD_target_data:
   case OMPD_distribute_simd:
   case OMPD_dispatch:
     CaptureRegions.push_back(OMPD_unknown);
     break;
   case OMPD_tile:
+  case OMPD_unroll:
     // loop transformations do not introduce captures.
     break;
   case OMPD_threadprivate:
   case OMPD_allocate:
   case OMPD_taskyield:
   case OMPD_barrier:
   case OMPD_taskwait:
   case OMPD_cancellation_point:

clang/lib/CodeGen/CGOpenMPRuntime.cpp

emitNumTeamsForTargetDirective(CodeGenFunction &CGF,
   case OMPD_cancel:
   case OMPD_cancellation_point:
   case OMPD_ordered:
   case OMPD_threadprivate:
   case OMPD_allocate:
   case OMPD_task:
   case OMPD_simd:
   case OMPD_tile:
+  case OMPD_unroll:
   case OMPD_sections:
   case OMPD_section:
   case OMPD_single:
   case OMPD_master:
   case OMPD_critical:
   case OMPD_taskyield:
   case OMPD_barrier:
   case OMPD_taskwait:
emitNumThreadsForTargetDirective(CodeGenFunction &CGF,
   case OMPD_cancel:
   case OMPD_cancellation_point:
   case OMPD_ordered:
   case OMPD_threadprivate:
   case OMPD_allocate:
   case OMPD_task:
   case OMPD_simd:
   case OMPD_tile:
+  case OMPD_unroll:
   case OMPD_sections:
   case OMPD_section:
   case OMPD_single:
   case OMPD_master:
   case OMPD_critical:
   case OMPD_taskyield:
   case OMPD_barrier:
   case OMPD_taskwait:
if (const auto *NestedDir =
     case OMPD_cancel:
     case OMPD_cancellation_point:
     case OMPD_ordered:
     case OMPD_threadprivate:
     case OMPD_allocate:
     case OMPD_task:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_critical:
     case OMPD_taskyield:
     case OMPD_barrier:
     case OMPD_taskwait:
if (RequiresDeviceCodegen) {
     case OMPD_cancel:
     case OMPD_cancellation_point:
     case OMPD_ordered:
     case OMPD_threadprivate:
     case OMPD_allocate:
     case OMPD_task:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_critical:
     case OMPD_taskyield:
     case OMPD_barrier:
     case OMPD_taskwait:
auto &&ThenGen = [this, &D, Device, &InputInfo, &MapTypesArray,
     case OMPD_cancel:
     case OMPD_cancellation_point:
     case OMPD_ordered:
     case OMPD_threadprivate:
     case OMPD_allocate:
     case OMPD_task:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_critical:
     case OMPD_taskyield:
     case OMPD_barrier:
     case OMPD_taskwait:

clang/lib/CodeGen/CGStmt.cpp

case Stmt::OMPParallelDirectiveClass:
     EmitOMPParallelDirective(cast<OMPParallelDirective>(*S));
     break;
   case Stmt::OMPSimdDirectiveClass:
     EmitOMPSimdDirective(cast<OMPSimdDirective>(*S));
     break;
   case Stmt::OMPTileDirectiveClass:
     EmitOMPTileDirective(cast<OMPTileDirective>(*S));
     break;
+  case Stmt::OMPUnrollDirectiveClass:
+    EmitOMPUnrollDirective(cast<OMPUnrollDirective>(*S));
+    break;
   case Stmt::OMPForDirectiveClass:
     EmitOMPForDirective(cast<OMPForDirective>(*S));
     break;
   case Stmt::OMPForSimdDirectiveClass:
     EmitOMPForSimdDirective(cast<OMPForSimdDirective>(*S));
     break;
   case Stmt::OMPSectionsDirectiveClass:
     EmitOMPSectionsDirective(cast<OMPSectionsDirective>(*S));

clang/lib/CodeGen/CGStmtOpenMP.cpp

if (auto *LD = dyn_cast<OMPLoopDirective>(&S)) {
               CGF.EmitStmt(CXXFor->getRangeStmt());
               CGF.EmitStmt(CXXFor->getEndStmt());
             }
             return false;
           });
       PreInits = cast_or_null<DeclStmt>(LD->getPreInits());
     } else if (const auto *Tile = dyn_cast<OMPTileDirective>(&S)) {
       PreInits = cast_or_null<DeclStmt>(Tile->getPreInits());
+    } else if (const auto *Unroll = dyn_cast<OMPUnrollDirective>(&S)) {
+      PreInits = cast_or_null<DeclStmt>(Unroll->getPreInits());
     } else {
       llvm_unreachable("Unknown loop-based directive kind.");
     }
     if (PreInits) {
       for (const auto *I : PreInits->decls())
         CGF.EmitVarDecl(cast<VarDecl>(*I));
     }
     PreCondVars.restore(CGF);
if (const auto *CS = dyn_cast<CompoundStmt>(SimplifiedS)) {
     CodeGenFunction::LexicalScope Scope(CGF, S->getSourceRange());
     for (const Stmt *CurStmt : CS->body())
       emitBody(CGF, CurStmt, NextLoop, MaxLevel, Level);
     return;
   }
   if (SimplifiedS == NextLoop) {
     if (auto *Dir = dyn_cast<OMPTileDirective>(SimplifiedS))
       SimplifiedS = Dir->getTransformedStmt();
+    if (auto *Dir = dyn_cast<OMPUnrollDirective>(SimplifiedS))
+      SimplifiedS = Dir->getTransformedStmt();
     if (const auto *CanonLoop = dyn_cast<OMPCanonicalLoop>(SimplifiedS))
       SimplifiedS = CanonLoop->getLoopStmt();
     if (const auto *For = dyn_cast<ForStmt>(SimplifiedS)) {
       S = For->getBody();
     } else {
       assert(isa<CXXForRangeStmt>(SimplifiedS) &&
              "Expected canonical for loop or range-based for loop.");
       const auto *CXXFor = cast<CXXForRangeStmt>(SimplifiedS);

 }
 
 void CodeGenFunction::EmitOMPTileDirective(const OMPTileDirective &S) {
   // Emit the de-sugared statement.
   OMPTransformDirectiveScopeRAII TileScope(*this, &S);
   EmitStmt(S.getTransformedStmt());
 }
 
+void CodeGenFunction::EmitOMPUnrollDirective(const OMPUnrollDirective &S) {
+  // This function is only called if the unrolled loop is not consumed by any
+  // other loop-associated construct. Such a loop-associated construct will have
+  // used the transformed AST.
+
+  // Set the unroll metadata for the next emitted loop.
+  LoopStack.setUnrollState(LoopAttributes::Enable);
+
+  auto *FullClause = S.getSingleClause<OMPFullClause>();
+  if (FullClause)
+    LoopStack.setUnrollState(LoopAttributes::Full);
+
+  auto *PartialClause = S.getSingleClause<OMPPartialClause>();
+  uint64_t Factor = 0;
+  if (PartialClause) {
+    if (Expr *FactorExpr = PartialClause->getFactor()) {
+      RValue FactorRVal = EmitAnyExpr(FactorExpr, AggValueSlot::ignored(),
+                                      /*ignoreResult=*/true);
+      Factor =
+          cast<llvm::ConstantInt>(FactorRVal.getScalarVal())->getZExtValue();
+      assert(Factor >= 1 && "Only positive factors are valid");
+      LoopStack.setUnrollCount(Factor);
+    }
+  }
+
+  EmitStmt(S.getAssociatedStmt());
+}
+
 void CodeGenFunction::EmitOMPOuterLoop(
     bool DynamicOrOrdered, bool IsMonotonic, const OMPLoopDirective &S,
     CodeGenFunction::OMPPrivateScope &LoopScope,
     const CodeGenFunction::OMPLoopArguments &LoopArgs,
     const CodeGenFunction::CodeGenLoopTy &CodeGenLoop,
     const CodeGenFunction::CodeGenOrderedTy &CodeGenOrdered) {
   CGOpenMPRuntime &RT = CGM.getOpenMPRuntime();
 
static void emitOMPAtomicExpr(CodeGenFunction &CGF, OpenMPClauseKind Kind,
   case OMPC_firstprivate:
   case OMPC_lastprivate:
   case OMPC_reduction:
   case OMPC_task_reduction:
   case OMPC_in_reduction:
   case OMPC_safelen:
   case OMPC_simdlen:
   case OMPC_sizes:
+  case OMPC_full:
+  case OMPC_partial:
   case OMPC_allocator:
   case OMPC_allocate:
   case OMPC_collapse:
   case OMPC_default:
   case OMPC_seq_cst:
   case OMPC_acq_rel:
   case OMPC_acquire:
   case OMPC_release:

clang/lib/CodeGen/CodeGenFunction.h

   };
   void EmitOMPTargetTaskBasedDirective(const OMPExecutableDirective &S,
                                        const RegionCodeGenTy &BodyGen,
                                        OMPTargetDataInfo &InputInfo);
 
   void EmitOMPParallelDirective(const OMPParallelDirective &S);
   void EmitOMPSimdDirective(const OMPSimdDirective &S);
   void EmitOMPTileDirective(const OMPTileDirective &S);
+  void EmitOMPUnrollDirective(const OMPUnrollDirective &S);
   void EmitOMPForDirective(const OMPForDirective &S);
   void EmitOMPForSimdDirective(const OMPForSimdDirective &S);
   void EmitOMPSectionsDirective(const OMPSectionsDirective &S);
   void EmitOMPSectionDirective(const OMPSectionDirective &S);
   void EmitOMPSingleDirective(const OMPSingleDirective &S);
   void EmitOMPMasterDirective(const OMPMasterDirective &S);
   void EmitOMPMaskedDirective(const OMPMaskedDirective &S);
   void EmitOMPCriticalDirective(const OMPCriticalDirective &S);

clang/lib/Parse/ParseOpenMP.cpp

case OMPD_declare_target: {
     return Actions.BuildDeclaratorGroup(Decls);
   }
   case OMPD_unknown:
     Diag(Tok, diag::err_omp_unknown_directive);
     break;
   case OMPD_parallel:
   case OMPD_simd:
   case OMPD_tile:
+  case OMPD_unroll:
   case OMPD_task:
   case OMPD_taskyield:
   case OMPD_barrier:
   case OMPD_taskwait:
   case OMPD_taskgroup:
   case OMPD_flush:
   case OMPD_depobj:
   case OMPD_scan:
if ((StmtCtx & ParsedStmtContext::AllowStandaloneOpenMPDirectives) ==
           << getOpenMPDirectiveName(DKind) << 0;
     }
     HasAssociatedStatement = false;
     // Fall through for further analysis.
     LLVM_FALLTHROUGH;
   case OMPD_parallel:
   case OMPD_simd:
   case OMPD_tile:
+  case OMPD_unroll:
   case OMPD_for:
   case OMPD_for_simd:
   case OMPD_sections:
   case OMPD_single:
   case OMPD_section:
   case OMPD_master:
   case OMPD_critical:
   case OMPD_parallel_for:
OMPClause *Parser::ParseOpenMPClause(OpenMPDirectiveKind DKind,
   case OMPC_num_tasks:
   case OMPC_hint:
   case OMPC_allocator:
   case OMPC_depobj:
   case OMPC_detach:
   case OMPC_novariants:
   case OMPC_nocontext:
   case OMPC_filter:
+  case OMPC_partial:
     // OpenMP [2.5, Restrictions]
     //  At most one num_threads clause can appear on the directive.
     // OpenMP [2.8.1, simd construct, Restrictions]
     //  Only one safelen  clause can appear on a simd directive.
     //  Only one simdlen  clause can appear on a simd directive.
     //  Only one collapse clause can appear on a simd directive.
     // OpenMP [2.11.1, task Construct, Restrictions]
     //  At most one if clause can appear on the directive.
case OMPC_partial:
     // At most one novariants clause can appear on a dispatch directive.
     // At most one nocontext clause can appear on a dispatch directive.
     if (!FirstClause) {
       Diag(Tok, diag::err_omp_more_one_clause)
           << getOpenMPDirectiveName(DKind) << getOpenMPClauseName(CKind) << 0;
       ErrorFound = true;
     }
 
-    if (CKind == OMPC_ordered && PP.LookAhead(/*N=*/0).isNot(tok::l_paren))
+    if ((CKind == OMPC_ordered || CKind == OMPC_partial) &&
+        PP.LookAhead(/*N=*/0).isNot(tok::l_paren))
       Clause = ParseOpenMPClause(CKind, WrongDirective);
     else
       Clause = ParseOpenMPSingleExprClause(CKind, WrongDirective);
     break;
   case OMPC_default:
   case OMPC_proc_bind:
   case OMPC_atomic_default_mem_order:
   case OMPC_order:
OMPClause *Parser::ParseOpenMPClause(OpenMPDirectiveKind DKind,
   case OMPC_relaxed:
   case OMPC_threads:
   case OMPC_simd:
   case OMPC_nogroup:
   case OMPC_unified_address:
   case OMPC_unified_shared_memory:
   case OMPC_reverse_offload:
   case OMPC_dynamic_allocators:
+  case OMPC_full:
     // OpenMP [2.7.1, Restrictions, p. 9]
     //  Only one ordered clause can appear on a loop directive.
     // OpenMP [2.7.1, Restrictions, C/C++, p. 4]
     //  Only one nowait clause can appear on a for directive.
     // OpenMP [5.0, Requires directive, Restrictions]
     //   Each of the requires clauses can appear at most once on the directive.
     if (!FirstClause) {
       Diag(Tok, diag::err_omp_more_one_clause)
OMPClause *Parser::ParseOpenMPClause(OpenMPDirectiveKind DKind,
   case OMPC_threadprivate:
   case OMPC_uniform:
   case OMPC_match:
     if (!WrongDirective)
       Diag(Tok, diag::err_omp_unexpected_clause)
           << getOpenMPClauseName(CKind) << getOpenMPDirectiveName(DKind);
     SkipUntil(tok::comma, tok::annot_pragma_openmp_end, StopBeforeMatch);
     break;
   default:
     break;
   }
   return ErrorFound ? nullptr : Clause;
 }
 
 /// Parses simple expression in parens for single-expression clauses of OpenMP
 /// constructs.
 /// \param RLoc Returned location of right paren.
 ExprResult Parser::ParseOpenMPParensExpr(StringRef ClauseName,

clang/lib/Sema/SemaExceptionSpec.cpp

CanThrowResult Sema::canThrow(const Stmt *S) {
   case Stmt::OMPParallelMasterDirectiveClass:
   case Stmt::OMPParallelMasterTaskLoopDirectiveClass:
   case Stmt::OMPParallelMasterTaskLoopSimdDirectiveClass:
   case Stmt::OMPParallelSectionsDirectiveClass:
   case Stmt::OMPSectionDirectiveClass:
   case Stmt::OMPSectionsDirectiveClass:
   case Stmt::OMPSimdDirectiveClass:
   case Stmt::OMPTileDirectiveClass:
+  case Stmt::OMPUnrollDirectiveClass:
   case Stmt::OMPSingleDirectiveClass:
   case Stmt::OMPTargetDataDirectiveClass:
   case Stmt::OMPTargetDirectiveClass:
   case Stmt::OMPTargetEnterDataDirectiveClass:
   case Stmt::OMPTargetExitDataDirectiveClass:
   case Stmt::OMPTargetParallelDirectiveClass:
   case Stmt::OMPTargetParallelForDirectiveClass:
   case Stmt::OMPTargetParallelForSimdDirectiveClass:

clang/lib/Sema/SemaOpenMP.cpp

void VisitOMPExecutableDirective(OMPExecutableDirective *S) {
     VisitSubCaptures(S);
   }
 
   void VisitOMPTileDirective(OMPTileDirective *S) {
     // #pragma omp tile does not introduce data sharing.
     VisitStmt(S);
   }
 
+  void VisitOMPUnrollDirective(OMPUnrollDirective *S) {
+    // #pragma omp unroll does not introduce data sharing.
+    VisitStmt(S);
+  }
+
   void VisitStmt(Stmt *S) {
     for (Stmt *C : S->children()) {
       if (C) {
         // Check implicitly captured variables in the task-based directives to
         // check if they must be firstprivatized.
         Visit(C);
       }
     }
case OMPD_target_simd: {
     break;
   }
   case OMPD_atomic:
   case OMPD_critical:
   case OMPD_section:
   case OMPD_master:
   case OMPD_masked:
   case OMPD_tile:
+  case OMPD_unroll:
     break;
   case OMPD_simd:
   case OMPD_for:
   case OMPD_for_simd:
   case OMPD_sections:
   case OMPD_single:
   case OMPD_taskgroup:
   case OMPD_distribute:
Res = ActOnOpenMPSimdDirective(ClausesWithImplicit, AStmt, StartLoc, EndLoc,
                                    VarsWithInheritedDSA);
     if (LangOpts.OpenMP >= 50)
       AllowedNameModifiers.push_back(OMPD_simd);
     break;
   case OMPD_tile:
     Res =
         ActOnOpenMPTileDirective(ClausesWithImplicit, AStmt, StartLoc, EndLoc);
     break;
+  case OMPD_unroll:
+    Res = ActOnOpenMPUnrollDirective(ClausesWithImplicit, AStmt, StartLoc,
+                                     EndLoc);
+    break;
   case OMPD_for:
     Res = ActOnOpenMPForDirective(ClausesWithImplicit, AStmt, StartLoc, EndLoc,
                                   VarsWithInheritedDSA);
     break;
   case OMPD_for_simd:
     Res = ActOnOpenMPForSimdDirective(ClausesWithImplicit, AStmt, StartLoc,
                                       EndLoc, VarsWithInheritedDSA);
     if (LangOpts.OpenMP >= 50)
if (!OMPLoopBasedDirective::doForAllLoops(
               auto *DRE = cast<DeclRefExpr>(IterSpaces[Cnt].CounterVar);
               if (isa<OMPCapturedExprDecl>(DRE->getDecl())) {
                 Captures[DRE] = DRE;
               }
             }
             return false;
           },
           [&SemaRef, &Captures](OMPLoopBasedDirective *Transform) {
-            Stmt *DependentPreInits =
-                cast<OMPTileDirective>(Transform)->getPreInits();
+            Stmt *DependentPreInits;
+            if (auto *Dir = dyn_cast<OMPTileDirective>(Transform)) {
+              DependentPreInits = Dir->getPreInits();
+            } else if (auto *Dir = dyn_cast<OMPUnrollDirective>(Transform)) {
+              DependentPreInits = Dir->getPreInits();
+            } else
+              llvm_unreachable("Unexpected loop transformation");
+            if (!DependentPreInits)
+              return;
             for (Decl *C : cast<DeclStmt>(DependentPreInits)->getDeclGroup()) {
               auto *D = cast<VarDecl>(C);
               DeclRefExpr *Ref = buildDeclRefExpr(SemaRef, D, D->getType(),
                                                   Transform->getBeginLoc());
               Captures[Ref] = Ref;
             }
           }))
     return 0;
Sema::ActOnOpenMPParallelSectionsDirective(ArrayRef<OMPClause *> Clauses,
 
   setFunctionHasBranchProtectedScope();
 
   return OMPParallelSectionsDirective::Create(
       Context, StartLoc, EndLoc, Clauses, AStmt,
       DSAStack->getTaskgroupReductionRef(), DSAStack->isCancelRegion());
 }
 
-/// detach and mergeable clauses are mutially exclusive, check for it.
-static bool checkDetachMergeableClauses(Sema &S,
-                                        ArrayRef<OMPClause *> Clauses) {
+/// Find and diagnose mutually exclusive clause kinds.
+static bool checkMutuallyExclusiveClauses(
+    Sema &S, ArrayRef<OMPClause *> Clauses,
+    ArrayRef<OpenMPClauseKind> MutuallyExclusiveClauses) {
   const OMPClause *PrevClause = nullptr;
   bool ErrorFound = false;
   for (const OMPClause *C : Clauses) {
-    if (C->getClauseKind() == OMPC_detach ||
-        C->getClauseKind() == OMPC_mergeable) {
+    if (llvm::is_contained(MutuallyExclusiveClauses, C->getClauseKind())) {
       if (!PrevClause) {
         PrevClause = C;
       } else if (PrevClause->getClauseKind() != C->getClauseKind()) {
         S.Diag(C->getBeginLoc(), diag::err_omp_clauses_mutually_exclusive)
             << getOpenMPClauseName(C->getClauseKind())
             << getOpenMPClauseName(PrevClause->getClauseKind());
         S.Diag(PrevClause->getBeginLoc(), diag::note_omp_previous_clause)
             << getOpenMPClauseName(PrevClause->getClauseKind());
         ErrorFound = true;
       }
     }
   }
   return ErrorFound;
 }
 
 StmtResult Sema::ActOnOpenMPTaskDirective(ArrayRef<OMPClause *> Clauses,
                                           Stmt *AStmt, SourceLocation StartLoc,
                                           SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();
 
   // OpenMP 5.0, 2.10.1 task Construct
   // If a detach clause appears on the directive, then a mergeable clause cannot
   // appear on the same directive.
-  if (checkDetachMergeableClauses(*this, Clauses))
+  if (checkMutuallyExclusiveClauses(*this, Clauses,
+                                    {OMPC_detach, OMPC_mergeable}))
     return StmtError();
 
   auto *CS = cast<CapturedStmt>(AStmt);
   // 1.2.2 OpenMP Language Terminology
   // Structured block - An executable statement with a single entry at the
   // top and a single exit at the bottom.
   // The point of exit cannot be a branch out of the structured block.
   // longjmp() and throw() must not violate the entry/exit criteria.
if (DSAStack->isParentOrderedRegion()) {
     Diag(StartLoc, diag::err_omp_parent_cancel_region_ordered) << 1;
     return StmtError();
   }
   DSAStack->setParentCancelRegion(/*Cancel=*/true);
   return OMPCancelDirective::Create(Context, StartLoc, EndLoc, Clauses,
                                     CancelRegion);
 }
 
-static bool checkGrainsizeNumTasksClauses(Sema &S,
-                                          ArrayRef<OMPClause *> Clauses) {
-  const OMPClause *PrevClause = nullptr;
-  bool ErrorFound = false;
-  for (const OMPClause *C : Clauses) {
-    if (C->getClauseKind() == OMPC_grainsize ||
-        C->getClauseKind() == OMPC_num_tasks) {
-      if (!PrevClause)
-        PrevClause = C;
-      else if (PrevClause->getClauseKind() != C->getClauseKind()) {
-        S.Diag(C->getBeginLoc(), diag::err_omp_clauses_mutually_exclusive)
-            << getOpenMPClauseName(C->getClauseKind())
-            << getOpenMPClauseName(PrevClause->getClauseKind());
-        S.Diag(PrevClause->getBeginLoc(), diag::note_omp_previous_clause)
-            << getOpenMPClauseName(PrevClause->getClauseKind());
-        ErrorFound = true;
-      }
-    }
-  }
-  return ErrorFound;
-}
-
 static bool checkReductionClauseWithNogroup(Sema &S,
                                             ArrayRef<OMPClause *> Clauses) {
   const OMPClause *ReductionClause = nullptr;
   const OMPClause *NogroupClause = nullptr;
   for (const OMPClause *C : Clauses) {
     if (C->getClauseKind() == OMPC_reduction) {
       ReductionClause = C;
       if (NogroupClause)
if (NestedLoopCount == 0)
     return StmtError();
 
   assert((CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkGrainsizeNumTasksClauses(*this, Clauses))
+  if (checkMutuallyExclusiveClauses(*this, Clauses,
+                                    {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
   if (checkReductionClauseWithNogroup(*this, Clauses))
     return StmtError();
 
   setFunctionHasBranchProtectedScope();
for (OMPClause *C : Clauses) {
                                      DSAStack))
           return StmtError();
     }
   }
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkGrainsizeNumTasksClauses(*this, Clauses))
+  if (checkMutuallyExclusiveClauses(*this, Clauses,
+                                    {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
   if (checkReductionClauseWithNogroup(*this, Clauses))
     return StmtError();
   if (checkSimdlenSafelenSpecified(*this, Clauses))
     return StmtError();
if (NestedLoopCount == 0)
     return StmtError();
 
   assert((CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkGrainsizeNumTasksClauses(*this, Clauses))
+  if (checkMutuallyExclusiveClauses(*this, Clauses,
+                                    {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
   if (checkReductionClauseWithNogroup(*this, Clauses))
     return StmtError();
 
   setFunctionHasBranchProtectedScope();
for (OMPClause *C : Clauses) {
                                      DSAStack))
           return StmtError();
     }
   }
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkGrainsizeNumTasksClauses(*this, Clauses))
+  if (checkMutuallyExclusiveClauses(*this, Clauses,
+                                    {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
   if (checkReductionClauseWithNogroup(*this, Clauses))
     return StmtError();
   if (checkSimdlenSafelenSpecified(*this, Clauses))
     return StmtError();
if (NestedLoopCount == 0)
     return StmtError();
 
   assert((CurContext->isDependentContext() || B.builtAll()) &&
          "omp for loop exprs were not built");
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkGrainsizeNumTasksClauses(*this, Clauses))
+  if (checkMutuallyExclusiveClauses(*this, Clauses,
+                                    {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
   if (checkReductionClauseWithNogroup(*this, Clauses))
     return StmtError();
 
   setFunctionHasBranchProtectedScope();
for (OMPClause *C : Clauses) {
                                      DSAStack))
           return StmtError();
     }
   }
 
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // The grainsize clause and num_tasks clause are mutually exclusive and may
   // not appear on the same taskloop directive.
-  if (checkGrainsizeNumTasksClauses(*this, Clauses))
+  if (checkMutuallyExclusiveClauses(*this, Clauses,
+                                    {OMPC_grainsize, OMPC_num_tasks}))
     return StmtError();
   // OpenMP, [2.9.2 taskloop Construct, Restrictions]
   // If a reduction clause is present on the taskloop directive, the nogroup
   // clause must not be specified.
   if (checkReductionClauseWithNogroup(*this, Clauses))
     return StmtError();
   if (checkSimdlenSafelenSpecified(*this, Clauses))
     return StmtError();
StmtResult Sema::ActOnOpenMPTargetTeamsDistributeSimdDirective(
   if (checkSimdlenSafelenSpecified(*this, Clauses))
     return StmtError();
 
   setFunctionHasBranchProtectedScope();
   return OMPTargetTeamsDistributeSimdDirective::Create(
       Context, StartLoc, EndLoc, NestedLoopCount, Clauses, AStmt, B);
 }
 
+bool Sema::checkTransformableLoopNest(
+    OpenMPDirectiveKind Kind, Stmt *AStmt, int NumLoops,
+    SmallVectorImpl<OMPLoopBasedDirective::HelperExprs> &LoopHelpers,
+    Stmt *&Body,
+    SmallVectorImpl<std::vector<llvm::PointerUnion<Stmt *, Decl *>>>
+        &OriginalInits) {
+  OriginalInits.emplace_back();
+  bool Result = OMPLoopBasedDirective::doForAllLoops(
+      AStmt->IgnoreContainers(), /*TryImperfectlyNestedLoops=*/false, NumLoops,
+      [this, &LoopHelpers, &Body, &OriginalInits, Kind](unsigned Cnt,
+                                                        Stmt *CurStmt) {
+        VarsWithInheritedDSAType TmpDSA;
+        unsigned SingleNumLoops =
+            checkOpenMPLoop(Kind, nullptr, nullptr, CurStmt, *this, *DSAStack,
+                            TmpDSA, LoopHelpers[Cnt]);
+        if (SingleNumLoops == 0)
+          return true;
+        assert(SingleNumLoops == 1 && "Expect single loop iteration space");
+        if (auto *For = dyn_cast<ForStmt>(CurStmt)) {
+          OriginalInits.back().push_back(For->getInit());
+          Body = For->getBody();
+        } else {
+          assert(isa<CXXForRangeStmt>(CurStmt) &&
+                 "Expected canonical for or range-based for loops.");
+          auto *CXXFor = cast<CXXForRangeStmt>(CurStmt);
+          OriginalInits.back().push_back(CXXFor->getBeginStmt());
+          Body = CXXFor->getBody();
+        }
+        OriginalInits.emplace_back();
+        return false;
+      },
+      [&OriginalInits](OMPLoopBasedDirective *Transform) {
+        Stmt *DependentPreInits;
+        if (auto *Dir = dyn_cast<OMPTileDirective>(Transform))
+          DependentPreInits = Dir->getPreInits();
+        else if (auto *Dir = dyn_cast<OMPUnrollDirective>(Transform))
+          DependentPreInits = Dir->getPreInits();
+        else
+          llvm_unreachable("Unhandled loop transformation");
+        if (!DependentPreInits)
+          return;
+        for (Decl *C : cast<DeclStmt>(DependentPreInits)->getDeclGroup())
+          OriginalInits.back().push_back(C);
+      });
+  assert(OriginalInits.back().empty());
+  OriginalInits.pop_back();
+  return Result;
+}
+
 StmtResult Sema::ActOnOpenMPTileDirective(ArrayRef<OMPClause *> Clauses,
                                           Stmt *AStmt, SourceLocation StartLoc,
                                           SourceLocation EndLoc) {
   auto SizesClauses =
       OMPExecutableDirective::getClausesOfKind<OMPSizesClause>(Clauses);
   if (SizesClauses.empty()) {
     // A missing 'sizes' clause is already reported by the parser.
     return StmtError();
   }
   const OMPSizesClause *SizesClause = *SizesClauses.begin();
   unsigned NumLoops = SizesClause->getNumSizes();
 
   // Empty statement should only be possible if there already was an error.
   if (!AStmt)
     return StmtError();
 
   // Verify and diagnose loop nest.
   SmallVector<OMPLoopBasedDirective::HelperExprs, 4> LoopHelpers(NumLoops);
-  SmallVector<Decl *, 4> PreInits;
   Stmt *Body = nullptr;
-  SmallVector<Stmt *, 4> OriginalInits;
-  if (!OMPLoopBasedDirective::doForAllLoops(
-          AStmt->IgnoreContainers(), /*TryImperfectlyNestedLoops=*/false,
-          NumLoops,
-          [this, &LoopHelpers, &Body, &OriginalInits](unsigned Cnt,
-                                                      Stmt *CurStmt) {
-            VarsWithInheritedDSAType TmpDSA;
-            unsigned SingleNumLoops =
-                checkOpenMPLoop(OMPD_tile, nullptr, nullptr, CurStmt, *this,
-                                *DSAStack, TmpDSA, LoopHelpers[Cnt]);
-            if (SingleNumLoops == 0)
-              return true;
-            assert(SingleNumLoops == 1 && "Expect single loop iteration space");
-            if (auto *For = dyn_cast<ForStmt>(CurStmt)) {
-              OriginalInits.push_back(For->getInit());
-              Body = For->getBody();
-            } else {
-              assert(isa<CXXForRangeStmt>(CurStmt) &&
-                     "Expected canonical for or range-based for loops.");
-              auto *CXXFor = cast<CXXForRangeStmt>(CurStmt);
-              OriginalInits.push_back(CXXFor->getBeginStmt());
-              Body = CXXFor->getBody();
-            }
-            return false;
-          },
-          [this, &PreInits](OMPLoopBasedDirective *Transform) {
-            Stmt *DependentPreInits =
-                cast<OMPTileDirective>(Transform)->getPreInits();
-            for (Decl *C : cast<DeclStmt>(DependentPreInits)->getDeclGroup())
-              PreInits.push_back(C);
-          }))
+  SmallVector<std::vector<llvm::PointerUnion<Stmt *, Decl *>>, 4> OriginalInits;
+  if (!checkTransformableLoopNest(OMPD_tile, AStmt, NumLoops, LoopHelpers, Body,
+                                  OriginalInits))
     return StmtError();
 
   // Delay tiling to when template is completely instantiated.
   if (CurContext->isDependentContext())
     return OMPTileDirective::Create(Context, StartLoc, EndLoc, Clauses,
                                     NumLoops, AStmt, nullptr, nullptr);
 
+  SmallVector<Decl *, 4> PreInits;
 
   // Create iteration variables for the generated loops.
   SmallVector<VarDecl *, 4> FloorIndVars;
   SmallVector<VarDecl *, 4> TileIndVars;
   FloorIndVars.resize(NumLoops);
   TileIndVars.resize(NumLoops);
   for (unsigned I = 0; I < NumLoops; ++I) {
     OMPLoopBasedDirective::HelperExprs &LoopHelper = LoopHelpers[I];
-    if (auto *PI = cast_or_null<DeclStmt>(LoopHelper.PreInits))
-      PreInits.append(PI->decl_begin(), PI->decl_end());
+
     assert(LoopHelper.Counters.size() == 1 &&
            "Expect single-dimensional loop iteration space");
     auto *OrigCntVar = cast<DeclRefExpr>(LoopHelper.Counters.front());
     std::string OrigVarName = OrigCntVar->getNameInfo().getAsString();
     DeclRefExpr *IterVarRef = cast<DeclRefExpr>(LoopHelper.IterationVarRef);
     QualType CntTy = IterVarRef->getType();
 
     // Iteration variable for the floor (i.e. outer) loop.
// Iteration variable for the tile (i.e. inner) loop.
 
       // Reuse the iteration variable created by checkOpenMPLoop. It is also
       // used by the expressions to derive the original iteration variable's
       // value from the logical iteration number.
       auto *TileCntDecl = cast<VarDecl>(IterVarRef->getDecl());
       TileCntDecl->setDeclName(&PP.getIdentifierTable().get(TileCntName));
       TileIndVars[I] = TileCntDecl;
     }
-    if (auto *PI = dyn_cast_or_null<DeclStmt>(OriginalInits[I]))
+    for (auto &P : OriginalInits[I]) {
+      if (auto *D = P.dyn_cast<Decl *>())
+        PreInits.push_back(D);
+      else if (auto *PI = dyn_cast_or_null<DeclStmt>(P.dyn_cast<Stmt *>()))
+        PreInits.append(PI->decl_begin(), PI->decl_end());
+    }
+    if (auto *PI = cast_or_null<DeclStmt>(LoopHelper.PreInits))
       PreInits.append(PI->decl_begin(), PI->decl_end());
     // Gather declarations for the data members used as counters.
     for (Expr *CounterRef : LoopHelper.Counters) {
       auto *CounterDecl = cast<DeclRefExpr>(CounterRef)->getDecl();
       if (isa<OMPCapturedExprDecl>(CounterDecl))
         PreInits.push_back(CounterDecl);
     }
   }
Inner = new (Context)
                 LoopHelper.Init->getBeginLoc(), LoopHelper.Inc->getEndLoc());
   }
 
   return OMPTileDirective::Create(Context, StartLoc, EndLoc, Clauses, NumLoops,
                                   AStmt, Inner,
                                   buildPreInits(Context, PreInits));
 }
 
+/// Determine whether an expression is constant without emitting diagnostic.
+static bool isConstantExpression(Sema &SemaRef, Expr *E) {
+  struct ConstTripcountDiagnoser : public Sema::VerifyICEDiagnoser {
+    ConstTripcountDiagnoser() : VerifyICEDiagnoser(/*Suppress=*/true) {}
+    Sema::SemaDiagnosticBuilder diagnoseNotICE(Sema &S,
+                                               SourceLocation Loc) override {
+      llvm_unreachable("Diagnostic suppressed");
+    }
+  } Diagnoser;
+  ExprResult ConstResult = SemaRef.VerifyIntegerConstantExpression(
+      E, nullptr, Diagnoser, Sema::AllowFold);
+  return ConstResult.isUsable();
+}
+
+StmtResult Sema::ActOnOpenMPUnrollDirective(ArrayRef<OMPClause *> Clauses,
+                                            Stmt *AStmt,
+                                            SourceLocation StartLoc,
+                                            SourceLocation EndLoc) {
+  // Empty statement should only be possible if there already was an error.
+  if (!AStmt)
+    return StmtError();
+
+  if (checkMutuallyExclusiveClauses(*this, Clauses, {OMPC_partial, OMPC_full}))
+    return StmtError();
+
+  const OMPFullClause *FullClause =
+      OMPExecutableDirective::getSingleClause<OMPFullClause>(Clauses);
+  const OMPPartialClause *PartialClause =
+      OMPExecutableDirective::getSingleClause<OMPPartialClause>(Clauses);
+  assert(!(FullClause && PartialClause) &&
+         "mutual exclusivity must have been checked before");
+
+  constexpr unsigned NumLoops = 1;
+  Stmt *Body = nullptr;
+  SmallVector<OMPLoopBasedDirective::HelperExprs, NumLoops> LoopHelpers(
+      NumLoops);
+  SmallVector<std::vector<llvm::PointerUnion<Stmt *, Decl *>>, NumLoops + 1>
+      OriginalInits;
+  if (!checkTransformableLoopNest(OMPD_unroll, AStmt, NumLoops, LoopHelpers,
+                                  Body, OriginalInits))
+    return StmtError();
+
+  // Delay unrolling to when template is completely instantiated.
+  if (CurContext->isDependentContext())
+    return OMPUnrollDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
+                                      nullptr, nullptr);
+
+  OMPLoopBasedDirective::HelperExprs &LoopHelper = LoopHelpers.front();
+
+  if (FullClause) {
+    if (!isConstantExpression(*this, LoopHelper.NumIterations)) {
+      Diag(AStmt->getBeginLoc(), diag::err_omp_unroll_full_variable_trip_count);
+      Diag(FullClause->getBeginLoc(), diag::note_omp_directive_here)
+          << "#pragma omp unroll full";
+      return StmtError();
+    }
+  }
+
+  // The generated loop may only be passed to other loop-associated directive
+  // when a partial clause is specified. Without the requirement it is
+  // sufficient to generate loop unroll metadata at code-generation.
+  if (!PartialClause)
+    return OMPUnrollDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
+                                      nullptr, nullptr);
+
+  // Otherwise, we need to provide a de-sugared/transformed AST that can be
+  // associated with another loop directive.
+  //
+  // The canonical loop analysis return by checkTransformableLoopNest assumes
+  // the following structure to be the same loop without transformations or
+  // directives applied: \code OriginalInits; LoopHelper.PreInits;
+  // LoopHelper.Counters;
+  // for (; IV < LoopHelper.NumIterations; ++IV) {
+  //   LoopHelper.Updates;
+  //   Body;
+  // }
+  // \endcode
+  // where IV is a variable declared and initialized to 0 in LoopHelper.PreInits
+  // and referenced by LoopHelper.IterationVarRef.
+  //
+  // The unrolling directive transforms this into the following loop:
+  // \code
+  // OriginalInits;         \
+  // LoopHelper.PreInits;    > NewPreInits
+  // LoopHelper.Counters;   /
+  // for (auto UIV = 0; UIV < LoopHelper.NumIterations; UIV+=Factor) {
+  //   #pragma clang loop unroll_count(Factor)
+  //   for (IV = UIV; IV < UIV + Factor && UIV < LoopHelper.NumIterations; ++IV)
+  //   {
+  //     LoopHelper.Updates;
+  //     Body;
+  //   }
+  // }
+  // \endcode
+  // where UIV is a new logical iteration counter. IV must be the same VarDecl
+  // as the original LoopHelper.IterationVarRef because LoopHelper.Updates
+  // references it. If the partially unrolled loop is associated with another
+  // loop directive (like an OMPForDirective), it will use checkOpenMPLoop to
+  // analyze this loop, i.e. the outer loop must fulfill the constraints of an
+  // OpenMP canonical loop. The inner loop is not an associable canonical loop
+  // and only exists to defer its unrolling to LLVM's LoopUnroll instead of
+  // doing it in the frontend (by adding loop metadata). NewPreInits becomes a
+  // property of the OMPLoopBasedDirective instead of statements in
+  // CompoundStatement. This is to allow the loop to become a non-outermost loop
+  // of a canonical loop nest where these PreInits are emitted before the
+  // outermost directive.
+
+  // Determine the PreInit declarations.
+  SmallVector<Decl *, 4> PreInits;
+  assert(OriginalInits.size() == 1 &&
+         "Expecting a single-dimensional loop iteration space");
+  for (auto &P : OriginalInits[0]) {
+    if (auto *D = P.dyn_cast<Decl *>())
+      PreInits.push_back(D);
+    else if (auto *PI = dyn_cast_or_null<DeclStmt>(P.dyn_cast<Stmt *>()))
+      PreInits.append(PI->decl_begin(), PI->decl_end());
+  }
+  if (auto *PI = cast_or_null<DeclStmt>(LoopHelper.PreInits))
+    PreInits.append(PI->decl_begin(), PI->decl_end());
+  // Gather declarations for the data members used as counters.
+  for (Expr *CounterRef : LoopHelper.Counters) {
+    auto *CounterDecl = cast<DeclRefExpr>(CounterRef)->getDecl();
+    if (isa<OMPCapturedExprDecl>(CounterDecl))
+      PreInits.push_back(CounterDecl);
+  }
+
+  auto *IterationVarRef = cast<DeclRefExpr>(LoopHelper.IterationVarRef);
+  QualType IVTy = IterationVarRef->getType();
+  assert(LoopHelper.Counters.size() == 1 &&
+         "Expecting a single-dimensional loop iteration space");
+  auto *OrigVar = cast<DeclRefExpr>(LoopHelper.Counters.front());
+
+  // Determine the unroll factor.
+  uint64_t Factor;
+  SourceLocation FactorLoc;
+  if (auto *FactorConst =
+          cast_or_null<ConstantExpr>(PartialClause->getFactor())) {
+    Factor = FactorConst->getResultAsAPSInt().getZExtValue();
+    FactorLoc = PartialClause->getFactor()->getExprLoc();
+  } else {
+    // TODO: Use a better profitability model.
+    Factor = 2;
+  }
+  assert(Factor > 0 && "Expected positive unroll factor");
+  auto makeFactorExpr = [this, Factor, IVTy, FactorLoc]() {
+    return IntegerLiteral::Create(
+        Context, llvm::APInt(Context.getIntWidth(IVTy), Factor), IVTy,
+        FactorLoc);
+  };
+
+  // Iteration variable SourceLocations.
+
+  SourceLocation OrigVarLoc = OrigVar->getExprLoc();
+  SourceLocation OrigVarLocBegin = OrigVar->getBeginLoc();
+  SourceLocation OrigVarLocEnd = OrigVar->getEndLoc();
+
+  // Internal variable names.
+  std::string OrigVarName = OrigVar->getNameInfo().getAsString();
+  std::string OuterIVName = (Twine(".unrolled.iv.") + OrigVarName).str();
+  std::string InnerIVName = (Twine(".unroll_inner.iv.") + OrigVarName).str();
+  std::string InnerTripCountName =
+      (Twine(".unroll_inner.tripcount.") + OrigVarName).str();
+
+  // Create the iteration variable for the unrolled loop.
+  VarDecl *OuterIVDecl =
+      buildVarDecl(*this, {}, IVTy, OuterIVName, nullptr, OrigVar);
+  auto makeOuterRef = [this, OuterIVDecl, IVTy, OrigVarLoc]() {
+    return buildDeclRefExpr(*this, OuterIVDecl, IVTy, OrigVarLoc);
+  };
+
+  // Iteration variable for the inner loop: Reuse the iteration variable created
+  // by checkOpenMPLoop.
+  auto *InnerIVDecl = cast<VarDecl>(IterationVarRef->getDecl());
+  InnerIVDecl->setDeclName(&PP.getIdentifierTable().get(InnerIVName));
+  auto makeInnerRef = [this, InnerIVDecl, IVTy, OrigVarLoc]() {
+    return buildDeclRefExpr(*this, InnerIVDecl, IVTy, OrigVarLoc);
+  };
+
+  // Make a copy of the NumIterations expression for each use: By the AST
+  // constraints, every expression object in a DeclContext must be unique.
+  CaptureVars CopyTransformer(*this);
+  auto makeNumIterations = [&CopyTransformer, &LoopHelper]() -> Expr * {
+    return AssertSuccess(
+        CopyTransformer.TransformExpr(LoopHelper.NumIterations));
+  };
+
+  // Inner For init-statement: auto .unroll_inner.iv = .unrolled.iv
+  ExprResult LValueConv = DefaultLvalueConversion(makeOuterRef());
+  AddInitializerToDecl(InnerIVDecl, LValueConv.get(), /*DirectInit=*/false);
+  StmtResult InnerInit = new (Context)
+      DeclStmt(DeclGroupRef(InnerIVDecl), OrigVarLocBegin, OrigVarLocEnd);
+  if (!InnerInit.isUsable())
+    return StmtError();
+
+  // Inner For cond-expression:
+  // \code
+  //   .unroll_inner.iv < .unrolled.iv + Factor &&
+  //   .unroll_inner.iv < NumIterations
+  // \endcode
+  // This conjunction of two conditions allows ScalarEvolution to derive the
+  // maximum trip count of the inner loop.
+  ExprResult EndOfTile = BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(),
+                                    BO_Add, makeOuterRef(), makeFactorExpr());
+  if (!EndOfTile.isUsable())
+    return StmtError();
+  ExprResult InnerCond1 = BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(),
+                                     BO_LE, makeInnerRef(), EndOfTile.get());
+  if (!InnerCond1.isUsable())
+    return StmtError();
+  ExprResult InnerCond2 =
+      BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LE, makeInnerRef(),
+                 makeNumIterations());
+  if (!InnerCond2.isUsable())
+    return StmtError();
+  ExprResult InnerCond =
+      BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LAnd,
+                 InnerCond1.get(), InnerCond2.get());
+  if (!InnerCond.isUsable())
+    return StmtError();
+
+  // Inner For incr-statement: ++.unroll_inner.iv
+  ExprResult InnerIncr = BuildUnaryOp(CurScope, LoopHelper.Inc->getExprLoc(),
+                                      UO_PreInc, makeInnerRef());
+  if (!InnerIncr.isUsable())
+    return StmtError();
+
+  // Inner For statement.
+  SmallVector<Stmt *> InnerBodyStmts;
+  InnerBodyStmts.append(LoopHelper.Updates.begin(), LoopHelper.Updates.end());
+  InnerBodyStmts.push_back(Body);
+  CompoundStmt *InnerBody = CompoundStmt::Create(
+      Context, InnerBodyStmts, Body->getBeginLoc(), Body->getEndLoc());
+  ForStmt *InnerFor = new (Context)
+      ForStmt(Context, InnerInit.get(), InnerCond.get(), nullptr,
+              InnerIncr.get(), InnerBody, LoopHelper.Init->getBeginLoc(),
+              LoopHelper.Init->getBeginLoc(), LoopHelper.Inc->getEndLoc());
+
+  // Unroll metadata for the inner loop.
+  // This needs to take into account the remainder portion of the unrolled loop,
+  // hence `unroll(full)` does not apply here, even though the LoopUnroll pass
+  // supports multiple loop exits. Instead, unroll using a factor equivalent to
+  // the maximum trip count, which will also generate a remainder loop. Just
+  // `unroll(enable)` (which could have been useful if the user has not
+  // specified a concrete factor; even though the outer loop cannot be
+  // influenced anymore, would avoid more code bloat than necessary) will refuse
+  // the loop because "Won't unroll; remainder loop could not be generated when
+  // assuming runtime trip count". Even if it did work, it must not choose a
+  // larger unroll factor than the maximum loop length, or it would always just
+  // execute the remainder loop.
+  LoopHintAttr *UnrollHintAttr =
+      LoopHintAttr::CreateImplicit(Context, LoopHintAttr::UnrollCount,
+                                   LoopHintAttr::Numeric, makeFactorExpr());
+  AttributedStmt *InnerUnrolled =
+      AttributedStmt::Create(Context, StartLoc, {UnrollHintAttr}, InnerFor);
+
+  // Outer For init-statement: auto .unrolled.iv = 0
+  AddInitializerToDecl(
+      OuterIVDecl, ActOnIntegerConstant(LoopHelper.Init->getExprLoc(), 0).get(),
+      /*DirectInit=*/false);
+  StmtResult OuterInit = new (Context)
+      DeclStmt(DeclGroupRef(OuterIVDecl), OrigVarLocBegin, OrigVarLocEnd);
+  if (!OuterInit.isUsable())
+    return StmtError();
+
+  // Outer For cond-expression: .unrolled.iv < NumIterations
+  ExprResult OuterConde =
+      BuildBinOp(CurScope, LoopHelper.Cond->getExprLoc(), BO_LT, makeOuterRef(),
+                 makeNumIterations());
+  if (!OuterConde.isUsable())
+    return StmtError();
+
+  // Outer For incr-statement: .unrolled.iv += Factor
+  ExprResult OuterIncr =
+      BuildBinOp(CurScope, LoopHelper.Inc->getExprLoc(), BO_AddAssign,
+                 makeOuterRef(), makeFactorExpr());
+  if (!OuterIncr.isUsable())
+    return StmtError();
+
+  // Outer For statement.
+  ForStmt *OuterFor = new (Context)
+      ForStmt(Context, OuterInit.get(), OuterConde.get(), nullptr,
+              OuterIncr.get(), InnerUnrolled, LoopHelper.Init->getBeginLoc(),
+              LoopHelper.Init->getBeginLoc(), LoopHelper.Inc->getEndLoc());
+
+  return OMPUnrollDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
+                                    OuterFor, buildPreInits(Context, PreInits));
+}
+
 OMPClause *Sema::ActOnOpenMPSingleExprClause(OpenMPClauseKind Kind, Expr *Expr,
                                              SourceLocation StartLoc,
                                              SourceLocation LParenLoc,
                                              SourceLocation EndLoc) {
   OMPClause *Res = nullptr;
   switch (Kind) {
   case OMPC_final:
     Res = ActOnOpenMPFinalClause(Expr, StartLoc, LParenLoc, EndLoc);
case OMPC_novariants:
     Res = ActOnOpenMPNovariantsClause(Expr, StartLoc, LParenLoc, EndLoc);
     break;
   case OMPC_nocontext:
     Res = ActOnOpenMPNocontextClause(Expr, StartLoc, LParenLoc, EndLoc);
     break;
   case OMPC_filter:
     Res = ActOnOpenMPFilterClause(Expr, StartLoc, LParenLoc, EndLoc);
     break;
+  case OMPC_partial:
+    Res = ActOnOpenMPPartialClause(Expr, StartLoc, LParenLoc, EndLoc);
+    break;
   case OMPC_device:
   case OMPC_if:
   case OMPC_default:
   case OMPC_proc_bind:
   case OMPC_schedule:
   case OMPC_private:
   case OMPC_firstprivate:
   case OMPC_lastprivate:
case OMPC_if:
     case OMPD_declare_simd:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_teams:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_for:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
     case OMPD_taskgroup:
case OMPC_num_threads:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_teams:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_for:
     case OMPD_for_simd:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
case OMPC_num_teams:
     case OMPD_declare_simd:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_for:
     case OMPD_for_simd:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
case OMPC_thread_limit:
     case OMPD_declare_simd:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_for:
     case OMPD_for_simd:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
case OMPC_schedule:
     case OMPD_declare_simd:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
     case OMPD_taskgroup:
     case OMPD_distribute:
case OMPC_dist_schedule:
     case OMPD_declare_simd:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_for:
     case OMPD_for_simd:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
case OMPC_device:
     case OMPD_declare_simd:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_for:
     case OMPD_for_simd:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
case OMPC_priority:
     case OMPD_declare_simd:
     case OMPD_declare_variant:
     case OMPD_begin_declare_variant:
     case OMPD_end_declare_variant:
     case OMPD_declare_target:
     case OMPD_end_declare_target:
     case OMPD_simd:
     case OMPD_tile:
+    case OMPD_unroll:
     case OMPD_for:
     case OMPD_for_simd:
     case OMPD_sections:
     case OMPD_section:
     case OMPD_single:
     case OMPD_master:
     case OMPD_masked:
     case OMPD_critical:
if (!NumForLoopsResult.isUsable())
       return nullptr;
   }
 
   DSAStack->setAssociatedLoops(SizeExprs.size());
   return OMPSizesClause::Create(Context, StartLoc, LParenLoc, EndLoc,
                                 SizeExprs);
 }
 
+OMPClause *Sema::ActOnOpenMPFullClause(SourceLocation StartLoc,
+                                       SourceLocation EndLoc) {
+  return OMPFullClause::Create(Context, StartLoc, EndLoc);
+}
+
+OMPClause *Sema::ActOnOpenMPPartialClause(Expr *FactorExpr,
+                                          SourceLocation StartLoc,
+                                          SourceLocation LParenLoc,
+                                          SourceLocation EndLoc) {
+  if (FactorExpr) {
+    // If an argument is specified, it must be a constant (or an unevaluated
+    // template expression).
+    ExprResult FactorResult = VerifyPositiveIntegerConstantInClause(
+        FactorExpr, OMPC_partial, /*StrictlyPositive=*/true);
+    if (FactorResult.isInvalid())
+      return nullptr;
+    FactorExpr = FactorResult.get();
+  }
+
+  return OMPPartialClause::Create(Context, StartLoc, LParenLoc, EndLoc,
+                                  FactorExpr);
+}
+
 OMPClause *Sema::ActOnOpenMPSingleExprWithArgClause(
     OpenMPClauseKind Kind, ArrayRef<unsigned> Argument, Expr *Expr,
     SourceLocation StartLoc, SourceLocation LParenLoc,
     ArrayRef<SourceLocation> ArgumentLoc, SourceLocation DelimLoc,
     SourceLocation EndLoc) {
   OMPClause *Res = nullptr;
   switch (Kind) {
   case OMPC_schedule:
OMPClause *Sema::ActOnOpenMPClause(OpenMPClauseKind Kind,
   case OMPC_dynamic_allocators:
     Res = ActOnOpenMPDynamicAllocatorsClause(StartLoc, EndLoc);
     break;
   case OMPC_destroy:
     Res = ActOnOpenMPDestroyClause(/*InteropVar=*/nullptr, StartLoc,
                                    /*LParenLoc=*/SourceLocation(),
                                    /*VarLoc=*/SourceLocation(), EndLoc);
     break;
+  case OMPC_full:
+    Res = ActOnOpenMPFullClause(StartLoc, EndLoc);
+    break;
+  case OMPC_partial:
+    Res = ActOnOpenMPPartialClause(nullptr, StartLoc, /*LParenLoc=*/{}, EndLoc);
+    break;
   case OMPC_if:
   case OMPC_final:
   case OMPC_num_threads:
   case OMPC_safelen:
   case OMPC_simdlen:
   case OMPC_sizes:
   case OMPC_allocator:
   case OMPC_collapse:

clang/lib/Sema/TreeTransform.h

#include "llvm/Frontend/OpenMP/OMP.inc"
 
   OMPClause *RebuildOMPSizesClause(ArrayRef<Expr *> Sizes,
                                    SourceLocation StartLoc,
                                    SourceLocation LParenLoc,
                                    SourceLocation EndLoc) {
     return getSema().ActOnOpenMPSizesClause(Sizes, StartLoc, LParenLoc, EndLoc);
   }
 
+  /// Build a new OpenMP 'full' clause.
+  OMPClause *RebuildOMPFullClause(SourceLocation StartLoc,
+                                  SourceLocation EndLoc) {
+    return getSema().ActOnOpenMPFullClause(StartLoc, EndLoc);
+  }
+
+  /// Build a new OpenMP 'partial' clause.
+  OMPClause *RebuildOMPPartialClause(Expr *Factor, SourceLocation StartLoc,
+                                     SourceLocation LParenLoc,
+                                     SourceLocation EndLoc) {
+    return getSema().ActOnOpenMPPartialClause(Factor, StartLoc, LParenLoc,
+                                              EndLoc);
+  }
+
   /// Build a new OpenMP 'allocator' clause.
   ///
   /// By default, performs semantic analysis to build the new OpenMP clause.
   /// Subclasses may override this routine to provide different behavior.
   OMPClause *RebuildOMPAllocatorClause(Expr *A, SourceLocation StartLoc,
                                        SourceLocation LParenLoc,
                                        SourceLocation EndLoc) {
     return getSema().ActOnOpenMPAllocatorClause(A, StartLoc, LParenLoc, EndLoc);
getDerived().getSema().StartOpenMPDSABlock(D->getDirectiveKind(), DirName,
                                              nullptr, D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
   getDerived().getSema().EndOpenMPDSABlock(Res.get());
   return Res;
 }
 
 template <typename Derived>
 StmtResult
+TreeTransform<Derived>::TransformOMPUnrollDirective(OMPUnrollDirective *D) {
+  DeclarationNameInfo DirName;
+  getDerived().getSema().StartOpenMPDSABlock(D->getDirectiveKind(), DirName,
+                                             nullptr, D->getBeginLoc());
+  StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
+  getDerived().getSema().EndOpenMPDSABlock(Res.get());
+  return Res;
+}
+
+template <typename Derived>
+StmtResult
 TreeTransform<Derived>::TransformOMPForDirective(OMPForDirective *D) {
   DeclarationNameInfo DirName;
   getDerived().getSema().StartOpenMPDSABlock(OMPD_for, DirName, nullptr,
                                              D->getBeginLoc());
   StmtResult Res = getDerived().TransformOMPExecutableDirective(D);
   getDerived().getSema().EndOpenMPDSABlock(Res.get());
   return Res;
 }
OMPClause *TreeTransform<Derived>::TransformOMPSizesClause(OMPSizesClause *C) {
 
   if (!Changed && !getDerived().AlwaysRebuild())
     return C;
   return RebuildOMPSizesClause(TransformedSizes, C->getBeginLoc(),
                                C->getLParenLoc(), C->getEndLoc());
 }
 
 template <typename Derived>
+OMPClause *TreeTransform<Derived>::TransformOMPFullClause(OMPFullClause *C) {
+  if (!getDerived().AlwaysRebuild())
+    return C;
+  return RebuildOMPFullClause(C->getBeginLoc(), C->getEndLoc());
+}
+
+template <typename Derived>
+OMPClause *
+TreeTransform<Derived>::TransformOMPPartialClause(OMPPartialClause *C) {
+  ExprResult T = getDerived().TransformExpr(C->getFactor());
+  if (T.isInvalid())
+    return nullptr;
+  Expr *Factor = T.get();
+  bool Changed = Factor != C->getFactor();
+
+  if (!Changed && !getDerived().AlwaysRebuild())
+    return C;
+  return RebuildOMPPartialClause(Factor, C->getBeginLoc(), C->getLParenLoc(),
+                                 C->getEndLoc());
+}
+
+template <typename Derived>
 OMPClause *
 TreeTransform<Derived>::TransformOMPCollapseClause(OMPCollapseClause *C) {
   ExprResult E = getDerived().TransformExpr(C->getNumForLoops());
   if (E.isInvalid())
     return nullptr;
   return getDerived().RebuildOMPCollapseClause(
       E.get(), C->getBeginLoc(), C->getLParenLoc(), C->getEndLoc());
 }

clang/lib/Serialization/ASTReader.cpp

OMPClause *OMPClauseReader::readClause() {
   case llvm::omp::OMPC_simdlen:
     C = new (Context) OMPSimdlenClause();
     break;
   case llvm::omp::OMPC_sizes: {
     unsigned NumSizes = Record.readInt();
     C = OMPSizesClause::CreateEmpty(Context, NumSizes);
     break;
   }
+  case llvm::omp::OMPC_full:
+    C = OMPFullClause::CreateEmpty(Context);
+    break;
+  case llvm::omp::OMPC_partial:
+    C = OMPPartialClause::CreateEmpty(Context);
+    break;
   case llvm::omp::OMPC_allocator:
     C = new (Context) OMPAllocatorClause();
     break;
   case llvm::omp::OMPC_collapse:
     C = new (Context) OMPCollapseClause();
     break;
   case llvm::omp::OMPC_default:
     C = new (Context) OMPDefaultClause();

 }
 
 void OMPClauseReader::VisitOMPSizesClause(OMPSizesClause *C) {
   for (Expr *&E : C->getSizesRefs())
     E = Record.readSubExpr();
   C->setLParenLoc(Record.readSourceLocation());
 }
 
+void OMPClauseReader::VisitOMPFullClause(OMPFullClause *C) {}
+
+void OMPClauseReader::VisitOMPPartialClause(OMPPartialClause *C) {
+  C->setFactor(Record.readSubExpr());
+  C->setLParenLoc(Record.readSourceLocation());
+}
+
 void OMPClauseReader::VisitOMPAllocatorClause(OMPAllocatorClause *C) {
   C->setAllocator(Record.readExpr());
   C->setLParenLoc(Record.readSourceLocation());
 }
 
 void OMPClauseReader::VisitOMPCollapseClause(OMPCollapseClause *C) {
   C->setNumForLoops(Record.readSubExpr());
   C->setLParenLoc(Record.readSourceLocation());

clang/lib/Serialization/ASTReaderStmt.cpp

 void ASTStmtReader::VisitOMPSimdDirective(OMPSimdDirective *D) {
   VisitOMPLoopDirective(D);
 }
 
 void ASTStmtReader::VisitOMPTileDirective(OMPTileDirective *D) {
   VisitOMPLoopBasedDirective(D);
 }
 
+void ASTStmtReader::VisitOMPUnrollDirective(OMPUnrollDirective *D) {
+  VisitOMPLoopBasedDirective(D);
+}
+
 void ASTStmtReader::VisitOMPForDirective(OMPForDirective *D) {
   VisitOMPLoopDirective(D);
   D->setHasCancel(Record.readBool());
 }
 
 void ASTStmtReader::VisitOMPForSimdDirective(OMPForSimdDirective *D) {
   VisitOMPLoopDirective(D);
 }
while (true) {
 
     case STMT_OMP_TILE_DIRECTIVE: {
       unsigned NumLoops = Record[ASTStmtReader::NumStmtFields];
       unsigned NumClauses = Record[ASTStmtReader::NumStmtFields + 1];
       S = OMPTileDirective::CreateEmpty(Context, NumClauses, NumLoops);
       break;
     }
 
+    case STMT_OMP_UNROLL_DIRECTIVE: {
+      unsigned NumLoops = Record[ASTStmtReader::NumStmtFields];
+      assert(NumLoops == 1 && "Unroll directive accepts only a single loop");
+      unsigned NumClauses = Record[ASTStmtReader::NumStmtFields + 1];
+      S = OMPUnrollDirective::CreateEmpty(Context, NumClauses);
+      break;
+    }
+
     case STMT_OMP_FOR_DIRECTIVE: {
       unsigned CollapsedNum = Record[ASTStmtReader::NumStmtFields];
       unsigned NumClauses = Record[ASTStmtReader::NumStmtFields + 1];
       S = OMPForDirective::CreateEmpty(Context, NumClauses, CollapsedNum,
                                        Empty);
       break;
     }
 

clang/lib/Serialization/ASTWriter.cpp

 
 void OMPClauseWriter::VisitOMPSizesClause(OMPSizesClause *C) {
   Record.push_back(C->getNumSizes());
   for (Expr *Size : C->getSizesRefs())
     Record.AddStmt(Size);
   Record.AddSourceLocation(C->getLParenLoc());
 }
 
+void OMPClauseWriter::VisitOMPFullClause(OMPFullClause *C) {}
+
+void OMPClauseWriter::VisitOMPPartialClause(OMPPartialClause *C) {
+  Record.AddStmt(C->getFactor());
+  Record.AddSourceLocation(C->getLParenLoc());
+}
+
 void OMPClauseWriter::VisitOMPAllocatorClause(OMPAllocatorClause *C) {
   Record.AddStmt(C->getAllocator());
   Record.AddSourceLocation(C->getLParenLoc());
 }
 
 void OMPClauseWriter::VisitOMPCollapseClause(OMPCollapseClause *C) {
   Record.AddStmt(C->getNumForLoops());
   Record.AddSourceLocation(C->getLParenLoc());

clang/lib/Serialization/ASTWriterStmt.cpp

void ASTStmtWriter::VisitOMPSimdDirective(OMPSimdDirective *D) {
   Code = serialization::STMT_OMP_SIMD_DIRECTIVE;
 }
 
 void ASTStmtWriter::VisitOMPTileDirective(OMPTileDirective *D) {
   VisitOMPLoopBasedDirective(D);
   Code = serialization::STMT_OMP_TILE_DIRECTIVE;
 }
 
+void ASTStmtWriter::VisitOMPUnrollDirective(OMPUnrollDirective *D) {
+  VisitOMPLoopBasedDirective(D);
+  Code = serialization::STMT_OMP_UNROLL_DIRECTIVE;
+}
+
 void ASTStmtWriter::VisitOMPForDirective(OMPForDirective *D) {
   VisitOMPLoopDirective(D);
   Record.writeBool(D->hasCancel());
   Code = serialization::STMT_OMP_FOR_DIRECTIVE;
 }
 
 void ASTStmtWriter::VisitOMPForSimdDirective(OMPForSimdDirective *D) {
   VisitOMPLoopDirective(D);

clang/lib/StaticAnalyzer/Core/ExprEngine.cpp

switch (S->getStmtClass()) {
     case Stmt::OMPTargetTeamsDistributeDirectiveClass:
     case Stmt::OMPTargetTeamsDistributeParallelForDirectiveClass:
     case Stmt::OMPTargetTeamsDistributeParallelForSimdDirectiveClass:
     case Stmt::OMPTargetTeamsDistributeSimdDirectiveClass:
     case Stmt::OMPTileDirectiveClass:
     case Stmt::OMPInteropDirectiveClass:
     case Stmt::OMPDispatchDirectiveClass:
     case Stmt::OMPMaskedDirectiveClass:
-    case Stmt::CapturedStmtClass: {
+    case Stmt::CapturedStmtClass:
+    case Stmt::OMPUnrollDirectiveClass: {
       const ExplodedNode *node = Bldr.generateSink(S, Pred, Pred->getState());
       Engine.addAbortedBlock(node, currBldrCtx->getBlock());
       break;
     }
 
     case Stmt::ParenExprClass:
       llvm_unreachable("ParenExprs already handled.");
     case Stmt::GenericSelectionExprClass:

clang/test/OpenMP/tile_codegen.cpp

 // IR-NEXT:  entry:
 // IR-NEXT:    call void @_Z4foo7IiLi3ELi5EEvT_S0_(i32 0, i32 42)
 // IR-NEXT:    ret void
 //
 // IR-LABEL: define linkonce_odr void @_Z4foo7IiLi3ELi5EEvT_S0_(
 // IR-NEXT:  entry:
 // IR-NEXT:    [[START_ADDR:%.*]] = alloca i32, align 4
 // IR-NEXT:    [[END_ADDR:%.*]] = alloca i32, align 4
+// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
 // IR-NEXT:    [[CAPTURE_EXPR:%.+]] = alloca i32, align 4
 // IR-NEXT:    [[CAPTURE_EXPR1:%.+]] = alloca i32, align 4
 // IR-NEXT:    [[CAPTURE_EXPR2:%.+]] = alloca i32, align 4
-// IR-NEXT:    [[I:%.*]] = alloca i32, align 4
 // IR-NEXT:    [[DOTFLOOR_0_IV_I:%.*]] = alloca i32, align 4
 // IR-NEXT:    [[DOTTILE_0_IV_I:%.*]] = alloca i32, align 4
 // IR-NEXT:    store i32 [[START:%.*]], i32* [[START_ADDR]], align 4
 // IR-NEXT:    store i32 [[END:%.*]], i32* [[END_ADDR]], align 4
+// IR-NEXT:    [[TMP4:%.+]] = load i32, i32* [[START_ADDR]], align 4
+// IR-NEXT:    store i32 [[TMP4]], i32* [[I]], align 4
 // IR-NEXT:    [[TMP0:%.+]] = load i32, i32* [[START_ADDR]], align 4
 // IR-NEXT:    store i32 [[TMP0]], i32* [[CAPTURE_EXPR]], align 4
 // IR-NEXT:    [[TMP1:%.+]] = load i32, i32* [[END_ADDR]], align 4
 // IR-NEXT:    store i32 [[TMP1]], i32* [[CAPTURE_EXPR1]], align 4
 // IR-NEXT:    [[TMP2:%.+]] = load i32, i32* [[CAPTURE_EXPR1]], align 4
 // IR-NEXT:    [[TMP3:%.+]] = load i32, i32* [[CAPTURE_EXPR]], align 4
 // IR-NEXT:    [[SUB:%.+]] = sub i32 [[TMP2]], [[TMP3]]
 // IR-NEXT:    [[SUB3:%.+]] = sub i32 [[SUB]], 1
 // IR-NEXT:    [[ADD:%.+]] = add i32 [[SUB3]], 3
 // IR-NEXT:    [[DIV:%.+]] = udiv i32 [[ADD]], 3
 // IR-NEXT:    [[SUB4:%.+]] = sub i32 [[DIV]], 1
 // IR-NEXT:    store i32 [[SUB4]], i32* [[CAPTURE_EXPR2]], align 4
-// IR-NEXT:    [[TMP4:%.+]] = load i32, i32* [[START_ADDR]], align 4
-// IR-NEXT:    store i32 [[TMP4]], i32* [[I]], align 4
 // IR-NEXT:    store i32 0, i32* [[DOTFLOOR_0_IV_I]], align 4
 // IR-NEXT:    br label %[[FOR_COND:.*]]
 // IR:         [[FOR_COND]]:
 // IR-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTFLOOR_0_IV_I]], align 4
 // IR-NEXT:    [[TMP6:%.+]] = load i32, i32* [[CAPTURE_EXPR2]], align 4
 // IR-NEXT:    [[ADD3:%.*]] = add i32 [[TMP6]], 1
 // IR-NEXT:    [[CMP:%.*]] = icmp ult i32 [[TMP0]], [[ADD3]]
 // IR-NEXT:    br i1 [[CMP]], label %[[FOR_BODY:.*]], label %[[FOR_END25:.*]]

clang/test/OpenMP/tile_codegen_for_dependent.cpp

 // Check code generation
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 
 // Check same results after serialization round-trip
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 // expected-no-diagnostics
 
 // The loop trip count used by #pragma omp for depends on code generated
 // by #pragma omp file. Check that theses PreInits are emitted before
 // the code generated by #pragma omp for.
 
 #ifndef HEADER
 #define HEADER
 
 // placeholder for loop body code.
-
 extern "C" void body(...) {}
 
 
-
 // IR-LABEL: @func(
 // IR-NEXT:  [[ENTRY:.*]]:
 // IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IV:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_1:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_2:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_3:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_6:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_8:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTFLOOR_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_LB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_UB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_STRIDE:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IS_LAST:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTFLOOR_0_IV_I12:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTTILE_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP0:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
 // IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
 // IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
 // IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP1]], i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP4]], %[[TMP5]]
+// IR-NEXT:    store i32 %[[TMP1]], i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP4]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP5]], %[[TMP6]]
 // IR-NEXT:    %[[SUB4:.+]] = sub i32 %[[SUB]], 1
-// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP6]]
 // IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP7]]
+// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP7]]
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP8]]
 // IR-NEXT:    %[[SUB5:.+]] = sub i32 %[[DIV]], 1
 // IR-NEXT:    store i32 %[[SUB5]], i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP8]], i32* %[[I]], align 4
 // IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
 // IR-NEXT:    %[[ADD7:.+]] = add i32 %[[TMP9]], 1
 // IR-NEXT:    store i32 %[[ADD7]], i32* %[[DOTCAPTURE_EXPR_6]], align 4
 // IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
 // IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP10]], -3
 // IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 4
 // IR-NEXT:    %[[SUB11:.+]] = sub i32 %[[DIV10]], 1
 // IR-NEXT:    store i32 %[[SUB11]], i32* %[[DOTCAPTURE_EXPR_8]], align 4

clang/test/OpenMP/tile_codegen_tile_for.cpp

 
 // IR-LABEL: @func(
 // IR-NEXT:  [[ENTRY:.*]]:
 // IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IV:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_1:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_2:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_3:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTFLOOR_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_6:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_8:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_12:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_14:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTFLOOR_0_IV__FLOOR_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_LB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_UB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_STRIDE:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IS_LAST:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTTILE_0_IV__FLOOR_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTTILE_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP0:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
 // IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
 // IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
 // IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP1]], i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP4]], %[[TMP5]]
+// IR-NEXT:    store i32 %[[TMP1]], i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP4]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP5]], %[[TMP6]]
 // IR-NEXT:    %[[SUB4:.+]] = sub i32 %[[SUB]], 1
-// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP6]]
 // IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP7]]
+// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP7]]
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP8]]
 // IR-NEXT:    %[[SUB5:.+]] = sub i32 %[[DIV]], 1
 // IR-NEXT:    store i32 %[[SUB5]], i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP8]], i32* %[[I]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV_I]], align 4
 // IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
 // IR-NEXT:    %[[ADD7:.+]] = add i32 %[[TMP9]], 1
 // IR-NEXT:    store i32 %[[ADD7]], i32* %[[DOTCAPTURE_EXPR_6]], align 4
 // IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
 // IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP10]], -3
 // IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 4
 // IR-NEXT:    %[[SUB11:.+]] = sub i32 %[[DIV10]], 1
 // IR-NEXT:    store i32 %[[SUB11]], i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV_I]], align 4
 // IR-NEXT:    %[[TMP11:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
 // IR-NEXT:    %[[ADD13:.+]] = add i32 %[[TMP11]], 1
 // IR-NEXT:    store i32 %[[ADD13]], i32* %[[DOTCAPTURE_EXPR_12]], align 4
 // IR-NEXT:    %[[TMP12:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_12]], align 4
 // IR-NEXT:    %[[SUB15:.+]] = sub i32 %[[TMP12]], -2
 // IR-NEXT:    %[[DIV16:.+]] = udiv i32 %[[SUB15]], 3
 // IR-NEXT:    %[[SUB17:.+]] = sub i32 %[[DIV16]], 1
 // IR-NEXT:    store i32 %[[SUB17]], i32* %[[DOTCAPTURE_EXPR_14]], align 4

 #pragma omp for
 #pragma omp tile sizes(3)
 #pragma omp tile sizes(4)
   for (int i = start; i < end; i += step)
     body(start, end, step, i);
 }
 
 #endif /* HEADER */
+// IR: ![[META0:[0-9]+]] = !{i32 1, !"wchar_size", i32 4}
+// IR: ![[META1:[0-9]+]] = !{!"clang version {{[^"]*}}"}
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOP4]] = distinct !{![[LOOP4]], ![[LOOPPROP3]]}

clang/test/OpenMP/unroll_ast_print.cpp

+// Check no warnings/errors
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -fsyntax-only -verify %s
+// expected-no-diagnostics
+
+// Check AST and unparsing
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -ast-dump  %s | FileCheck %s --check-prefix=DUMP
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -ast-print %s | FileCheck %s --check-prefix=PRINT --match-full-lines
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -ast-dump-all %s | FileCheck %s --check-prefix=DUMP
+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -ast-print    %s | FileCheck %s --check-prefix=PRINT --match-full-lines
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+void body(...);
+
+
+// PRINT-LABEL: void func_unroll() {
+// DUMP-LABEL:  FunctionDecl {{.*}} func_unroll
+void func_unroll() {
+  // PRINT:  #pragma omp unroll
+  // DUMP:   OMPUnrollDirective
+  #pragma omp unroll
+  // PRINT-NEXT: for (int i = 7; i < 17; i += 3)
+  // DUMP-NEXT: ForStmt
+  for (int i = 7; i < 17; i += 3)
+    // PRINT-NEXT: body(i);
+    // DUMP: CallExpr
+    body(i);
+}
+
+
+// PRINT-LABEL: void func_unroll_full() {
+// DUMP-LABEL:  FunctionDecl {{.*}} func_unroll_full 
+void func_unroll_full() {
+  // PRINT:     #pragma omp unroll full
+  // DUMP:      OMPUnrollDirective
+  // DUMP-NEXT:   OMPFullClause
+  #pragma omp unroll full
+  // PRINT-NEXT: for (int i = 7; i < 17; i += 3)
+  // DUMP-NEXT: ForStmt
+  for (int i = 7; i < 17; i += 3)
+    // PRINT-NEXT: body(i);
+    // DUMP: CallExpr
+    body(i);
+}
+
+
+// PRINT-LABEL: void func_unroll_partial() {
+// DUMP-LABEL:  FunctionDecl {{.*}} func_unroll_partial 
+void func_unroll_partial() {
+  // PRINT:     #pragma omp unroll partial
+  // DUMP:      OMPUnrollDirective
+  // DUMP-NEXT:   OMPPartialClause
+  // DUMP-NEXT:     <<<NULL>>>
+  #pragma omp unroll partial
+  // PRINT-NEXT: for (int i = 7; i < 17; i += 3)
+  // DUMP-NEXT: ForStmt
+  for (int i = 7; i < 17; i += 3)
+    // PRINT: body(i);
+    // DUMP: CallExpr
+    body(i);
+}
+
+
+// PRINT-LABEL: void func_unroll_partial_factor() {
+// DUMP-LABEL:  FunctionDecl {{.*}} func_unroll_partial_factor 
+void func_unroll_partial_factor() {
+  // PRINT:     #pragma omp unroll partial(4)
+  // DUMP:      OMPUnrollDirective
+  // DUMP-NEXT:   OMPPartialClause
+  // DUMP-NEXT:     ConstantExpr
+  // DUMP-NEXT:       value: Int 4
+  // DUMP-NEXT:       IntegerLiteral {{.*}} 4
+  #pragma omp unroll partial(4)
+  // PRINT-NEXT: for (int i = 7; i < 17; i += 3)
+  // DUMP-NEXT: ForStmt
+  for (int i = 7; i < 17; i += 3)
+    // PRINT-NEXT: body(i);
+    // DUMP: CallExpr
+    body(i);
+}
+
+
+// PRINT-LABEL: void func_unroll_partial_factor_for() {
+// DUMP-LABEL:  FunctionDecl {{.*}} func_unroll_partial_factor_for 
+void func_unroll_partial_factor_for() {
+  // PRINT:     #pragma omp for
+  // DUMP:      OMPForDirective
+  #pragma omp for
+  // PRINT:       #pragma omp unroll partial(2)
+  // DUMP:        OMPUnrollDirective
+  // DUMP-NEXT:     OMPPartialClause
+  #pragma omp unroll partial(2)
+  // PRINT-NEXT: for (int i = 7; i < 17; i += 3)
+  // DUMP: ForStmt
+  for (int i = 7; i < 17; i += 3)
+    // PRINT-NEXT: body(i);
+    // DUMP: CallExpr
+    body(i);
+}
+
+
+// PRINT-LABEL: template <typename T, T Start, T End, T Step, int Factor> void unroll_templated() {
+// DUMP-LABEL:  FunctionTemplateDecl {{.*}} unroll_templated
+template<typename T, T Start, T End, T Step, int Factor>
+void unroll_templated() {
+  // PRINT: #pragma omp unroll partial(Factor)
+  // DUMP:      OMPUnrollDirective
+  // DUMP-NEXT: OMPPartialClause
+  // DUMP-NEXT:   DeclRefExpr {{.*}} 'Factor' 'int'
+  #pragma omp unroll partial(Factor)
+    // PRINT-NEXT: for (T i = Start; i < End; i += Step)
+    // DUMP-NEXT:  ForStmt
+    for (T i = Start; i < End; i += Step)
+      // PRINT-NEXT: body(i);
+      // DUMP:  CallExpr
+      body(i);
+}
+void unroll_template() {
+  unroll_templated<int,0,1024,1,4>();
+}
+
+#endif

clang/test/OpenMP/unroll_codegen_factor.cpp

+// Check code generation
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...) {}
+
+
+// IR-LABEL: @func(
+// IR-NEXT:  [[ENTRY:.*]]:
+// IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP0:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP0]], i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_COND]]:
+// IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[CMP:.+]] = icmp slt i32 %[[TMP1]], %[[TMP2]]
+// IR-NEXT:    br i1 %[[CMP]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_BODY]]:
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    call void (...) @body(i32 %[[TMP3]], i32 %[[TMP4]], i32 %[[TMP5]], i32 %[[TMP6]])
+// IR-NEXT:    br label %[[FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_INC]]:
+// IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[ADD:.+]] = add nsw i32 %[[TMP8]], %[[TMP7]]
+// IR-NEXT:    store i32 %[[ADD]], i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP2:[0-9]+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_END]]:
+// IR-NEXT:    ret void
+// IR-NEXT:  }
+extern "C" void func(int start, int end, int step) {
+  #pragma omp unroll partial(4)
+  for (int i = start; i < end; i+=step)
+    body(start, end, step, i);
+}
+
+#endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]], ![[LOOPPROP5:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.count", i32 4}
+// IR: ![[LOOPPROP5]] = !{!"llvm.loop.unroll.enable"}

clang/test/OpenMP/unroll_codegen_for_collapse_outer.cpp

+// Check code generation
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...) {}
+
+
+// IR-LABEL: @func(
+// IR-NEXT:  [[ENTRY:.*]]:
+// IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTOMP_IV:.+]] = alloca i64, align 8
+// IR-NEXT:    %[[TMP:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[TMP1:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_2:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_3:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[J:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_4:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_5:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_6:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_7:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_10:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_12:.+]] = alloca i64, align 8
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_J:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTOMP_LB:.+]] = alloca i64, align 8
+// IR-NEXT:    %[[DOTOMP_UB:.+]] = alloca i64, align 8
+// IR-NEXT:    %[[DOTOMP_STRIDE:.+]] = alloca i64, align 8
+// IR-NEXT:    %[[DOTOMP_IS_LAST:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I22:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_J23:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLL_INNER_IV_J:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[TMP0:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
+// IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP1_1:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP1_1]], i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP4]], i32* %[[J]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP5]], i32* %[[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP6]], i32* %[[DOTCAPTURE_EXPR_5]], align 4
+// IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP7]], i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_5]], align 4
+// IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP8]], %[[TMP9]]
+// IR-NEXT:    %[[SUB8:.+]] = sub i32 %[[SUB]], 1
+// IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB8]], %[[TMP10]]
+// IR-NEXT:    %[[TMP11:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP11]]
+// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[DIV]], 1
+// IR-NEXT:    store i32 %[[SUB9]], i32* %[[DOTCAPTURE_EXPR_7]], align 4
+// IR-NEXT:    %[[TMP12:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_7]], align 4
+// IR-NEXT:    %[[ADD11:.+]] = add i32 %[[TMP12]], 1
+// IR-NEXT:    store i32 %[[ADD11]], i32* %[[DOTCAPTURE_EXPR_10]], align 4
+// IR-NEXT:    %[[TMP13:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[TMP14:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[SUB13:.+]] = sub i32 %[[TMP13]], %[[TMP14]]
+// IR-NEXT:    %[[SUB14:.+]] = sub i32 %[[SUB13]], 1
+// IR-NEXT:    %[[TMP15:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[ADD15:.+]] = add i32 %[[SUB14]], %[[TMP15]]
+// IR-NEXT:    %[[TMP16:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[DIV16:.+]] = udiv i32 %[[ADD15]], %[[TMP16]]
+// IR-NEXT:    %[[CONV:.+]] = zext i32 %[[DIV16]] to i64
+// IR-NEXT:    %[[TMP17:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_10]], align 4
+// IR-NEXT:    %[[SUB17:.+]] = sub i32 %[[TMP17]], -1
+// IR-NEXT:    %[[DIV18:.+]] = udiv i32 %[[SUB17]], 2
+// IR-NEXT:    %[[CONV19:.+]] = zext i32 %[[DIV18]] to i64
+// IR-NEXT:    %[[MUL:.+]] = mul nsw i64 %[[CONV]], %[[CONV19]]
+// IR-NEXT:    %[[SUB20:.+]] = sub nsw i64 %[[MUL]], 1
+// IR-NEXT:    store i64 %[[SUB20]], i64* %[[DOTCAPTURE_EXPR_12]], align 8
+// IR-NEXT:    %[[TMP18:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    store i32 %[[TMP18]], i32* %[[I]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTUNROLLED_IV_J]], align 4
+// IR-NEXT:    %[[TMP19:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP20:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[CMP:.+]] = icmp slt i32 %[[TMP19]], %[[TMP20]]
+// IR-NEXT:    br i1 %[[CMP]], label %[[LAND_LHS_TRUE:.+]], label %[[OMP_PRECOND_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_LHS_TRUE]]:
+// IR-NEXT:    %[[TMP21:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_10]], align 4
+// IR-NEXT:    %[[CMP21:.+]] = icmp ult i32 0, %[[TMP21]]
+// IR-NEXT:    br i1 %[[CMP21]], label %[[OMP_PRECOND_THEN:.+]], label %[[OMP_PRECOND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_PRECOND_THEN]]:
+// IR-NEXT:    store i64 0, i64* %[[DOTOMP_LB]], align 8
+// IR-NEXT:    %[[TMP22:.+]] = load i64, i64* %[[DOTCAPTURE_EXPR_12]], align 8
+// IR-NEXT:    store i64 %[[TMP22]], i64* %[[DOTOMP_UB]], align 8
+// IR-NEXT:    store i64 1, i64* %[[DOTOMP_STRIDE]], align 8
+// IR-NEXT:    store i32 0, i32* %[[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @1, i32 %[[TMP0]], i32 34, i32* %[[DOTOMP_IS_LAST]], i64* %[[DOTOMP_LB]], i64* %[[DOTOMP_UB]], i64* %[[DOTOMP_STRIDE]], i64 1, i64 1)
+// IR-NEXT:    %[[TMP23:.+]] = load i64, i64* %[[DOTOMP_UB]], align 8
+// IR-NEXT:    %[[TMP24:.+]] = load i64, i64* %[[DOTCAPTURE_EXPR_12]], align 8
+// IR-NEXT:    %[[CMP24:.+]] = icmp sgt i64 %[[TMP23]], %[[TMP24]]
+// IR-NEXT:    br i1 %[[CMP24]], label %[[COND_TRUE:.+]], label %[[COND_FALSE:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[COND_TRUE]]:
+// IR-NEXT:    %[[TMP25:.+]] = load i64, i64* %[[DOTCAPTURE_EXPR_12]], align 8
+// IR-NEXT:    br label %[[COND_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[COND_FALSE]]:
+// IR-NEXT:    %[[TMP26:.+]] = load i64, i64* %[[DOTOMP_UB]], align 8
+// IR-NEXT:    br label %[[COND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[COND_END]]:
+// IR-NEXT:    %[[COND:.+]] = phi i64 [ %[[TMP25]], %[[COND_TRUE]] ], [ %[[TMP26]], %[[COND_FALSE]] ]
+// IR-NEXT:    store i64 %[[COND]], i64* %[[DOTOMP_UB]], align 8
+// IR-NEXT:    %[[TMP27:.+]] = load i64, i64* %[[DOTOMP_LB]], align 8
+// IR-NEXT:    store i64 %[[TMP27]], i64* %[[DOTOMP_IV]], align 8
+// IR-NEXT:    br label %[[OMP_INNER_FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_COND]]:
+// IR-NEXT:    %[[TMP28:.+]] = load i64, i64* %[[DOTOMP_IV]], align 8
+// IR-NEXT:    %[[TMP29:.+]] = load i64, i64* %[[DOTOMP_UB]], align 8
+// IR-NEXT:    %[[CMP25:.+]] = icmp sle i64 %[[TMP28]], %[[TMP29]]
+// IR-NEXT:    br i1 %[[CMP25]], label %[[OMP_INNER_FOR_BODY:.+]], label %[[OMP_INNER_FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_BODY]]:
+// IR-NEXT:    %[[TMP30:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[CONV26:.+]] = sext i32 %[[TMP30]] to i64
+// IR-NEXT:    %[[TMP31:.+]] = load i64, i64* %[[DOTOMP_IV]], align 8
+// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_10]], align 4
+// IR-NEXT:    %[[SUB27:.+]] = sub i32 %[[TMP32]], -1
+// IR-NEXT:    %[[DIV28:.+]] = udiv i32 %[[SUB27]], 2
+// IR-NEXT:    %[[MUL29:.+]] = mul i32 1, %[[DIV28]]
+// IR-NEXT:    %[[CONV30:.+]] = zext i32 %[[MUL29]] to i64
+// IR-NEXT:    %[[DIV31:.+]] = sdiv i64 %[[TMP31]], %[[CONV30]]
+// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[CONV32:.+]] = sext i32 %[[TMP33]] to i64
+// IR-NEXT:    %[[MUL33:.+]] = mul nsw i64 %[[DIV31]], %[[CONV32]]
+// IR-NEXT:    %[[ADD34:.+]] = add nsw i64 %[[CONV26]], %[[MUL33]]
+// IR-NEXT:    %[[CONV35:.+]] = trunc i64 %[[ADD34]] to i32
+// IR-NEXT:    store i32 %[[CONV35]], i32* %[[I22]], align 4
+// IR-NEXT:    %[[TMP34:.+]] = load i64, i64* %[[DOTOMP_IV]], align 8
+// IR-NEXT:    %[[TMP35:.+]] = load i64, i64* %[[DOTOMP_IV]], align 8
+// IR-NEXT:    %[[TMP36:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_10]], align 4
+// IR-NEXT:    %[[SUB36:.+]] = sub i32 %[[TMP36]], -1
+// IR-NEXT:    %[[DIV37:.+]] = udiv i32 %[[SUB36]], 2
+// IR-NEXT:    %[[MUL38:.+]] = mul i32 1, %[[DIV37]]
+// IR-NEXT:    %[[CONV39:.+]] = zext i32 %[[MUL38]] to i64
+// IR-NEXT:    %[[DIV40:.+]] = sdiv i64 %[[TMP35]], %[[CONV39]]
+// IR-NEXT:    %[[TMP37:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_10]], align 4
+// IR-NEXT:    %[[SUB41:.+]] = sub i32 %[[TMP37]], -1
+// IR-NEXT:    %[[DIV42:.+]] = udiv i32 %[[SUB41]], 2
+// IR-NEXT:    %[[MUL43:.+]] = mul i32 1, %[[DIV42]]
+// IR-NEXT:    %[[CONV44:.+]] = zext i32 %[[MUL43]] to i64
+// IR-NEXT:    %[[MUL45:.+]] = mul nsw i64 %[[DIV40]], %[[CONV44]]
+// IR-NEXT:    %[[SUB46:.+]] = sub nsw i64 %[[TMP34]], %[[MUL45]]
+// IR-NEXT:    %[[MUL47:.+]] = mul nsw i64 %[[SUB46]], 2
+// IR-NEXT:    %[[ADD48:.+]] = add nsw i64 0, %[[MUL47]]
+// IR-NEXT:    %[[CONV49:.+]] = trunc i64 %[[ADD48]] to i32
+// IR-NEXT:    store i32 %[[CONV49]], i32* %[[DOTUNROLLED_IV_J23]], align 4
+// IR-NEXT:    %[[TMP38:.+]] = load i32, i32* %[[DOTUNROLLED_IV_J23]], align 4
+// IR-NEXT:    store i32 %[[TMP38]], i32* %[[DOTUNROLL_INNER_IV_J]], align 4
+// IR-NEXT:    br label %[[FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_COND]]:
+// IR-NEXT:    %[[TMP39:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_J]], align 4
+// IR-NEXT:    %[[TMP40:.+]] = load i32, i32* %[[DOTUNROLLED_IV_J23]], align 4
+// IR-NEXT:    %[[ADD50:.+]] = add i32 %[[TMP40]], 2
+// IR-NEXT:    %[[CMP51:.+]] = icmp ule i32 %[[TMP39]], %[[ADD50]]
+// IR-NEXT:    br i1 %[[CMP51]], label %[[LAND_RHS:.+]], label %[[LAND_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_RHS]]:
+// IR-NEXT:    %[[TMP41:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_J]], align 4
+// IR-NEXT:    %[[TMP42:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_7]], align 4
+// IR-NEXT:    %[[ADD52:.+]] = add i32 %[[TMP42]], 1
+// IR-NEXT:    %[[CMP53:.+]] = icmp ule i32 %[[TMP41]], %[[ADD52]]
+// IR-NEXT:    br label %[[LAND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_END]]:
+// IR-NEXT:    %[[TMP43:.+]] = phi i1 [ false, %[[FOR_COND]] ], [ %[[CMP53]], %[[LAND_RHS]] ]
+// IR-NEXT:    br i1 %[[TMP43]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_BODY]]:
+// IR-NEXT:    %[[TMP44:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_4]], align 4
+// IR-NEXT:    %[[TMP45:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_J]], align 4
+// IR-NEXT:    %[[TMP46:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[MUL54:.+]] = mul i32 %[[TMP45]], %[[TMP46]]
+// IR-NEXT:    %[[ADD55:.+]] = add i32 %[[TMP44]], %[[MUL54]]
+// IR-NEXT:    store i32 %[[ADD55]], i32* %[[J]], align 4
+// IR-NEXT:    %[[TMP47:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    %[[TMP48:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[TMP49:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP50:.+]] = load i32, i32* %[[I22]], align 4
+// IR-NEXT:    %[[TMP51:.+]] = load i32, i32* %[[J]], align 4
+// IR-NEXT:    call void (...) @body(i32 %[[TMP47]], i32 %[[TMP48]], i32 %[[TMP49]], i32 %[[TMP50]], i32 %[[TMP51]])
+// IR-NEXT:    br label %[[FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_INC]]:
+// IR-NEXT:    %[[TMP52:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_J]], align 4
+// IR-NEXT:    %[[INC:.+]] = add i32 %[[TMP52]], 1
+// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTUNROLL_INNER_IV_J]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP2:[0-9]+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_END]]:
+// IR-NEXT:    br label %[[OMP_BODY_CONTINUE:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_BODY_CONTINUE]]:
+// IR-NEXT:    br label %[[OMP_INNER_FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_INC]]:
+// IR-NEXT:    %[[TMP53:.+]] = load i64, i64* %[[DOTOMP_IV]], align 8
+// IR-NEXT:    %[[ADD56:.+]] = add nsw i64 %[[TMP53]], 1
+// IR-NEXT:    store i64 %[[ADD56]], i64* %[[DOTOMP_IV]], align 8
+// IR-NEXT:    br label %[[OMP_INNER_FOR_COND]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_END]]:
+// IR-NEXT:    br label %[[OMP_LOOP_EXIT:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_LOOP_EXIT]]:
+// IR-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @1, i32 %[[TMP0]])
+// IR-NEXT:    br label %[[OMP_PRECOND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_PRECOND_END]]:
+// IR-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @3, i32 %[[TMP0]])
+// IR-NEXT:    ret void
+// IR-NEXT:  }
+extern "C" void func(int start, int end, int step) {
+  #pragma omp for collapse(2)
+  for (int i = start; i < end; i+=step) {
+    #pragma omp unroll partial
+    for (int j = start; j < end; j+=step)
+        body(start, end, step, i, j);
+  }
+}
+
+#endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.count", i32 2}

clang/test/OpenMP/unroll_codegen_for_partial.cpp

 // Check code generation
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 
 // Check same results after serialization round-trip
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 // expected-no-diagnostics
 
-// The loop trip count used by #pragma omp for depends on code generated
-// by #pragma omp file. Check that theses PreInits are emitted before
-// the code generated by #pragma omp for.
-
 #ifndef HEADER
 #define HEADER
 
 // placeholder for loop body code.
-
 extern "C" void body(...) {}
 
 
-
 // IR-LABEL: @func(
 // IR-NEXT:  [[ENTRY:.*]]:
 // IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IV:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_1:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_2:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_3:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_6:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_8:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_LB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_UB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_STRIDE:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IS_LAST:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV_I12:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTTILE_0_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_I12:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLL_INNER_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP0:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
 // IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
 // IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
 // IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[START_ADDR]], align 4
 // IR-NEXT:    store i32 %[[TMP1]], i32* %[[DOTCAPTURE_EXPR_]], align 4
 // IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
 // IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
 // IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
 // IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
 // IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
 // IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP4]], %[[TMP5]]
 // IR-NEXT:    %[[SUB4:.+]] = sub i32 %[[SUB]], 1
 // IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
 // IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP6]]
 // IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
 // IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP7]]
 // IR-NEXT:    %[[SUB5:.+]] = sub i32 %[[DIV]], 1
 // IR-NEXT:    store i32 %[[SUB5]], i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP8]], i32* %[[I]], align 4
-// IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[ADD7:.+]] = add i32 %[[TMP9]], 1
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[ADD7:.+]] = add i32 %[[TMP8]], 1
 // IR-NEXT:    store i32 %[[ADD7]], i32* %[[DOTCAPTURE_EXPR_6]], align 4
-// IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
-// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP10]], -3
-// IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 4
+// IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP9]], -1
+// IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 2
 // IR-NEXT:    %[[SUB11:.+]] = sub i32 %[[DIV10]], 1
 // IR-NEXT:    store i32 %[[SUB11]], i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP11:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
-// IR-NEXT:    %[[CMP:.+]] = icmp ult i32 0, %[[TMP11]]
+// IR-NEXT:    store i32 0, i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[CMP:.+]] = icmp ult i32 0, %[[TMP10]]
 // IR-NEXT:    br i1 %[[CMP]], label %[[OMP_PRECOND_THEN:.+]], label %[[OMP_PRECOND_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_PRECOND_THEN]]:
 // IR-NEXT:    store i32 0, i32* %[[DOTOMP_LB]], align 4
-// IR-NEXT:    %[[TMP12:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    store i32 %[[TMP12]], i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    %[[TMP11:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
+// IR-NEXT:    store i32 %[[TMP11]], i32* %[[DOTOMP_UB]], align 4
 // IR-NEXT:    store i32 1, i32* %[[DOTOMP_STRIDE]], align 4
 // IR-NEXT:    store i32 0, i32* %[[DOTOMP_IS_LAST]], align 4
 // IR-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @1, i32 %[[TMP0]], i32 34, i32* %[[DOTOMP_IS_LAST]], i32* %[[DOTOMP_LB]], i32* %[[DOTOMP_UB]], i32* %[[DOTOMP_STRIDE]], i32 1, i32 1)
-// IR-NEXT:    %[[TMP13:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
-// IR-NEXT:    %[[TMP14:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    %[[CMP13:.+]] = icmp ugt i32 %[[TMP13]], %[[TMP14]]
+// IR-NEXT:    %[[TMP12:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    %[[TMP13:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
+// IR-NEXT:    %[[CMP13:.+]] = icmp ugt i32 %[[TMP12]], %[[TMP13]]
 // IR-NEXT:    br i1 %[[CMP13]], label %[[COND_TRUE:.+]], label %[[COND_FALSE:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[COND_TRUE]]:
-// IR-NEXT:    %[[TMP15:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
+// IR-NEXT:    %[[TMP14:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
 // IR-NEXT:    br label %[[COND_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[COND_FALSE]]:
-// IR-NEXT:    %[[TMP16:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    %[[TMP15:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
 // IR-NEXT:    br label %[[COND_END]]
 // IR-EMPTY:
 // IR-NEXT:  [[COND_END]]:
-// IR-NEXT:    %[[COND:.+]] = phi i32 [ %[[TMP15]], %[[COND_TRUE]] ], [ %[[TMP16]], %[[COND_FALSE]] ]
+// IR-NEXT:    %[[COND:.+]] = phi i32 [ %[[TMP14]], %[[COND_TRUE]] ], [ %[[TMP15]], %[[COND_FALSE]] ]
 // IR-NEXT:    store i32 %[[COND]], i32* %[[DOTOMP_UB]], align 4
-// IR-NEXT:    %[[TMP17:.+]] = load i32, i32* %[[DOTOMP_LB]], align 4
-// IR-NEXT:    store i32 %[[TMP17]], i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[TMP16:.+]] = load i32, i32* %[[DOTOMP_LB]], align 4
+// IR-NEXT:    store i32 %[[TMP16]], i32* %[[DOTOMP_IV]], align 4
 // IR-NEXT:    br label %[[OMP_INNER_FOR_COND:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_COND]]:
-// IR-NEXT:    %[[TMP18:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
-// IR-NEXT:    %[[TMP19:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
-// IR-NEXT:    %[[ADD14:.+]] = add i32 %[[TMP19]], 1
-// IR-NEXT:    %[[CMP15:.+]] = icmp ult i32 %[[TMP18]], %[[ADD14]]
+// IR-NEXT:    %[[TMP17:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[TMP18:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    %[[ADD14:.+]] = add i32 %[[TMP18]], 1
+// IR-NEXT:    %[[CMP15:.+]] = icmp ult i32 %[[TMP17]], %[[ADD14]]
 // IR-NEXT:    br i1 %[[CMP15]], label %[[OMP_INNER_FOR_BODY:.+]], label %[[OMP_INNER_FOR_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_BODY]]:
-// IR-NEXT:    %[[TMP20:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
-// IR-NEXT:    %[[MUL:.+]] = mul i32 %[[TMP20]], 4
+// IR-NEXT:    %[[TMP19:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[MUL:.+]] = mul i32 %[[TMP19]], 2
 // IR-NEXT:    %[[ADD16:.+]] = add i32 0, %[[MUL]]
-// IR-NEXT:    store i32 %[[ADD16]], i32* %[[DOTFLOOR_0_IV_I12]], align 4
-// IR-NEXT:    %[[TMP21:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I12]], align 4
-// IR-NEXT:    store i32 %[[TMP21]], i32* %[[DOTTILE_0_IV_I]], align 4
+// IR-NEXT:    store i32 %[[ADD16]], i32* %[[DOTUNROLLED_IV_I12]], align 4
+// IR-NEXT:    %[[TMP20:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I12]], align 4
+// IR-NEXT:    store i32 %[[TMP20]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    br label %[[FOR_COND:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_COND]]:
-// IR-NEXT:    %[[TMP22:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP23:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[ADD17:.+]] = add i32 %[[TMP23]], 1
-// IR-NEXT:    %[[TMP24:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I12]], align 4
-// IR-NEXT:    %[[ADD18:.+]] = add nsw i32 %[[TMP24]], 4
-// IR-NEXT:    %[[CMP19:.+]] = icmp ult i32 %[[ADD17]], %[[ADD18]]
-// IR-NEXT:    br i1 %[[CMP19]], label %[[COND_TRUE20:.+]], label %[[COND_FALSE22:.+]]
-// IR-EMPTY:
-// IR-NEXT:  [[COND_TRUE20]]:
-// IR-NEXT:    %[[TMP25:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[ADD21:.+]] = add i32 %[[TMP25]], 1
-// IR-NEXT:    br label %[[COND_END24:.+]]
-// IR-EMPTY:
-// IR-NEXT:  [[COND_FALSE22]]:
-// IR-NEXT:    %[[TMP26:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I12]], align 4
-// IR-NEXT:    %[[ADD23:.+]] = add nsw i32 %[[TMP26]], 4
-// IR-NEXT:    br label %[[COND_END24]]
-// IR-EMPTY:
-// IR-NEXT:  [[COND_END24]]:
-// IR-NEXT:    %[[COND25:.+]] = phi i32 [ %[[ADD21]], %[[COND_TRUE20]] ], [ %[[ADD23]], %[[COND_FALSE22]] ]
-// IR-NEXT:    %[[CMP26:.+]] = icmp ult i32 %[[TMP22]], %[[COND25]]
-// IR-NEXT:    br i1 %[[CMP26]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
+// IR-NEXT:    %[[TMP21:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP22:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I12]], align 4
+// IR-NEXT:    %[[ADD17:.+]] = add i32 %[[TMP22]], 2
+// IR-NEXT:    %[[CMP18:.+]] = icmp ule i32 %[[TMP21]], %[[ADD17]]
+// IR-NEXT:    br i1 %[[CMP18]], label %[[LAND_RHS:.+]], label %[[LAND_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_RHS]]:
+// IR-NEXT:    %[[TMP23:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP24:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[ADD19:.+]] = add i32 %[[TMP24]], 1
+// IR-NEXT:    %[[CMP20:.+]] = icmp ule i32 %[[TMP23]], %[[ADD19]]
+// IR-NEXT:    br label %[[LAND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_END]]:
+// IR-NEXT:    %[[TMP25:.+]] = phi i1 [ false, %[[FOR_COND]] ], [ %[[CMP20]], %[[LAND_RHS]] ]
+// IR-NEXT:    br i1 %[[TMP25]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_BODY]]:
-// IR-NEXT:    %[[TMP27:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[TMP28:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP29:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[MUL27:.+]] = mul i32 %[[TMP28]], %[[TMP29]]
-// IR-NEXT:    %[[ADD28:.+]] = add i32 %[[TMP27]], %[[MUL27]]
-// IR-NEXT:    store i32 %[[ADD28]], i32* %[[I]], align 4
-// IR-NEXT:    %[[TMP30:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    %[[TMP31:.+]] = load i32, i32* %[[END_ADDR]], align 4
-// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
-// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[I]], align 4
-// IR-NEXT:    call void (...) @body(i32 %[[TMP30]], i32 %[[TMP31]], i32 %[[TMP32]], i32 %[[TMP33]])
+// IR-NEXT:    %[[TMP26:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP27:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP28:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[MUL21:.+]] = mul i32 %[[TMP27]], %[[TMP28]]
+// IR-NEXT:    %[[ADD22:.+]] = add i32 %[[TMP26]], %[[MUL21]]
+// IR-NEXT:    store i32 %[[ADD22]], i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP29:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    %[[TMP30:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[TMP31:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    call void (...) @body(i32 %[[TMP29]], i32 %[[TMP30]], i32 %[[TMP31]], i32 %[[TMP32]])
 // IR-NEXT:    br label %[[FOR_INC:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_INC]]:
-// IR-NEXT:    %[[TMP34:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    %[[INC:.+]] = add nsw i32 %[[TMP34]], 1
-// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTTILE_0_IV_I]], align 4
+// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[INC:.+]] = add i32 %[[TMP33]], 1
+// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP2:[0-9]+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_END]]:
 // IR-NEXT:    br label %[[OMP_BODY_CONTINUE:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_BODY_CONTINUE]]:
 // IR-NEXT:    br label %[[OMP_INNER_FOR_INC:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_INC]]:
-// IR-NEXT:    %[[TMP35:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
-// IR-NEXT:    %[[ADD29:.+]] = add i32 %[[TMP35]], 1
-// IR-NEXT:    store i32 %[[ADD29]], i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[TMP34:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[ADD23:.+]] = add i32 %[[TMP34]], 1
+// IR-NEXT:    store i32 %[[ADD23]], i32* %[[DOTOMP_IV]], align 4
 // IR-NEXT:    br label %[[OMP_INNER_FOR_COND]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_END]]:
 // IR-NEXT:    br label %[[OMP_LOOP_EXIT:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_LOOP_EXIT]]:
 // IR-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @1, i32 %[[TMP0]])
 // IR-NEXT:    br label %[[OMP_PRECOND_END]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_PRECOND_END]]:
 // IR-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @3, i32 %[[TMP0]])
 // IR-NEXT:    ret void
 // IR-NEXT:  }
 extern "C" void func(int start, int end, int step) {
+  int i;
   #pragma omp for
-#pragma omp tile sizes(4)
-  for (int i = start; i < end; i += step)
+  #pragma omp unroll partial
+  for (i = start; i < end; i+=step)
     body(start, end, step, i);
 }
 
 #endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.count", i32 2}

clang/test/OpenMP/unroll_codegen_full.cpp

+// Check code generation
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...) {}
+
+
+// IR-LABEL: @func(
+// IR-NEXT:  [[ENTRY:.*]]:
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    store i32 7, i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_COND]]:
+// IR-NEXT:    %[[TMP0:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[CMP:.+]] = icmp slt i32 %[[TMP0]], 17
+// IR-NEXT:    br i1 %[[CMP]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_BODY]]:
+// IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    call void (...) @body(i32 %[[TMP1]])
+// IR-NEXT:    br label %[[FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_INC]]:
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[ADD:.+]] = add nsw i32 %[[TMP2]], 3
+// IR-NEXT:    store i32 %[[ADD]], i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP2:[0-9]+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_END]]:
+// IR-NEXT:    ret void
+// IR-NEXT:  }
+extern "C" void func() {
+  #pragma omp unroll full
+  for (int i = 7; i < 17; i += 3)
+    body(i);
+}
+
+#endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.full"}

clang/test/OpenMP/unroll_codegen_heuristic.cpp

+// Check code generation
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...) {}
+
+
+// IR-LABEL: @func(
+// IR-NEXT:  [[ENTRY:.*]]:
+// IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP0:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP0]], i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_COND]]:
+// IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[CMP:.+]] = icmp slt i32 %[[TMP1]], %[[TMP2]]
+// IR-NEXT:    br i1 %[[CMP]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_BODY]]:
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    call void (...) @body(i32 %[[TMP3]], i32 %[[TMP4]], i32 %[[TMP5]], i32 %[[TMP6]])
+// IR-NEXT:    br label %[[FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_INC]]:
+// IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[ADD:.+]] = add nsw i32 %[[TMP8]], %[[TMP7]]
+// IR-NEXT:    store i32 %[[ADD]], i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP2:[0-9]+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_END]]:
+// IR-NEXT:    ret void
+// IR-NEXT:  }
+extern "C" void func(int start, int end, int step) {
+  #pragma omp unroll
+  for (int i = start; i < end; i+=step)
+    body(start, end, step, i);
+}
+
+#endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.enable"}

clang/test/OpenMP/unroll_codegen_parallel_for_factor.cpp

+// Check code generation
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...) {}
+
+
+// IR-LABEL: @func(
+// IR-NEXT:  [[ENTRY:.*]]:
+// IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @2, i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* %[[END_ADDR]], i32* %[[STEP_ADDR]], i32* %[[START_ADDR]])
+// IR-NEXT:    ret void
+// IR-NEXT:  }
+extern "C" void func(int start, int end, int step) {
+  #pragma omp parallel for
+  #pragma omp unroll partial(7)
+  for (int i = start; i < end; i+=step)
+    body(start, end, step, i);
+}
+
+
+// IR-LABEL: @.omp_outlined.(
+// IR-NEXT:  [[ENTRY:.*]]:
+// IR-NEXT:    %[[DOTGLOBAL_TID__ADDR:.+]] = alloca i32*, align 8
+// IR-NEXT:    %[[DOTBOUND_TID__ADDR:.+]] = alloca i32*, align 8
+// IR-NEXT:    %[[END_ADDR:.+]] = alloca i32*, align 8
+// IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32*, align 8
+// IR-NEXT:    %[[START_ADDR:.+]] = alloca i32*, align 8
+// IR-NEXT:    %[[DOTOMP_IV:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[TMP:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_1:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_2:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_3:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_6:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTCAPTURE_EXPR_8:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTOMP_LB:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTOMP_UB:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTOMP_STRIDE:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTOMP_IS_LAST:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_I12:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLL_INNER_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    store i32* %[[DOTGLOBAL_TID_:.+]], i32** %[[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    store i32* %[[DOTBOUND_TID_:.+]], i32** %[[DOTBOUND_TID__ADDR]], align 8
+// IR-NEXT:    store i32* %[[END:.+]], i32** %[[END_ADDR]], align 8
+// IR-NEXT:    store i32* %[[STEP:.+]], i32** %[[STEP_ADDR]], align 8
+// IR-NEXT:    store i32* %[[START:.+]], i32** %[[START_ADDR]], align 8
+// IR-NEXT:    %[[TMP0:.+]] = load i32*, i32** %[[END_ADDR]], align 8
+// IR-NEXT:    %[[TMP1:.+]] = load i32*, i32** %[[STEP_ADDR]], align 8
+// IR-NEXT:    %[[TMP2:.+]] = load i32*, i32** %[[START_ADDR]], align 8
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[TMP2]], align 4
+// IR-NEXT:    store i32 %[[TMP3]], i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[TMP2]], align 4
+// IR-NEXT:    store i32 %[[TMP4]], i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[TMP0]], align 4
+// IR-NEXT:    store i32 %[[TMP5]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[TMP1]], align 4
+// IR-NEXT:    store i32 %[[TMP6]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP7]], %[[TMP8]]
+// IR-NEXT:    %[[SUB4:.+]] = sub i32 %[[SUB]], 1
+// IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP9]]
+// IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP10]]
+// IR-NEXT:    %[[SUB5:.+]] = sub i32 %[[DIV]], 1
+// IR-NEXT:    store i32 %[[SUB5]], i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[TMP11:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[ADD7:.+]] = add i32 %[[TMP11]], 1
+// IR-NEXT:    store i32 %[[ADD7]], i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[TMP12:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP12]], -6
+// IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 7
+// IR-NEXT:    %[[SUB11:.+]] = sub i32 %[[DIV10]], 1
+// IR-NEXT:    store i32 %[[SUB11]], i32* %[[DOTCAPTURE_EXPR_8]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[TMP13:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
+// IR-NEXT:    %[[CMP:.+]] = icmp ult i32 0, %[[TMP13]]
+// IR-NEXT:    br i1 %[[CMP]], label %[[OMP_PRECOND_THEN:.+]], label %[[OMP_PRECOND_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_PRECOND_THEN]]:
+// IR-NEXT:    store i32 0, i32* %[[DOTOMP_LB]], align 4
+// IR-NEXT:    %[[TMP14:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
+// IR-NEXT:    store i32 %[[TMP14]], i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    store i32 1, i32* %[[DOTOMP_STRIDE]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTOMP_IS_LAST]], align 4
+// IR-NEXT:    %[[TMP15:.+]] = load i32*, i32** %[[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    %[[TMP16:.+]] = load i32, i32* %[[TMP15]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @1, i32 %[[TMP16]], i32 34, i32* %[[DOTOMP_IS_LAST]], i32* %[[DOTOMP_LB]], i32* %[[DOTOMP_UB]], i32* %[[DOTOMP_STRIDE]], i32 1, i32 1)
+// IR-NEXT:    %[[TMP17:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    %[[TMP18:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
+// IR-NEXT:    %[[CMP13:.+]] = icmp ugt i32 %[[TMP17]], %[[TMP18]]
+// IR-NEXT:    br i1 %[[CMP13]], label %[[COND_TRUE:.+]], label %[[COND_FALSE:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[COND_TRUE]]:
+// IR-NEXT:    %[[TMP19:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
+// IR-NEXT:    br label %[[COND_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[COND_FALSE]]:
+// IR-NEXT:    %[[TMP20:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    br label %[[COND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[COND_END]]:
+// IR-NEXT:    %[[COND:.+]] = phi i32 [ %[[TMP19]], %[[COND_TRUE]] ], [ %[[TMP20]], %[[COND_FALSE]] ]
+// IR-NEXT:    store i32 %[[COND]], i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    %[[TMP21:.+]] = load i32, i32* %[[DOTOMP_LB]], align 4
+// IR-NEXT:    store i32 %[[TMP21]], i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    br label %[[OMP_INNER_FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_COND]]:
+// IR-NEXT:    %[[TMP22:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[TMP23:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
+// IR-NEXT:    %[[ADD14:.+]] = add i32 %[[TMP23]], 1
+// IR-NEXT:    %[[CMP15:.+]] = icmp ult i32 %[[TMP22]], %[[ADD14]]
+// IR-NEXT:    br i1 %[[CMP15]], label %[[OMP_INNER_FOR_BODY:.+]], label %[[OMP_INNER_FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_BODY]]:
+// IR-NEXT:    %[[TMP24:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[MUL:.+]] = mul i32 %[[TMP24]], 7
+// IR-NEXT:    %[[ADD16:.+]] = add i32 0, %[[MUL]]
+// IR-NEXT:    store i32 %[[ADD16]], i32* %[[DOTUNROLLED_IV_I12]], align 4
+// IR-NEXT:    %[[TMP25:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I12]], align 4
+// IR-NEXT:    store i32 %[[TMP25]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    br label %[[FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_COND]]:
+// IR-NEXT:    %[[TMP26:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP27:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I12]], align 4
+// IR-NEXT:    %[[ADD17:.+]] = add i32 %[[TMP27]], 7
+// IR-NEXT:    %[[CMP18:.+]] = icmp ule i32 %[[TMP26]], %[[ADD17]]
+// IR-NEXT:    br i1 %[[CMP18]], label %[[LAND_RHS:.+]], label %[[LAND_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_RHS]]:
+// IR-NEXT:    %[[TMP28:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP29:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
+// IR-NEXT:    %[[ADD19:.+]] = add i32 %[[TMP29]], 1
+// IR-NEXT:    %[[CMP20:.+]] = icmp ule i32 %[[TMP28]], %[[ADD19]]
+// IR-NEXT:    br label %[[LAND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_END]]:
+// IR-NEXT:    %[[TMP30:.+]] = phi i1 [ false, %[[FOR_COND]] ], [ %[[CMP20]], %[[LAND_RHS]] ]
+// IR-NEXT:    br i1 %[[TMP30]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_BODY]]:
+// IR-NEXT:    %[[TMP31:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[MUL21:.+]] = mul i32 %[[TMP32]], %[[TMP33]]
+// IR-NEXT:    %[[ADD22:.+]] = add i32 %[[TMP31]], %[[MUL21]]
+// IR-NEXT:    store i32 %[[ADD22]], i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP34:.+]] = load i32, i32* %[[TMP2]], align 4
+// IR-NEXT:    %[[TMP35:.+]] = load i32, i32* %[[TMP0]], align 4
+// IR-NEXT:    %[[TMP36:.+]] = load i32, i32* %[[TMP1]], align 4
+// IR-NEXT:    %[[TMP37:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    call void (...) @body(i32 %[[TMP34]], i32 %[[TMP35]], i32 %[[TMP36]], i32 %[[TMP37]])
+// IR-NEXT:    br label %[[FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_INC]]:
+// IR-NEXT:    %[[TMP38:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[INC:.+]] = add i32 %[[TMP38]], 1
+// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP2:[0-9]+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_END]]:
+// IR-NEXT:    br label %[[OMP_BODY_CONTINUE:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_BODY_CONTINUE]]:
+// IR-NEXT:    br label %[[OMP_INNER_FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_INC]]:
+// IR-NEXT:    %[[TMP39:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[ADD23:.+]] = add i32 %[[TMP39]], 1
+// IR-NEXT:    store i32 %[[ADD23]], i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    br label %[[OMP_INNER_FOR_COND]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_INNER_FOR_END]]:
+// IR-NEXT:    br label %[[OMP_LOOP_EXIT:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_LOOP_EXIT]]:
+// IR-NEXT:    %[[TMP40:.+]] = load i32*, i32** %[[DOTGLOBAL_TID__ADDR]], align 8
+// IR-NEXT:    %[[TMP41:.+]] = load i32, i32* %[[TMP40]], align 4
+// IR-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @1, i32 %[[TMP41]])
+// IR-NEXT:    br label %[[OMP_PRECOND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[OMP_PRECOND_END]]:
+// IR-NEXT:    ret void
+// IR-NEXT:  }
+
+#endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.count", i32 7}

clang/test/OpenMP/unroll_codegen_partial.cpp

+// Check code generation
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+
+// Check same results after serialization round-trip
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
+// RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+// placeholder for loop body code.
+extern "C" void body(...) {}
+
+
+// IR-LABEL: @func(
+// IR-NEXT:  [[ENTRY:.*]]:
+// IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP0:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP0]], i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_COND]]:
+// IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[CMP:.+]] = icmp slt i32 %[[TMP1]], %[[TMP2]]
+// IR-NEXT:    br i1 %[[CMP]], label %[[FOR_BODY:.+]], label %[[FOR_END:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_BODY]]:
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    call void (...) @body(i32 %[[TMP3]], i32 %[[TMP4]], i32 %[[TMP5]], i32 %[[TMP6]])
+// IR-NEXT:    br label %[[FOR_INC:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_INC]]:
+// IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[I]], align 4
+// IR-NEXT:    %[[ADD:.+]] = add nsw i32 %[[TMP8]], %[[TMP7]]
+// IR-NEXT:    store i32 %[[ADD]], i32* %[[I]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP2:[0-9]+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_END]]:
+// IR-NEXT:    ret void
+// IR-NEXT:  }
+extern "C" void func(int start, int end, int step) {
+  #pragma omp unroll partial
+  for (int i = start; i < end; i+=step)
+    body(start, end, step, i);
+}
+
+#endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.enable"}

clang/test/OpenMP/unroll_codegen_tile_for.cpp

 // Check code generation
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 
 // Check same results after serialization round-trip
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 // expected-no-diagnostics
 
-// Account for multiple transformations of a loop before consumed by
-// #pragma omp for.
-
 #ifndef HEADER
 #define HEADER
 
 // placeholder for loop body code.
 extern "C" void body(...) {}
 
 
 // IR-LABEL: @func(
 // IR-NEXT:  [[ENTRY:.*]]:
 // IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IV:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_1:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_2:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_3:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_6:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_8:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_12:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_14:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV__FLOOR_0_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTFLOOR_0_IV__UNROLLED_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_LB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_UB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_STRIDE:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IS_LAST:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTTILE_0_IV__FLOOR_0_IV_I:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTTILE_0_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTFLOOR_0_IV__UNROLLED_IV_I18:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTTILE_0_IV__UNROLLED_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLL_INNER_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP0:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
 // IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
 // IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
 // IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP1]], i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP4]], %[[TMP5]]
+// IR-NEXT:    store i32 %[[TMP1]], i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP4]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP5]], %[[TMP6]]
 // IR-NEXT:    %[[SUB4:.+]] = sub i32 %[[SUB]], 1
-// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP6]]
 // IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP7]]
+// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP7]]
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP8]]
 // IR-NEXT:    %[[SUB5:.+]] = sub i32 %[[DIV]], 1
 // IR-NEXT:    store i32 %[[SUB5]], i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP8]], i32* %[[I]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTUNROLLED_IV_I]], align 4
 // IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
 // IR-NEXT:    %[[ADD7:.+]] = add i32 %[[TMP9]], 1
 // IR-NEXT:    store i32 %[[ADD7]], i32* %[[DOTCAPTURE_EXPR_6]], align 4
 // IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
-// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP10]], -3
-// IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 4
+// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP10]], -1
+// IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 2
 // IR-NEXT:    %[[SUB11:.+]] = sub i32 %[[DIV10]], 1
 // IR-NEXT:    store i32 %[[SUB11]], i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV_I]], align 4
 // IR-NEXT:    %[[TMP11:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
 // IR-NEXT:    %[[ADD13:.+]] = add i32 %[[TMP11]], 1
 // IR-NEXT:    store i32 %[[ADD13]], i32* %[[DOTCAPTURE_EXPR_12]], align 4
 // IR-NEXT:    %[[TMP12:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_12]], align 4
-// IR-NEXT:    %[[SUB15:.+]] = sub i32 %[[TMP12]], -2
-// IR-NEXT:    %[[DIV16:.+]] = udiv i32 %[[SUB15]], 3
+// IR-NEXT:    %[[SUB15:.+]] = sub i32 %[[TMP12]], -3
+// IR-NEXT:    %[[DIV16:.+]] = udiv i32 %[[SUB15]], 4
 // IR-NEXT:    %[[SUB17:.+]] = sub i32 %[[DIV16]], 1
 // IR-NEXT:    store i32 %[[SUB17]], i32* %[[DOTCAPTURE_EXPR_14]], align 4
-// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV__UNROLLED_IV_I]], align 4
 // IR-NEXT:    %[[TMP13:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_12]], align 4
 // IR-NEXT:    %[[CMP:.+]] = icmp ult i32 0, %[[TMP13]]
 // IR-NEXT:    br i1 %[[CMP]], label %[[OMP_PRECOND_THEN:.+]], label %[[OMP_PRECOND_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_PRECOND_THEN]]:
 // IR-NEXT:    store i32 0, i32* %[[DOTOMP_LB]], align 4
 // IR-NEXT:    %[[TMP14:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_14]], align 4
 // IR-NEXT:    store i32 %[[TMP14]], i32* %[[DOTOMP_UB]], align 4

 // IR-NEXT:    %[[TMP20:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
 // IR-NEXT:    %[[TMP21:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
 // IR-NEXT:    %[[ADD20:.+]] = add i32 %[[TMP21]], 1
 // IR-NEXT:    %[[CMP21:.+]] = icmp ult i32 %[[TMP20]], %[[ADD20]]
 // IR-NEXT:    br i1 %[[CMP21]], label %[[OMP_INNER_FOR_BODY:.+]], label %[[OMP_INNER_FOR_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_BODY]]:
 // IR-NEXT:    %[[TMP22:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
-// IR-NEXT:    %[[MUL:.+]] = mul i32 %[[TMP22]], 3
+// IR-NEXT:    %[[MUL:.+]] = mul i32 %[[TMP22]], 4
 // IR-NEXT:    %[[ADD22:.+]] = add i32 0, %[[MUL]]
-// IR-NEXT:    store i32 %[[ADD22]], i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    %[[TMP23:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    store i32 %[[TMP23]], i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
+// IR-NEXT:    store i32 %[[ADD22]], i32* %[[DOTFLOOR_0_IV__UNROLLED_IV_I18]], align 4
+// IR-NEXT:    %[[TMP23:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__UNROLLED_IV_I18]], align 4
+// IR-NEXT:    store i32 %[[TMP23]], i32* %[[DOTTILE_0_IV__UNROLLED_IV_I]], align 4
 // IR-NEXT:    br label %[[FOR_COND:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_COND]]:
-// IR-NEXT:    %[[TMP24:.+]] = load i32, i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
+// IR-NEXT:    %[[TMP24:.+]] = load i32, i32* %[[DOTTILE_0_IV__UNROLLED_IV_I]], align 4
 // IR-NEXT:    %[[TMP25:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
 // IR-NEXT:    %[[ADD23:.+]] = add i32 %[[TMP25]], 1
-// IR-NEXT:    %[[TMP26:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    %[[ADD24:.+]] = add i32 %[[TMP26]], 3
+// IR-NEXT:    %[[TMP26:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__UNROLLED_IV_I18]], align 4
+// IR-NEXT:    %[[ADD24:.+]] = add i32 %[[TMP26]], 4
 // IR-NEXT:    %[[CMP25:.+]] = icmp ult i32 %[[ADD23]], %[[ADD24]]
 // IR-NEXT:    br i1 %[[CMP25]], label %[[COND_TRUE26:.+]], label %[[COND_FALSE28:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[COND_TRUE26]]:
 // IR-NEXT:    %[[TMP27:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
 // IR-NEXT:    %[[ADD27:.+]] = add i32 %[[TMP27]], 1
 // IR-NEXT:    br label %[[COND_END30:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[COND_FALSE28]]:
-// IR-NEXT:    %[[TMP28:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    %[[ADD29:.+]] = add i32 %[[TMP28]], 3
+// IR-NEXT:    %[[TMP28:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__UNROLLED_IV_I18]], align 4
+// IR-NEXT:    %[[ADD29:.+]] = add i32 %[[TMP28]], 4
 // IR-NEXT:    br label %[[COND_END30]]
 // IR-EMPTY:
 // IR-NEXT:  [[COND_END30]]:
 // IR-NEXT:    %[[COND31:.+]] = phi i32 [ %[[ADD27]], %[[COND_TRUE26]] ], [ %[[ADD29]], %[[COND_FALSE28]] ]
 // IR-NEXT:    %[[CMP32:.+]] = icmp ult i32 %[[TMP24]], %[[COND31]]
-// IR-NEXT:    br i1 %[[CMP32]], label %[[FOR_BODY:.+]], label %[[FOR_END51:.+]]
+// IR-NEXT:    br i1 %[[CMP32]], label %[[FOR_BODY:.+]], label %[[FOR_END45:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_BODY]]:
-// IR-NEXT:    %[[TMP29:.+]] = load i32, i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[MUL33:.+]] = mul i32 %[[TMP29]], 4
+// IR-NEXT:    %[[TMP29:.+]] = load i32, i32* %[[DOTTILE_0_IV__UNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[MUL33:.+]] = mul i32 %[[TMP29]], 2
 // IR-NEXT:    %[[ADD34:.+]] = add i32 0, %[[MUL33]]
-// IR-NEXT:    store i32 %[[ADD34]], i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP30:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    store i32 %[[TMP30]], i32* %[[DOTTILE_0_IV_I]], align 4
+// IR-NEXT:    store i32 %[[ADD34]], i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[TMP30:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    store i32 %[[TMP30]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    br label %[[FOR_COND35:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_COND35]]:
-// IR-NEXT:    %[[TMP31:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[ADD36:.+]] = add i32 %[[TMP32]], 1
-// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[ADD37:.+]] = add nsw i32 %[[TMP33]], 4
-// IR-NEXT:    %[[CMP38:.+]] = icmp ult i32 %[[ADD36]], %[[ADD37]]
-// IR-NEXT:    br i1 %[[CMP38]], label %[[COND_TRUE39:.+]], label %[[COND_FALSE41:.+]]
+// IR-NEXT:    %[[TMP31:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[ADD36:.+]] = add i32 %[[TMP32]], 2
+// IR-NEXT:    %[[CMP37:.+]] = icmp ule i32 %[[TMP31]], %[[ADD36]]
+// IR-NEXT:    br i1 %[[CMP37]], label %[[LAND_RHS:.+]], label %[[LAND_END:.+]]
 // IR-EMPTY:
-// IR-NEXT:  [[COND_TRUE39]]:
+// IR-NEXT:  [[LAND_RHS]]:
+// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    %[[TMP34:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[ADD40:.+]] = add i32 %[[TMP34]], 1
-// IR-NEXT:    br label %[[COND_END43:.+]]
+// IR-NEXT:    %[[ADD38:.+]] = add i32 %[[TMP34]], 1
+// IR-NEXT:    %[[CMP39:.+]] = icmp ule i32 %[[TMP33]], %[[ADD38]]
+// IR-NEXT:    br label %[[LAND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_END]]:
+// IR-NEXT:    %[[TMP35:.+]] = phi i1 [ false, %[[FOR_COND35]] ], [ %[[CMP39]], %[[LAND_RHS]] ]
+// IR-NEXT:    br i1 %[[TMP35]], label %[[FOR_BODY40:.+]], label %[[FOR_END:.+]]
 // IR-EMPTY:
-// IR-NEXT:  [[COND_FALSE41]]:
-// IR-NEXT:    %[[TMP35:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[ADD42:.+]] = add nsw i32 %[[TMP35]], 4
-// IR-NEXT:    br label %[[COND_END43]]
-// IR-EMPTY:
-// IR-NEXT:  [[COND_END43]]:
-// IR-NEXT:    %[[COND44:.+]] = phi i32 [ %[[ADD40]], %[[COND_TRUE39]] ], [ %[[ADD42]], %[[COND_FALSE41]] ]
-// IR-NEXT:    %[[CMP45:.+]] = icmp ult i32 %[[TMP31]], %[[COND44]]
-// IR-NEXT:    br i1 %[[CMP45]], label %[[FOR_BODY46:.+]], label %[[FOR_END:.+]]
-// IR-EMPTY:
-// IR-NEXT:  [[FOR_BODY46]]:
+// IR-NEXT:  [[FOR_BODY40]]:
 // IR-NEXT:    %[[TMP36:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[TMP37:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
+// IR-NEXT:    %[[TMP37:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    %[[TMP38:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[MUL47:.+]] = mul i32 %[[TMP37]], %[[TMP38]]
-// IR-NEXT:    %[[ADD48:.+]] = add i32 %[[TMP36]], %[[MUL47]]
-// IR-NEXT:    store i32 %[[ADD48]], i32* %[[I]], align 4
+// IR-NEXT:    %[[MUL41:.+]] = mul i32 %[[TMP37]], %[[TMP38]]
+// IR-NEXT:    %[[ADD42:.+]] = add i32 %[[TMP36]], %[[MUL41]]
+// IR-NEXT:    store i32 %[[ADD42]], i32* %[[I]], align 4
 // IR-NEXT:    %[[TMP39:.+]] = load i32, i32* %[[START_ADDR]], align 4
 // IR-NEXT:    %[[TMP40:.+]] = load i32, i32* %[[END_ADDR]], align 4
 // IR-NEXT:    %[[TMP41:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    %[[TMP42:.+]] = load i32, i32* %[[I]], align 4
 // IR-NEXT:    call void (...) @body(i32 %[[TMP39]], i32 %[[TMP40]], i32 %[[TMP41]], i32 %[[TMP42]])
 // IR-NEXT:    br label %[[FOR_INC:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_INC]]:
-// IR-NEXT:    %[[TMP43:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    %[[INC:.+]] = add nsw i32 %[[TMP43]], 1
-// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTTILE_0_IV_I]], align 4
+// IR-NEXT:    %[[TMP43:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[INC:.+]] = add i32 %[[TMP43]], 1
+// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    br label %[[FOR_COND35]], !llvm.loop ![[LOOP2:[0-9]+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_END]]:
-// IR-NEXT:    br label %[[FOR_INC49:.+]]
+// IR-NEXT:    br label %[[FOR_INC43:.+]]
 // IR-EMPTY:
-// IR-NEXT:  [[FOR_INC49]]:
-// IR-NEXT:    %[[TMP44:.+]] = load i32, i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[INC50:.+]] = add i32 %[[TMP44]], 1
-// IR-NEXT:    store i32 %[[INC50]], i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
-// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP4:[0-9]+]]
+// IR-NEXT:  [[FOR_INC43]]:
+// IR-NEXT:    %[[TMP44:.+]] = load i32, i32* %[[DOTTILE_0_IV__UNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[INC44:.+]] = add i32 %[[TMP44]], 1
+// IR-NEXT:    store i32 %[[INC44]], i32* %[[DOTTILE_0_IV__UNROLLED_IV_I]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP5:[0-9]+]]
 // IR-EMPTY:
-// IR-NEXT:  [[FOR_END51]]:
+// IR-NEXT:  [[FOR_END45]]:
 // IR-NEXT:    br label %[[OMP_BODY_CONTINUE:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_BODY_CONTINUE]]:
 // IR-NEXT:    br label %[[OMP_INNER_FOR_INC:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_INC]]:
 // IR-NEXT:    %[[TMP45:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
-// IR-NEXT:    %[[ADD52:.+]] = add i32 %[[TMP45]], 1
-// IR-NEXT:    store i32 %[[ADD52]], i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[ADD46:.+]] = add i32 %[[TMP45]], 1
+// IR-NEXT:    store i32 %[[ADD46]], i32* %[[DOTOMP_IV]], align 4
 // IR-NEXT:    br label %[[OMP_INNER_FOR_COND]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_END]]:
 // IR-NEXT:    br label %[[OMP_LOOP_EXIT:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_LOOP_EXIT]]:
 // IR-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @1, i32 %[[TMP0]])
 // IR-NEXT:    br label %[[OMP_PRECOND_END]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_PRECOND_END]]:
 // IR-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @3, i32 %[[TMP0]])
 // IR-NEXT:    ret void
 // IR-NEXT:  }
 extern "C" void func(int start, int end, int step) {
   #pragma omp for
-#pragma omp tile sizes(3)
   #pragma omp tile sizes(4)
-  for (int i = start; i < end; i += step)
+  #pragma omp unroll partial
+  for (int i = start; i < end; i+=step)
     body(start, end, step, i);
 }
 
 #endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.count", i32 2}
+// IR: ![[LOOP5]] = distinct !{![[LOOP5]], ![[LOOPPROP3]]}

clang/test/OpenMP/unroll_codegen_unroll_for.cpp

 // Check code generation
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 
 // Check same results after serialization round-trip
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -emit-pch -o %t %s
 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -fopenmp-version=51 -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
 // expected-no-diagnostics
 
-// Account for multiple transformations of a loop before consumed by
-// #pragma omp for.
-
 #ifndef HEADER
 #define HEADER
 
 // placeholder for loop body code.
 extern "C" void body(...) {}
 
 
 // IR-LABEL: @func(
 // IR-NEXT:  [[ENTRY:.*]]:
 // IR-NEXT:    %[[START_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[END_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[STEP_ADDR:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IV:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_1:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_2:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_3:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_6:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_8:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_12:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTCAPTURE_EXPR_14:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV__FLOOR_0_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV__UNROLLED_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_LB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_UB:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_STRIDE:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[DOTOMP_IS_LAST:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTTILE_0_IV__FLOOR_0_IV_I:.+]] = alloca i32, align 4
-// IR-NEXT:    %[[DOTTILE_0_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLLED_IV__UNROLLED_IV_I18:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLL_INNER_IV__UNROLLED_IV_I:.+]] = alloca i32, align 4
+// IR-NEXT:    %[[DOTUNROLL_INNER_IV_I:.+]] = alloca i32, align 4
 // IR-NEXT:    %[[TMP0:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @2)
 // IR-NEXT:    store i32 %[[START:.+]], i32* %[[START_ADDR]], align 4
 // IR-NEXT:    store i32 %[[END:.+]], i32* %[[END_ADDR]], align 4
 // IR-NEXT:    store i32 %[[STEP:.+]], i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    %[[TMP1:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP1]], i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[END_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
-// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP4]], %[[TMP5]]
+// IR-NEXT:    store i32 %[[TMP1]], i32* %[[I]], align 4
+// IR-NEXT:    %[[TMP2:.+]] = load i32, i32* %[[START_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP2]], i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[TMP3:.+]] = load i32, i32* %[[END_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP3]], i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP4:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
+// IR-NEXT:    store i32 %[[TMP4]], i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[TMP5:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_1]], align 4
+// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
+// IR-NEXT:    %[[SUB:.+]] = sub i32 %[[TMP5]], %[[TMP6]]
 // IR-NEXT:    %[[SUB4:.+]] = sub i32 %[[SUB]], 1
-// IR-NEXT:    %[[TMP6:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP6]]
 // IR-NEXT:    %[[TMP7:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP7]]
+// IR-NEXT:    %[[ADD:.+]] = add i32 %[[SUB4]], %[[TMP7]]
+// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
+// IR-NEXT:    %[[DIV:.+]] = udiv i32 %[[ADD]], %[[TMP8]]
 // IR-NEXT:    %[[SUB5:.+]] = sub i32 %[[DIV]], 1
 // IR-NEXT:    store i32 %[[SUB5]], i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[TMP8:.+]] = load i32, i32* %[[START_ADDR]], align 4
-// IR-NEXT:    store i32 %[[TMP8]], i32* %[[I]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTUNROLLED_IV_I]], align 4
 // IR-NEXT:    %[[TMP9:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
 // IR-NEXT:    %[[ADD7:.+]] = add i32 %[[TMP9]], 1
 // IR-NEXT:    store i32 %[[ADD7]], i32* %[[DOTCAPTURE_EXPR_6]], align 4
 // IR-NEXT:    %[[TMP10:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_6]], align 4
-// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP10]], -3
-// IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 4
+// IR-NEXT:    %[[SUB9:.+]] = sub i32 %[[TMP10]], -1
+// IR-NEXT:    %[[DIV10:.+]] = udiv i32 %[[SUB9]], 2
 // IR-NEXT:    %[[SUB11:.+]] = sub i32 %[[DIV10]], 1
 // IR-NEXT:    store i32 %[[SUB11]], i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV_I]], align 4
 // IR-NEXT:    %[[TMP11:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
 // IR-NEXT:    %[[ADD13:.+]] = add i32 %[[TMP11]], 1
 // IR-NEXT:    store i32 %[[ADD13]], i32* %[[DOTCAPTURE_EXPR_12]], align 4
 // IR-NEXT:    %[[TMP12:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_12]], align 4
-// IR-NEXT:    %[[SUB15:.+]] = sub i32 %[[TMP12]], -2
-// IR-NEXT:    %[[DIV16:.+]] = udiv i32 %[[SUB15]], 3
+// IR-NEXT:    %[[SUB15:.+]] = sub i32 %[[TMP12]], -1
+// IR-NEXT:    %[[DIV16:.+]] = udiv i32 %[[SUB15]], 2
 // IR-NEXT:    %[[SUB17:.+]] = sub i32 %[[DIV16]], 1
 // IR-NEXT:    store i32 %[[SUB17]], i32* %[[DOTCAPTURE_EXPR_14]], align 4
-// IR-NEXT:    store i32 0, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I]], align 4
+// IR-NEXT:    store i32 0, i32* %[[DOTUNROLLED_IV__UNROLLED_IV_I]], align 4
 // IR-NEXT:    %[[TMP13:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_12]], align 4
 // IR-NEXT:    %[[CMP:.+]] = icmp ult i32 0, %[[TMP13]]
 // IR-NEXT:    br i1 %[[CMP]], label %[[OMP_PRECOND_THEN:.+]], label %[[OMP_PRECOND_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_PRECOND_THEN]]:
 // IR-NEXT:    store i32 0, i32* %[[DOTOMP_LB]], align 4
 // IR-NEXT:    %[[TMP14:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_14]], align 4
 // IR-NEXT:    store i32 %[[TMP14]], i32* %[[DOTOMP_UB]], align 4

 // IR-NEXT:    %[[TMP20:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
 // IR-NEXT:    %[[TMP21:.+]] = load i32, i32* %[[DOTOMP_UB]], align 4
 // IR-NEXT:    %[[ADD20:.+]] = add i32 %[[TMP21]], 1
 // IR-NEXT:    %[[CMP21:.+]] = icmp ult i32 %[[TMP20]], %[[ADD20]]
 // IR-NEXT:    br i1 %[[CMP21]], label %[[OMP_INNER_FOR_BODY:.+]], label %[[OMP_INNER_FOR_END:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_BODY]]:
 // IR-NEXT:    %[[TMP22:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
-// IR-NEXT:    %[[MUL:.+]] = mul i32 %[[TMP22]], 3
+// IR-NEXT:    %[[MUL:.+]] = mul i32 %[[TMP22]], 2
 // IR-NEXT:    %[[ADD22:.+]] = add i32 0, %[[MUL]]
-// IR-NEXT:    store i32 %[[ADD22]], i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    %[[TMP23:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    store i32 %[[TMP23]], i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
+// IR-NEXT:    store i32 %[[ADD22]], i32* %[[DOTUNROLLED_IV__UNROLLED_IV_I18]], align 4
+// IR-NEXT:    %[[TMP23:.+]] = load i32, i32* %[[DOTUNROLLED_IV__UNROLLED_IV_I18]], align 4
+// IR-NEXT:    store i32 %[[TMP23]], i32* %[[DOTUNROLL_INNER_IV__UNROLLED_IV_I]], align 4
 // IR-NEXT:    br label %[[FOR_COND:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_COND]]:
-// IR-NEXT:    %[[TMP24:.+]] = load i32, i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP25:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    %[[ADD23:.+]] = add i32 %[[TMP25]], 1
-// IR-NEXT:    %[[TMP26:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    %[[ADD24:.+]] = add i32 %[[TMP26]], 3
-// IR-NEXT:    %[[CMP25:.+]] = icmp ult i32 %[[ADD23]], %[[ADD24]]
-// IR-NEXT:    br i1 %[[CMP25]], label %[[COND_TRUE26:.+]], label %[[COND_FALSE28:.+]]
+// IR-NEXT:    %[[TMP24:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV__UNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[TMP25:.+]] = load i32, i32* %[[DOTUNROLLED_IV__UNROLLED_IV_I18]], align 4
+// IR-NEXT:    %[[ADD23:.+]] = add i32 %[[TMP25]], 2
+// IR-NEXT:    %[[CMP24:.+]] = icmp ule i32 %[[TMP24]], %[[ADD23]]
+// IR-NEXT:    br i1 %[[CMP24]], label %[[LAND_RHS:.+]], label %[[LAND_END:.+]]
 // IR-EMPTY:
-// IR-NEXT:  [[COND_TRUE26]]:
+// IR-NEXT:  [[LAND_RHS]]:
+// IR-NEXT:    %[[TMP26:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV__UNROLLED_IV_I]], align 4
 // IR-NEXT:    %[[TMP27:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_8]], align 4
-// IR-NEXT:    %[[ADD27:.+]] = add i32 %[[TMP27]], 1
-// IR-NEXT:    br label %[[COND_END30:.+]]
-// IR-EMPTY:
-// IR-NEXT:  [[COND_FALSE28]]:
-// IR-NEXT:    %[[TMP28:.+]] = load i32, i32* %[[DOTFLOOR_0_IV__FLOOR_0_IV_I18]], align 4
-// IR-NEXT:    %[[ADD29:.+]] = add i32 %[[TMP28]], 3
-// IR-NEXT:    br label %[[COND_END30]]
-// IR-EMPTY:
-// IR-NEXT:  [[COND_END30]]:
-// IR-NEXT:    %[[COND31:.+]] = phi i32 [ %[[ADD27]], %[[COND_TRUE26]] ], [ %[[ADD29]], %[[COND_FALSE28]] ]
-// IR-NEXT:    %[[CMP32:.+]] = icmp ult i32 %[[TMP24]], %[[COND31]]
-// IR-NEXT:    br i1 %[[CMP32]], label %[[FOR_BODY:.+]], label %[[FOR_END51:.+]]
+// IR-NEXT:    %[[ADD25:.+]] = add i32 %[[TMP27]], 1
+// IR-NEXT:    %[[CMP26:.+]] = icmp ule i32 %[[TMP26]], %[[ADD25]]
+// IR-NEXT:    br label %[[LAND_END]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_END]]:
+// IR-NEXT:    %[[TMP28:.+]] = phi i1 [ false, %[[FOR_COND]] ], [ %[[CMP26]], %[[LAND_RHS]] ]
+// IR-NEXT:    br i1 %[[TMP28]], label %[[FOR_BODY:.+]], label %[[FOR_END41:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_BODY]]:
-// IR-NEXT:    %[[TMP29:.+]] = load i32, i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[MUL33:.+]] = mul i32 %[[TMP29]], 4
-// IR-NEXT:    %[[ADD34:.+]] = add i32 0, %[[MUL33]]
-// IR-NEXT:    store i32 %[[ADD34]], i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP30:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    store i32 %[[TMP30]], i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    br label %[[FOR_COND35:.+]]
-// IR-EMPTY:
-// IR-NEXT:  [[FOR_COND35]]:
-// IR-NEXT:    %[[TMP31:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[ADD36:.+]] = add i32 %[[TMP32]], 1
-// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[ADD37:.+]] = add nsw i32 %[[TMP33]], 4
-// IR-NEXT:    %[[CMP38:.+]] = icmp ult i32 %[[ADD36]], %[[ADD37]]
-// IR-NEXT:    br i1 %[[CMP38]], label %[[COND_TRUE39:.+]], label %[[COND_FALSE41:.+]]
+// IR-NEXT:    %[[TMP29:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV__UNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[MUL27:.+]] = mul i32 %[[TMP29]], 2
+// IR-NEXT:    %[[ADD28:.+]] = add i32 0, %[[MUL27]]
+// IR-NEXT:    store i32 %[[ADD28]], i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[TMP30:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    store i32 %[[TMP30]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    br label %[[FOR_COND29:.+]]
+// IR-EMPTY:
+// IR-NEXT:  [[FOR_COND29]]:
+// IR-NEXT:    %[[TMP31:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[TMP32:.+]] = load i32, i32* %[[DOTUNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[ADD30:.+]] = add i32 %[[TMP32]], 2
+// IR-NEXT:    %[[CMP31:.+]] = icmp ule i32 %[[TMP31]], %[[ADD30]]
+// IR-NEXT:    br i1 %[[CMP31]], label %[[LAND_RHS32:.+]], label %[[LAND_END35:.+]]
 // IR-EMPTY:
-// IR-NEXT:  [[COND_TRUE39]]:
+// IR-NEXT:  [[LAND_RHS32]]:
+// IR-NEXT:    %[[TMP33:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    %[[TMP34:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_3]], align 4
-// IR-NEXT:    %[[ADD40:.+]] = add i32 %[[TMP34]], 1
-// IR-NEXT:    br label %[[COND_END43:.+]]
+// IR-NEXT:    %[[ADD33:.+]] = add i32 %[[TMP34]], 1
+// IR-NEXT:    %[[CMP34:.+]] = icmp ule i32 %[[TMP33]], %[[ADD33]]
+// IR-NEXT:    br label %[[LAND_END35]]
+// IR-EMPTY:
+// IR-NEXT:  [[LAND_END35]]:
+// IR-NEXT:    %[[TMP35:.+]] = phi i1 [ false, %[[FOR_COND29]] ], [ %[[CMP34]], %[[LAND_RHS32]] ]
+// IR-NEXT:    br i1 %[[TMP35]], label %[[FOR_BODY36:.+]], label %[[FOR_END:.+]]
 // IR-EMPTY:
-// IR-NEXT:  [[COND_FALSE41]]:
-// IR-NEXT:    %[[TMP35:.+]] = load i32, i32* %[[DOTFLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[ADD42:.+]] = add nsw i32 %[[TMP35]], 4
-// IR-NEXT:    br label %[[COND_END43]]
-// IR-EMPTY:
-// IR-NEXT:  [[COND_END43]]:
-// IR-NEXT:    %[[COND44:.+]] = phi i32 [ %[[ADD40]], %[[COND_TRUE39]] ], [ %[[ADD42]], %[[COND_FALSE41]] ]
-// IR-NEXT:    %[[CMP45:.+]] = icmp ult i32 %[[TMP31]], %[[COND44]]
-// IR-NEXT:    br i1 %[[CMP45]], label %[[FOR_BODY46:.+]], label %[[FOR_END:.+]]
-// IR-EMPTY:
-// IR-NEXT:  [[FOR_BODY46]]:
+// IR-NEXT:  [[FOR_BODY36]]:
 // IR-NEXT:    %[[TMP36:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_]], align 4
-// IR-NEXT:    %[[TMP37:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
+// IR-NEXT:    %[[TMP37:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
 // IR-NEXT:    %[[TMP38:.+]] = load i32, i32* %[[DOTCAPTURE_EXPR_2]], align 4
-// IR-NEXT:    %[[MUL47:.+]] = mul i32 %[[TMP37]], %[[TMP38]]
-// IR-NEXT:    %[[ADD48:.+]] = add i32 %[[TMP36]], %[[MUL47]]
-// IR-NEXT:    store i32 %[[ADD48]], i32* %[[I]], align 4
+// IR-NEXT:    %[[MUL37:.+]] = mul i32 %[[TMP37]], %[[TMP38]]
+// IR-NEXT:    %[[ADD38:.+]] = add i32 %[[TMP36]], %[[MUL37]]
+// IR-NEXT:    store i32 %[[ADD38]], i32* %[[I]], align 4
 // IR-NEXT:    %[[TMP39:.+]] = load i32, i32* %[[START_ADDR]], align 4
 // IR-NEXT:    %[[TMP40:.+]] = load i32, i32* %[[END_ADDR]], align 4
 // IR-NEXT:    %[[TMP41:.+]] = load i32, i32* %[[STEP_ADDR]], align 4
 // IR-NEXT:    %[[TMP42:.+]] = load i32, i32* %[[I]], align 4
 // IR-NEXT:    call void (...) @body(i32 %[[TMP39]], i32 %[[TMP40]], i32 %[[TMP41]], i32 %[[TMP42]])
 // IR-NEXT:    br label %[[FOR_INC:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_INC]]:
-// IR-NEXT:    %[[TMP43:.+]] = load i32, i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    %[[INC:.+]] = add nsw i32 %[[TMP43]], 1
-// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTTILE_0_IV_I]], align 4
-// IR-NEXT:    br label %[[FOR_COND35]], !llvm.loop ![[LOOP2:[0-9]+]]
+// IR-NEXT:    %[[TMP43:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    %[[INC:.+]] = add i32 %[[TMP43]], 1
+// IR-NEXT:    store i32 %[[INC]], i32* %[[DOTUNROLL_INNER_IV_I]], align 4
+// IR-NEXT:    br label %[[FOR_COND29]], !llvm.loop ![[LOOP2:[0-9]+]]
 // IR-EMPTY:
 // IR-NEXT:  [[FOR_END]]:
-// IR-NEXT:    br label %[[FOR_INC49:.+]]
+// IR-NEXT:    br label %[[FOR_INC39:.+]]
 // IR-EMPTY:
-// IR-NEXT:  [[FOR_INC49]]:
-// IR-NEXT:    %[[TMP44:.+]] = load i32, i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
-// IR-NEXT:    %[[INC50:.+]] = add i32 %[[TMP44]], 1
-// IR-NEXT:    store i32 %[[INC50]], i32* %[[DOTTILE_0_IV__FLOOR_0_IV_I]], align 4
-// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP4:[0-9]+]]
+// IR-NEXT:  [[FOR_INC39]]:
+// IR-NEXT:    %[[TMP44:.+]] = load i32, i32* %[[DOTUNROLL_INNER_IV__UNROLLED_IV_I]], align 4
+// IR-NEXT:    %[[INC40:.+]] = add i32 %[[TMP44]], 1
+// IR-NEXT:    store i32 %[[INC40]], i32* %[[DOTUNROLL_INNER_IV__UNROLLED_IV_I]], align 4
+// IR-NEXT:    br label %[[FOR_COND]], !llvm.loop ![[LOOP5:[0-9]+]]
 // IR-EMPTY:
-// IR-NEXT:  [[FOR_END51]]:
+// IR-NEXT:  [[FOR_END41]]:
 // IR-NEXT:    br label %[[OMP_BODY_CONTINUE:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_BODY_CONTINUE]]:
 // IR-NEXT:    br label %[[OMP_INNER_FOR_INC:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_INC]]:
 // IR-NEXT:    %[[TMP45:.+]] = load i32, i32* %[[DOTOMP_IV]], align 4
-// IR-NEXT:    %[[ADD52:.+]] = add i32 %[[TMP45]], 1
-// IR-NEXT:    store i32 %[[ADD52]], i32* %[[DOTOMP_IV]], align 4
+// IR-NEXT:    %[[ADD42:.+]] = add i32 %[[TMP45]], 1
+// IR-NEXT:    store i32 %[[ADD42]], i32* %[[DOTOMP_IV]], align 4
 // IR-NEXT:    br label %[[OMP_INNER_FOR_COND]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_INNER_FOR_END]]:
 // IR-NEXT:    br label %[[OMP_LOOP_EXIT:.+]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_LOOP_EXIT]]:
 // IR-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @1, i32 %[[TMP0]])
 // IR-NEXT:    br label %[[OMP_PRECOND_END]]
 // IR-EMPTY:
 // IR-NEXT:  [[OMP_PRECOND_END]]:
 // IR-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @3, i32 %[[TMP0]])
 // IR-NEXT:    ret void
 // IR-NEXT:  }
 extern "C" void func(int start, int end, int step) {
   #pragma omp for
-#pragma omp tile sizes(3)
-#pragma omp tile sizes(4)
-  for (int i = start; i < end; i += step)
+  #pragma omp unroll partial
+  #pragma omp unroll partial
+  for (int i = start; i < end; i+=step)
     body(start, end, step, i);
 }
 
 #endif /* HEADER */
+
+
+// IR: ![[LOOP2]] = distinct !{![[LOOP2]], ![[LOOPPROP3:[0-9]+]], ![[LOOPPROP4:[0-9]+]]}
+// IR: ![[LOOPPROP3]] = !{!"llvm.loop.mustprogress"}
+// IR: ![[LOOPPROP4]] = !{!"llvm.loop.unroll.count", i32 2}
+// IR: ![[LOOP5]] = distinct !{![[LOOP5]], ![[LOOPPROP3]], ![[LOOPPROP4]]}

clang/test/OpenMP/unroll_messages.cpp

+// RUN: %clang_cc1 -triple x86_64-pc-linux-gnu -std=c++17 -fopenmp -fopenmp-version=51 -fsyntax-only -Wuninitialized -verify %s
+
+void func(int n) {
+  // expected-error@+2 {{statement after '#pragma omp unroll' must be a for loop}}
+  #pragma omp unroll
+  func(n);
+
+  // expected-error@+2 {{statement after '#pragma omp unroll' must be a for loop}}
+  #pragma omp unroll
+    ;
+
+  // expected-error@+2 {{the loop condition expression depends on the current loop control variable}}
+  #pragma omp unroll
+  for (int i = 0; i < 2*(i-4); ++i) {}
+
+  // expected-error@+2 {{condition of OpenMP for loop must be a relational comparison ('<', '<=', '>', '>=', or '!=') of loop variable 'i'}}
+  #pragma omp unroll
+  for (int i = 0; i/3 < 7; ++i) {}
+
+  // expected-warning@+1 {{extra tokens at the end of '#pragma omp unroll' are ignored}}
+  #pragma omp unroll foo
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{expected expression}} expected-error@+1 {{expected ')'}} expected-note@+1 {{to match this '('}}
+  #pragma omp unroll partial(
+  for (int i = 0; i < n; ++i) {}
+  
+  // expected-error@+1 {{expected ')'}} expected-note@+1 {{to match this '('}}
+  #pragma omp unroll partial(4
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{expected expression}} expected-error@+1 {{expected ')'}} expected-note@+1 {{to match this '('}}
+  #pragma omp unroll partial(4+
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{expected expression}} expected-error@+1 {{expected ')'}} expected-note@+1 {{to match this '('}}
+  #pragma omp unroll partial(for)
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{integral constant expression must have integral or unscoped enumeration type, not 'void (int)'}}
+  #pragma omp unroll partial(func)
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{expected expression}}
+  #pragma omp unroll partial()
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{expected ')'}} expected-note@+1 {{to match this '('}}
+  #pragma omp unroll partial(4,4)
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+3 {{expression is not an integral constant expression}} expected-note@+3 {{read of non-const variable 'a' is not allowed in a constant expression}}
+  // expected-note@+1 {{declared here}}
+  int a;
+  #pragma omp unroll partial(a)
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{argument to 'partial' clause must be a strictly positive integer value}} 
+  #pragma omp unroll partial(0)
+  for (int i = 0; i < n; ++i) {}
+    
+  // expected-error@+1 {{directive '#pragma omp unroll' cannot contain more than one 'partial' clause}} 
+  #pragma omp unroll partial partial
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{directive '#pragma omp unroll' cannot contain more than one 'partial' clause}} 
+  #pragma omp unroll partial(4) partial
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{directive '#pragma omp unroll' cannot contain more than one 'full' clause}}
+  #pragma omp unroll full full
+  for (int i = 0; i < 128; ++i) {}
+
+  // expected-error@+1 {{'full' and 'partial' clause are mutually exclusive and may not appear on the same directive}} expected-note@+1 {{'partial' clause is specified here}}
+  #pragma omp unroll partial full
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+1 {{'partial' and 'full' clause are mutually exclusive and may not appear on the same directive}} expected-note@+1 {{'full' clause is specified here}}
+  #pragma omp unroll full partial
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+2 {{loop to be fully unrolled must have a constant trip count}} expected-note@+1 {{'#pragma omp unroll full' directive found here}}
+  #pragma omp unroll full
+  for (int i = 0; i < n; ++i) {}
+
+  // expected-error@+2 {{statement after '#pragma omp for' must be a for loop}}
+  #pragma omp for
+  #pragma omp unroll
+  for (int i = 0; i < n; ++i) {}
+
+    // expected-error@+2 {{statement after '#pragma omp for' must be a for loop}}
+  #pragma omp for
+  #pragma omp unroll full
+  for (int i = 0; i < 128; ++i) {}
+
+  // expected-error@+2 {{statement after '#pragma omp unroll' must be a for loop}}
+  #pragma omp unroll
+  #pragma omp unroll
+  for (int i = 0; i < n; ++i) {}
+  
+  // expected-error@+2 {{statement after '#pragma omp tile' must be a for loop}}
+  #pragma omp tile sizes(4)
+  #pragma omp unroll
+  for (int i = 0; i < n; ++i) {}
+  
+  // expected-error@+4 {{expected 2 for loops after '#pragma omp for', but found only 1}} 
+  // expected-note@+1 {{as specified in 'collapse' clause}}
+  #pragma omp for collapse(2)
+  for (int i = 0; i < n; ++i) {
+    #pragma omp unroll full
+    for (int j = 0; j < 128; ++j) {}
+  }
+}
+
+
+template<typename T, int Factor>
+void templated_func(int n) {
+  // expected-error@+1 {{argument to 'partial' clause must be a strictly positive integer value}} 
+  #pragma omp unroll partial(Factor)
+  for (T i = 0; i < n; ++i) {}
+
+  // expected-error@+2 {{loop to be fully unrolled must have a constant trip count}} expected-note@+1 {{'#pragma omp unroll full' directive found here}}
+  #pragma omp unroll full
+  for (int i = 0; i < n; i-=Factor) {}
+}
+
+void template_inst(int n) {
+  // expected-note@+1 {{in instantiation of function template specialization 'templated_func<int, -1>' requested here}}
+  templated_func<int, -1>(n);
+}

clang/tools/libclang/CIndex.cpp

public:
   void VisitOpaqueValueExpr(const OpaqueValueExpr *E);
   void VisitLambdaExpr(const LambdaExpr *E);
   void VisitOMPExecutableDirective(const OMPExecutableDirective *D);
   void VisitOMPLoopBasedDirective(const OMPLoopBasedDirective *D);
   void VisitOMPLoopDirective(const OMPLoopDirective *D);
   void VisitOMPParallelDirective(const OMPParallelDirective *D);
   void VisitOMPSimdDirective(const OMPSimdDirective *D);
   void VisitOMPTileDirective(const OMPTileDirective *D);
+  void VisitOMPUnrollDirective(const OMPUnrollDirective *D);
   void VisitOMPForDirective(const OMPForDirective *D);
   void VisitOMPForSimdDirective(const OMPForSimdDirective *D);
   void VisitOMPSectionsDirective(const OMPSectionsDirective *D);
   void VisitOMPSectionDirective(const OMPSectionDirective *D);
   void VisitOMPSingleDirective(const OMPSingleDirective *D);
   void VisitOMPMasterDirective(const OMPMasterDirective *D);
   void VisitOMPCriticalDirective(const OMPCriticalDirective *D);
   void VisitOMPParallelForDirective(const OMPParallelForDirective *D);
void OMPClauseEnqueue::VisitOMPSimdlenClause(const OMPSimdlenClause *C) {
   Visitor->AddStmt(C->getSimdlen());
 }
 
 void OMPClauseEnqueue::VisitOMPSizesClause(const OMPSizesClause *C) {
   for (auto E : C->getSizesRefs())
     Visitor->AddStmt(E);
 }
 
+
+
+
+
+void OMPClauseEnqueue::VisitOMPFullClause(const OMPFullClause *C) {}
+
+
+void OMPClauseEnqueue::VisitOMPPartialClause(const OMPPartialClause *C) {
+  Visitor->AddStmt(C->getFactor());
+}
+
+
+
+
+
+
+
 void OMPClauseEnqueue::VisitOMPAllocatorClause(const OMPAllocatorClause *C) {
   Visitor->AddStmt(C->getAllocator());
 }
 
 void OMPClauseEnqueue::VisitOMPCollapseClause(const OMPCollapseClause *C) {
   Visitor->AddStmt(C->getNumForLoops());
 }
 

 void EnqueueVisitor::VisitOMPSimdDirective(const OMPSimdDirective *D) {
   VisitOMPLoopDirective(D);
 }
 
 void EnqueueVisitor::VisitOMPTileDirective(const OMPTileDirective *D) {
   VisitOMPLoopBasedDirective(D);
 }
 
+
+void EnqueueVisitor::VisitOMPUnrollDirective(const OMPUnrollDirective *D) {
+  VisitOMPLoopBasedDirective(D);
+}
+
+
 void EnqueueVisitor::VisitOMPForDirective(const OMPForDirective *D) {
   VisitOMPLoopDirective(D);
 }
 
 void EnqueueVisitor::VisitOMPForSimdDirective(const OMPForSimdDirective *D) {
   VisitOMPLoopDirective(D);
 }
 
CXString clang_getCursorKindSpelling(enum CXCursorKind Kind) {
   case CXCursor_OMPCanonicalLoop:
     return cxstring::createRef("OMPCanonicalLoop");
   case CXCursor_OMPParallelDirective:
     return cxstring::createRef("OMPParallelDirective");
   case CXCursor_OMPSimdDirective:
     return cxstring::createRef("OMPSimdDirective");
   case CXCursor_OMPTileDirective:
     return cxstring::createRef("OMPTileDirective");
+  case CXCursor_OMPUnrollDirective:
+    return cxstring::createRef("OMPUnrollDirective");
   case CXCursor_OMPForDirective:
     return cxstring::createRef("OMPForDirective");
   case CXCursor_OMPForSimdDirective:
     return cxstring::createRef("OMPForSimdDirective");
   case CXCursor_OMPSectionsDirective:
     return cxstring::createRef("OMPSectionsDirective");
   case CXCursor_OMPSectionDirective:
     return cxstring::createRef("OMPSectionDirective");

clang/tools/libclang/CXCursor.cpp

case Stmt::OMPParallelDirectiveClass:
     K = CXCursor_OMPParallelDirective;
     break;
   case Stmt::OMPSimdDirectiveClass:
     K = CXCursor_OMPSimdDirective;
     break;
   case Stmt::OMPTileDirectiveClass:
     K = CXCursor_OMPTileDirective;
     break;
+  case Stmt::OMPUnrollDirectiveClass:
+    K = CXCursor_OMPUnrollDirective;
+    break;
   case Stmt::OMPForDirectiveClass:
     K = CXCursor_OMPForDirective;
     break;
   case Stmt::OMPForSimdDirectiveClass:
     K = CXCursor_OMPForSimdDirective;
     break;
   case Stmt::OMPSectionsDirectiveClass:
     K = CXCursor_OMPSectionsDirective;

llvm/include/llvm/Frontend/OpenMP/OMP.td

def OMPC_Default : Clause<"default"> {
   let clangClass = "OMPDefaultClause";
   let flangClass = "OmpDefaultClause";
 }
 def OMPC_Private : Clause<"private"> {
   let clangClass = "OMPPrivateClause";
   let flangClass = "OmpObjectList";
 }
 def OMPC_Sizes: Clause<"sizes"> { let clangClass = "OMPSizesClause"; }
+def OMPC_Full: Clause<"full"> { let clangClass = "OMPFullClause"; }
+def OMPC_Partial: Clause<"partial"> { let clangClass = "OMPPartialClause"; }
 def OMPC_FirstPrivate : Clause<"firstprivate"> {
   let clangClass = "OMPFirstprivateClause";
   let flangClass = "OmpObjectList";
 }
 def OMPC_LastPrivate : Clause<"lastprivate"> {
   let clangClass = "OMPLastprivateClause";
   let flangClass = "OmpObjectList";
 }
let allowedOnceClauses = [
     VersionedClause<OMPC_If, 50>,
   ];
 }
 def OMP_Tile : Directive<"tile"> {
   let allowedOnceClauses = [
     VersionedClause<OMPC_Sizes, 51>,
   ];
 }
+def OMP_Unroll : Directive<"unroll"> {
+  let allowedOnceClauses = [
+    VersionedClause<OMPC_Full, 51>,
+    VersionedClause<OMPC_Partial, 51>,
+  ];
+}
 def OMP_For : Directive<"for"> {
   let allowedClauses = [
     VersionedClause<OMPC_Private>,
     VersionedClause<OMPC_LastPrivate>,
     VersionedClause<OMPC_FirstPrivate>,
     VersionedClause<OMPC_Reduction>,
     VersionedClause<OMPC_Collapse>,
     VersionedClause<OMPC_Schedule>,