[Clang][OpenMP] Add the codegen support for `atomic compare capture`

[tianshilei1992]

Summary

This patch adds the codegen support for atomic compare capture in clang.

[tianshilei1992]

add tests

[ABataev]

clang/include/clang/AST/StmtOpenMP.h

2835–2848

Transform these booleans to bitfields?

clang/lib/Sema/SemaOpenMP.cpp

11556

Need to be careful here, probably may need to cast to resulting type explicitly to avoid problems between int/float/etc. conversions.

[tianshilei1992]

clang/lib/Sema/SemaOpenMP.cpp

11556

I think we don't need that as D and E are supposed to be rvalue, and as shown here I only ignore implicit casts for rvalue.

[tianshilei1992]

clang/include/clang/AST/StmtOpenMP.h

2835–2848

Will do in another patch.

[nikic]

clang/lib/CodeGen/CGStmtOpenMP.cpp

6056

Please avoid adding new calls to getPointerElementType(). See https://github.com/llvm/llvm-project/commit/b1863d82454b2905db8b492bea0ce8a260362645 for a way to avoid it in this context.

[tianshilei1992]

rebase and fix comments

[tianshilei1992]

clang/lib/CodeGen/CGStmtOpenMP.cpp

6056

Got it. Thanks for the info.

[tianshilei1992]

rebase

[ABataev]

clang/include/clang/AST/StmtOpenMP.h

2917–2920

There are too many params already, better to gather them into a struct/class

clang/lib/CodeGen/CGStmtOpenMP.cpp

6216

Can this be fixed in a separate patch?

[tianshilei1992]

clang/include/clang/AST/StmtOpenMP.h

2917–2920

Sure, we could do that.

clang/lib/CodeGen/CGStmtOpenMP.cpp

6216

Well, I think it's part of this patch because we can't tell if it's compare or compare capture before, but now we can. If we really want to do that, we can have another patch including all changes in this patch related to OMPAtomicDirective.

[ABataev]

clang/lib/CodeGen/CGStmtOpenMP.cpp

6216

Kind of NFC? Would be good

[tianshilei1992]

rebase

[tianshilei1992]

clang/include/clang/AST/StmtOpenMP.h

2835–2848

In D123862.

clang/lib/CodeGen/CGStmtOpenMP.cpp

6216

I think we don't need an extra NFC patch for this part because it's not gonna be very neat w/o some other code in this patch.

[tianshilei1992]

rebase

[ABataev]

clang/lib/Sema/SemaOpenMP.cpp

11423

Why do we need to use IgnoreImpCasts() here and in other places?

[tianshilei1992]

clang/lib/Sema/SemaOpenMP.cpp

11423

Clang usually inserts implicit casts. For example, if we have:

char a, b, c;
#pragma omp atomic compare capture
  { r = a; if (a > c) { a = b; } }

Clang inserts an implicit cast from char to int for all statements except r = a. In this case, what should be the right solution? I'm not quite sure actually.

[tianshilei1992]

ping

clang/lib/Sema/SemaOpenMP.cpp

11423

https://godbolt.org/z/a6WWdx581
This shows how the AST looks like.

[ABataev]

clang/lib/Sema/SemaOpenMP.cpp

11423

I don't see casts to int in a=b (BO points to this expression,right?). The only cast, that maybe should be removed here, is LValueToRValue

[tianshilei1992]

fix comments

[ABataev]

clang/lib/CodeGen/CGStmtOpenMP.cpp

6049

1. auto *CI
2. What if this is not a constant, but just a value with int type? Is this possible?

clang/lib/Sema/SemaOpenMP.cpp

11556

Same question as before. It is assignment BO, do you really need to remove casts here?

11570

Maybe, IgnoreImplicitAsWritten?

[tianshilei1992]

clang/lib/CodeGen/CGStmtOpenMP.cpp

6049

What if this is not a constant, but just a value with int type? Is this possible?

I'm thinking of a similar case.

void foo() {
  int e, d;
  char x, v;
  if (x == e)
    x = d;
}

In this case, there are two casts:

1. In expression x == e, cast x to int.
2. In expression x = d, cast d to char.

We cannot ignore both of them. For the 2nd, no question. For the expression x == e, the problem is in e instead of x. e here is always int, further causing issue in codegen.

I'm thinking for those lvalues, we can only set them where they are used as lvalues in case any type lifting. For those rvalues, if their types are not same as x, but compatible with x, we have to insert cast, including truncate, before feed them into codegen. Does it sound good?

[tianshilei1992]

rebase

[tianshilei1992]

fix the wrong use of AtomicOpValue

[tianshilei1992]

ping

[ABataev]

LG

[tianshilei1992]

rebase

clang/include/clang/AST/StmtOpenMP.h

class OMPAtomicDirective : public OMPExecutableDirective {
   /// have atomic expressions of forms
   /// \code
   /// x = x binop expr;
   /// x = expr binop x;
   /// \endcode
   /// This field is true for the first form of the expression and false for the
   /// second. Required for correct codegen of non-associative operations (like
   /// << or >>).
   bool IsXLHSInRHSPart = false;
   /// Used for 'atomic update' or 'atomic capture' constructs. They may
   /// have atomic expressions of forms
   /// \code
   /// v = x; <update x>;
   /// <update x>; v = x;
   /// \endcode
   /// This field is true for the first(postfix) form of the expression and false
   /// otherwise.
   bool IsPostfixUpdate = false;
+  /// True if 'v' is updated only when the condition is false (compare capture
+  /// only).
+  bool IsFailOnly = false;
 
   /// Build directive with the given start and end location.
   ///
   /// \param StartLoc Starting location of the directive kind.
   /// \param EndLoc Ending location of the directive.
   ///
   OMPAtomicDirective(SourceLocation StartLoc, SourceLocation EndLoc)
       : OMPExecutableDirective(OMPAtomicDirectiveClass, llvm::omp::OMPD_atomic,
                                StartLoc, EndLoc) {}
 
   /// Build an empty directive.
   ///
   explicit OMPAtomicDirective()
       : OMPExecutableDirective(OMPAtomicDirectiveClass, llvm::omp::OMPD_atomic,
                                SourceLocation(), SourceLocation()) {}
 
   enum DataPositionTy : size_t {
     POS_X = 0,
     POS_V,
     POS_E,
     POS_UpdateExpr,
     POS_D,
     POS_Cond,
+    POS_R,
   };
 
   /// Set 'x' part of the associated expression/statement.
   void setX(Expr *X) { Data->getChildren()[DataPositionTy::POS_X] = X; }
   /// Set helper expression of the form
   /// 'OpaqueValueExpr(x) binop OpaqueValueExpr(expr)' or
   /// 'OpaqueValueExpr(expr) binop OpaqueValueExpr(x)'.
   void setUpdateExpr(Expr *UE) {
     Data->getChildren()[DataPositionTy::POS_UpdateExpr] = UE;
   }
   /// Set 'v' part of the associated expression/statement.
   void setV(Expr *V) { Data->getChildren()[DataPositionTy::POS_V] = V; }
+  /// Set 'r' part of the associated expression/statement.
+  void setR(Expr *R) { Data->getChildren()[DataPositionTy::POS_R] = R; }
   /// Set 'expr' part of the associated expression/statement.
   void setExpr(Expr *E) { Data->getChildren()[DataPositionTy::POS_E] = E; }
   /// Set 'd' part of the associated expression/statement.
   void setD(Expr *D) { Data->getChildren()[DataPositionTy::POS_D] = D; }
   /// Set conditional expression in `atomic compare`.
   void setCond(Expr *C) { Data->getChildren()[DataPositionTy::POS_Cond] = C; }
 
 public:
   /// Creates directive with a list of \a Clauses and 'x', 'v' and 'expr'
   /// parts of the atomic construct (see Section 2.12.6, atomic Construct, for
   /// detailed description of 'x', 'v' and 'expr').
   ///
   /// \param C AST context.
   /// \param StartLoc Starting location of the directive kind.
   /// \param EndLoc Ending Location of the directive.
   /// \param Clauses List of clauses.
   /// \param AssociatedStmt Statement, associated with the directive.
   /// \param X 'x' part of the associated expression/statement.
   /// \param V 'v' part of the associated expression/statement.
+  /// \param R 'r' part of the associated expression/statement.
   /// \param E 'expr' part of the associated expression/statement.
   /// \param UE Helper expression of the form
   /// 'OpaqueValueExpr(x) binop OpaqueValueExpr(expr)' or
   /// 'OpaqueValueExpr(expr) binop OpaqueValueExpr(x)'.
   /// \param D 'd' part of the associated expression/statement.
   /// \param Cond Conditional expression in `atomic compare` construct.
   /// \param IsXLHSInRHSPart true if \a UE has the first form and false if the
   /// second.
   /// \param IsPostfixUpdate true if original value of 'x' must be stored in
   /// 'v', not an updated one.
   static OMPAtomicDirective *
   Create(const ASTContext &C, SourceLocation StartLoc, SourceLocation EndLoc,
          ArrayRef<OMPClause *> Clauses, Stmt *AssociatedStmt, Expr *X, Expr *V,
-         Expr *E, Expr *UE, Expr *D, Expr *Cond, bool IsXLHSInRHSPart,
-         bool IsPostfixUpdate);
+         Expr *R, Expr *E, Expr *UE, Expr *D, Expr *Cond, bool IsXLHSInRHSPart,
+         bool IsPostfixUpdate, bool IsFailOnly);
 
   /// Creates an empty directive with the place for \a NumClauses
   /// clauses.
   ///
   /// \param C AST context.
   /// \param NumClauses Number of clauses.
   ///
   static OMPAtomicDirective *CreateEmpty(const ASTContext &C,
public:
   const Expr *getUpdateExpr() const {
     return cast_or_null<Expr>(
         Data->getChildren()[DataPositionTy::POS_UpdateExpr]);
   }
   /// Return true if helper update expression has form
   /// 'OpaqueValueExpr(x) binop OpaqueValueExpr(expr)' and false if it has form
   /// 'OpaqueValueExpr(expr) binop OpaqueValueExpr(x)'.
   bool isXLHSInRHSPart() const { return IsXLHSInRHSPart; }
+  /// Return true if 'v' is updated only when the condition is evaluated false
+  /// (compare capture only).
+  bool isFailOnly() const { return IsFailOnly; }
   /// Return true if 'v' expression must be updated to original value of
   /// 'x', false if 'v' must be updated to the new value of 'x'.
   bool isPostfixUpdate() const { return IsPostfixUpdate; }
   /// Get 'v' part of the associated expression/statement.
   Expr *getV() {
     return cast_or_null<Expr>(Data->getChildren()[DataPositionTy::POS_V]);
   }
   const Expr *getV() const {
     return cast_or_null<Expr>(Data->getChildren()[DataPositionTy::POS_V]);
   }
+  /// Get 'r' part of the associated expression/statement.
+  Expr *getR() {
+    return cast_or_null<Expr>(Data->getChildren()[DataPositionTy::POS_R]);
+  }
+  const Expr *getR() const {
+    return cast_or_null<Expr>(Data->getChildren()[DataPositionTy::POS_R]);
+  }
   /// Get 'expr' part of the associated expression/statement.
   Expr *getExpr() {
     return cast_or_null<Expr>(Data->getChildren()[DataPositionTy::POS_E]);
   }
   const Expr *getExpr() const {
     return cast_or_null<Expr>(Data->getChildren()[DataPositionTy::POS_E]);
   }
   /// Get 'd' part of the associated expression/statement.

clang/lib/AST/StmtOpenMP.cpp

 OMPOrderedDirective *OMPOrderedDirective::CreateEmpty(const ASTContext &C,
                                                       unsigned NumClauses,
                                                       bool IsStandalone,
                                                       EmptyShell) {
   return createEmptyDirective<OMPOrderedDirective>(C, NumClauses,
                                                    !IsStandalone);
 }
 
-OMPAtomicDirective *
-OMPAtomicDirective::Create(const ASTContext &C, SourceLocation StartLoc,
-                           SourceLocation EndLoc, ArrayRef<OMPClause *> Clauses,
-                           Stmt *AssociatedStmt, Expr *X, Expr *V, Expr *E,
-                           Expr *UE, Expr *D, Expr *Cond, bool IsXLHSInRHSPart,
-                           bool IsPostfixUpdate) {
+OMPAtomicDirective *OMPAtomicDirective::Create(
+    const ASTContext &C, SourceLocation StartLoc, SourceLocation EndLoc,
+    ArrayRef<OMPClause *> Clauses, Stmt *AssociatedStmt, Expr *X, Expr *V,
+    Expr *R, Expr *E, Expr *UE, Expr *D, Expr *Cond, bool IsXLHSInRHSPart,
+    bool IsPostfixUpdate, bool IsFailOnly) {
   auto *Dir = createDirective<OMPAtomicDirective>(
-      C, Clauses, AssociatedStmt, /*NumChildren=*/6, StartLoc, EndLoc);
+      C, Clauses, AssociatedStmt, /*NumChildren=*/7, StartLoc, EndLoc);
   Dir->setX(X);
   Dir->setV(V);
+  Dir->setR(R);
   Dir->setExpr(E);
   Dir->setUpdateExpr(UE);
   Dir->setD(D);
   Dir->setCond(Cond);
   Dir->IsXLHSInRHSPart = IsXLHSInRHSPart;
   Dir->IsPostfixUpdate = IsPostfixUpdate;
+  Dir->IsFailOnly = IsFailOnly;
   return Dir;
 }
 
 OMPAtomicDirective *OMPAtomicDirective::CreateEmpty(const ASTContext &C,
                                                     unsigned NumClauses,
                                                     EmptyShell) {
   return createEmptyDirective<OMPAtomicDirective>(
-      C, NumClauses, /*HasAssociatedStmt=*/true, /*NumChildren=*/6);
+      C, NumClauses, /*HasAssociatedStmt=*/true, /*NumChildren=*/7);
 }
 
 OMPTargetDirective *OMPTargetDirective::Create(const ASTContext &C,
                                                SourceLocation StartLoc,
                                                SourceLocation EndLoc,
                                                ArrayRef<OMPClause *> Clauses,
                                                Stmt *AssociatedStmt) {
   return createDirective<OMPTargetDirective>(

clang/lib/CodeGen/CGStmtOpenMP.cpp

if (CGF.CGM.getLangOpts().OpenMP < 51) {
     case llvm::AtomicOrdering::Unordered:
       llvm_unreachable("Unexpected ordering.");
     }
   }
 }
 
 static void emitOMPAtomicCompareExpr(CodeGenFunction &CGF,
                                      llvm::AtomicOrdering AO, const Expr *X,
+                                     const Expr *V, const Expr *R,
                                      const Expr *E, const Expr *D,
                                      const Expr *CE, bool IsXBinopExpr,
+                                     bool IsPostfixUpdate, bool IsFailOnly,
                                      SourceLocation Loc) {
   llvm::OpenMPIRBuilder &OMPBuilder =
       CGF.CGM.getOpenMPRuntime().getOMPBuilder();
 
   OMPAtomicCompareOp Op;
   assert(isa<BinaryOperator>(CE) && "CE is not a BinaryOperator");
   switch (cast<BinaryOperator>(CE)->getOpcode()) {
   case BO_EQ:
     Op = OMPAtomicCompareOp::EQ;
     break;
   case BO_LT:
     Op = OMPAtomicCompareOp::MIN;
     break;
   case BO_GT:
     Op = OMPAtomicCompareOp::MAX;
     break;
   default:
     llvm_unreachable("unsupported atomic compare binary operator");
   }
 
   LValue XLVal = CGF.EmitLValue(X);
   Address XAddr = XLVal.getAddress(CGF);
   llvm::Value *EVal = CGF.EmitScalarExpr(E);
   llvm::Value *DVal = D ? CGF.EmitScalarExpr(D) : nullptr;
-
   llvm::OpenMPIRBuilder::AtomicOpValue XOpVal{
       XAddr.getPointer(), XAddr.getElementType(),
       X->getType().isVolatileQualified(),
       X->getType()->hasSignedIntegerRepresentation()};
+  llvm::OpenMPIRBuilder::AtomicOpValue VOpVal;
+  if (V) {
+    LValue LV = CGF.EmitLValue(V);
+    Address Addr = LV.getAddress(CGF);
+    VOpVal = {Addr.getPointer(), Addr.getElementType(),
+              V->getType().isVolatileQualified(),
+              V->getType()->hasSignedIntegerRepresentation()};
+  }
+  llvm::OpenMPIRBuilder::AtomicOpValue ROpVal;
+  if (R) {
+    LValue LV = CGF.EmitLValue(R);
+    Address Addr = LV.getAddress(CGF);
+    ROpVal = {Addr.getPointer(), Addr.getElementType(),
+              R->getType().isVolatileQualified(),
+              R->getType()->hasSignedIntegerRepresentation()};
+  }
 
   CGF.Builder.restoreIP(OMPBuilder.createAtomicCompare(
-      CGF.Builder, XOpVal, EVal, DVal, AO, Op, IsXBinopExpr));
+      CGF.Builder, XOpVal, VOpVal, ROpVal, EVal, DVal, AO, Op, IsXBinopExpr,
+      IsPostfixUpdate, IsFailOnly));
 }
 
 static void emitOMPAtomicExpr(CodeGenFunction &CGF, OpenMPClauseKind Kind,
                               llvm::AtomicOrdering AO, bool IsPostfixUpdate,
-                              const Expr *X, const Expr *V, const Expr *E,
-                              const Expr *UE, const Expr *D, const Expr *CE,
-                              bool IsXLHSInRHSPart, bool IsCompareCapture,
-                              SourceLocation Loc) {
+                              const Expr *X, const Expr *V, const Expr *R,
+                              const Expr *E, const Expr *UE, const Expr *D,
+                              const Expr *CE, bool IsXLHSInRHSPart,
+                              bool IsFailOnly, SourceLocation Loc) {
   switch (Kind) {
   case OMPC_read:
     emitOMPAtomicReadExpr(CGF, AO, X, V, Loc);
     break;
   case OMPC_write:
     emitOMPAtomicWriteExpr(CGF, AO, X, E, Loc);
     break;
   case OMPC_unknown:
   case OMPC_update:
     emitOMPAtomicUpdateExpr(CGF, AO, X, E, UE, IsXLHSInRHSPart, Loc);
     break;
   case OMPC_capture:
     emitOMPAtomicCaptureExpr(CGF, AO, IsPostfixUpdate, V, X, E, UE,
                              IsXLHSInRHSPart, Loc);
     break;
   case OMPC_compare: {
-    if (IsCompareCapture) {
-      // Emit an error here.
-      unsigned DiagID = CGF.CGM.getDiags().getCustomDiagID(
-          DiagnosticsEngine::Error,
-          "'atomic compare capture' is not supported for now");
-      CGF.CGM.getDiags().Report(DiagID);
-    } else {
-      emitOMPAtomicCompareExpr(CGF, AO, X, E, D, CE, IsXLHSInRHSPart, Loc);
-    }
+    emitOMPAtomicCompareExpr(CGF, AO, X, V, R, E, D, CE, IsXLHSInRHSPart,
+                             IsPostfixUpdate, IsFailOnly, Loc);
     break;
   }
   case OMPC_if:
   case OMPC_final:
   case OMPC_num_threads:
   case OMPC_private:
   case OMPC_firstprivate:
   case OMPC_lastprivate:
for (const OMPClause *C : S.clauses()) {
     // if it is first).
     OpenMPClauseKind K = C->getClauseKind();
     if (K == OMPC_seq_cst || K == OMPC_acq_rel || K == OMPC_acquire ||
         K == OMPC_release || K == OMPC_relaxed || K == OMPC_hint)
       continue;
     Kind = K;
     KindsEncountered.insert(K);
   }
-  bool IsCompareCapture = false;
+  // We just need to correct Kind here. No need to set a bool saying it is
+  // actually compare capture because we can tell from whether V and R are
+  // nullptr.
   if (KindsEncountered.contains(OMPC_compare) &&
-      KindsEncountered.contains(OMPC_capture)) {
-    IsCompareCapture = true;
+      KindsEncountered.contains(OMPC_capture))
     Kind = OMPC_compare;
-  }
   if (!MemOrderingSpecified) {
     llvm::AtomicOrdering DefaultOrder =
         CGM.getOpenMPRuntime().getDefaultMemoryOrdering();
     if (DefaultOrder == llvm::AtomicOrdering::Monotonic ||
         DefaultOrder == llvm::AtomicOrdering::SequentiallyConsistent ||
         (DefaultOrder == llvm::AtomicOrdering::AcquireRelease &&
          Kind == OMPC_capture)) {
       AO = DefaultOrder;
     } else if (DefaultOrder == llvm::AtomicOrdering::AcquireRelease) {
       if (Kind == OMPC_unknown || Kind == OMPC_update || Kind == OMPC_write) {
         AO = llvm::AtomicOrdering::Release;
       } else if (Kind == OMPC_read) {
         assert(Kind == OMPC_read && "Unexpected atomic kind.");
         AO = llvm::AtomicOrdering::Acquire;
       }
     }
   }
 
   LexicalScope Scope(*this, S.getSourceRange());
   EmitStopPoint(S.getAssociatedStmt());
   emitOMPAtomicExpr(*this, Kind, AO, S.isPostfixUpdate(), S.getX(), S.getV(),
-                    S.getExpr(), S.getUpdateExpr(), S.getD(), S.getCondExpr(),
-                    S.isXLHSInRHSPart(), IsCompareCapture, S.getBeginLoc());
+                    S.getR(), S.getExpr(), S.getUpdateExpr(), S.getD(),
+                    S.getCondExpr(), S.isXLHSInRHSPart(), S.isFailOnly(),
+                    S.getBeginLoc());
 }
 
 static void emitCommonOMPTargetDirective(CodeGenFunction &CGF,
                                          const OMPExecutableDirective &S,
                                          const RegionCodeGenTy &CodeGen) {
   assert(isOpenMPTargetExecutionDirective(S.getDirectiveKind()));
   CodeGenModule &CGM = CGF.CGM;
 

clang/lib/Sema/SemaOpenMP.cpp

 class OpenMPAtomicCompareCaptureChecker final
     : public OpenMPAtomicCompareChecker {
 public:
   OpenMPAtomicCompareCaptureChecker(Sema &S) : OpenMPAtomicCompareChecker(S) {}
 
   Expr *getV() const { return V; }
   Expr *getR() const { return R; }
   bool isFailOnly() const { return IsFailOnly; }
+  bool isPostfixUpdate() const { return IsPostfixUpdate; }
 
   /// Check if statement \a S is valid for <tt>atomic compare capture</tt>.
   bool checkStmt(Stmt *S, ErrorInfoTy &ErrorInfo);
 
 private:
   bool checkType(ErrorInfoTy &ErrorInfo);
 
   // NOTE: Form 3, 4, 5 in the following comments mean the 3rd, 4th, and 5th
private:
   bool checkForm45(Stmt *S, ErrorInfoTy &ErrorInfo);
 
   /// 'v' lvalue part of the source atomic expression.
   Expr *V = nullptr;
   /// 'r' lvalue part of the source atomic expression.
   Expr *R = nullptr;
   /// If 'v' is only updated when the comparison fails.
   bool IsFailOnly = false;
+  /// If original value of 'x' must be stored in 'v', not an updated one.
+  bool IsPostfixUpdate = false;
 };
 
 bool OpenMPAtomicCompareCaptureChecker::checkType(ErrorInfoTy &ErrorInfo) {
   if (!OpenMPAtomicCompareChecker::checkType(ErrorInfo))
     return false;
 
   if (V && !CheckValue(V, ErrorInfo, true))
     return false;
if (BO->getOpcode() != BO_Assign) {
     ErrorInfo.Error = ErrorTy::NotAnAssignment;
     ErrorInfo.ErrorLoc = BO->getExprLoc();
     ErrorInfo.NoteLoc = BO->getOperatorLoc();
     ErrorInfo.ErrorRange = ErrorInfo.NoteRange = BO->getSourceRange();
     return false;
   }
 
   X = BO->getLHS();
-  D = BO->getRHS();
+  D = BO->getRHS()->IgnoreImpCasts();
 
   auto *Cond = dyn_cast<BinaryOperator>(S->getCond());
   if (!Cond) {
     ErrorInfo.Error = ErrorTy::NotABinaryOp;
     ErrorInfo.ErrorLoc = ErrorInfo.NoteLoc = S->getCond()->getExprLoc();
     ErrorInfo.ErrorRange = ErrorInfo.NoteRange = S->getCond()->getSourceRange();
     return false;
   }
   if (Cond->getOpcode() != BO_EQ) {
     ErrorInfo.Error = ErrorTy::NotEQ;
     ErrorInfo.ErrorLoc = ErrorInfo.NoteLoc = Cond->getExprLoc();
     ErrorInfo.ErrorRange = ErrorInfo.NoteRange = Cond->getSourceRange();
     return false;
   }
 
   if (checkIfTwoExprsAreSame(ContextRef, X, Cond->getLHS())) {
-    E = Cond->getRHS();
+    E = Cond->getRHS()->IgnoreImpCasts();
   } else if (checkIfTwoExprsAreSame(ContextRef, X, Cond->getRHS())) {
-    E = Cond->getLHS();
+    E = Cond->getLHS()->IgnoreImpCasts();
   } else {
     ErrorInfo.Error = ErrorTy::InvalidComparison;
     ErrorInfo.ErrorLoc = ErrorInfo.NoteLoc = Cond->getExprLoc();
     ErrorInfo.ErrorRange = ErrorInfo.NoteRange = Cond->getSourceRange();
     return false;
   }
 
   C = Cond;
if (ThenBO->getOpcode() != BO_Assign) {
     ErrorInfo.Error = ErrorTy::NotAnAssignment;
     ErrorInfo.ErrorLoc = ThenBO->getExprLoc();
     ErrorInfo.NoteLoc = ThenBO->getOperatorLoc();
     ErrorInfo.ErrorRange = ErrorInfo.NoteRange = ThenBO->getSourceRange();
     return false;
   }
 
   X = ThenBO->getLHS();
-  D = ThenBO->getRHS();
+  D = ThenBO->getRHS()->IgnoreImpCasts();
 
   auto *BO = cast<BinaryOperator>(S1->getRHS()->IgnoreImpCasts());
   if (BO->getOpcode() != BO_EQ) {
     ErrorInfo.Error = ErrorTy::NotEQ;
     ErrorInfo.ErrorLoc = BO->getExprLoc();
     ErrorInfo.NoteLoc = BO->getOperatorLoc();
     ErrorInfo.ErrorRange = ErrorInfo.NoteRange = BO->getSourceRange();
     return false;
   }
 
   C = BO;
 
   if (checkIfTwoExprsAreSame(ContextRef, X, BO->getLHS())) {
-    E = BO->getRHS();
+    E = BO->getRHS()->IgnoreImpCasts();
   } else if (checkIfTwoExprsAreSame(ContextRef, X, BO->getRHS())) {
-    E = BO->getLHS();
+    E = BO->getLHS()->IgnoreImpCasts();
   } else {
     ErrorInfo.Error = ErrorTy::InvalidComparison;
     ErrorInfo.ErrorLoc = ErrorInfo.NoteLoc = BO->getExprLoc();
     ErrorInfo.ErrorRange = ErrorInfo.NoteRange = BO->getSourceRange();
     return false;
   }
 
   if (S2->getElse()) {
if (CS->size() == 1) {
     if (auto *BO = dyn_cast<BinaryOperator>(S1)) {
       // { v = x; cond-update-stmt } or form 45.
       UpdateStmt = S1;
       CondUpdateStmt = S2;
       // Check if form 45.
       if (dyn_cast<BinaryOperator>(BO->getRHS()->IgnoreImpCasts()) &&
           dyn_cast<IfStmt>(S2))
         return checkForm45(CS, ErrorInfo);
+      // It cannot be set before we the check for form45.
+      IsPostfixUpdate = true;
     } else {
       // { cond-update-stmt v = x; }
       UpdateStmt = S2;
       CondUpdateStmt = S1;
     }
 
     auto CheckCondUpdateStmt = [this, &ErrorInfo](Stmt *CUS) {
       auto *IS = dyn_cast<IfStmt>(CUS);
if (auto *EWC = dyn_cast<ExprWithCleanups>(Body))
     Body = EWC->getSubExpr();
 
   Expr *X = nullptr;
   Expr *V = nullptr;
   Expr *E = nullptr;
   Expr *UE = nullptr;
   Expr *D = nullptr;
   Expr *CE = nullptr;
+  Expr *R = nullptr;
   bool IsXLHSInRHSPart = false;
   bool IsPostfixUpdate = false;
+  bool IsFailOnly = false;
   // OpenMP [2.12.6, atomic Construct]
   // In the next expressions:
   // * x and v (as applicable) are both l-value expressions with scalar type.
   // * During the execution of an atomic region, multiple syntactic
   // occurrences of x must designate the same storage location.
   // * Neither of v and expr (as applicable) may access the storage location
   // designated by x.
   // * Neither of x and expr (as applicable) may access the storage location
if (IsCompareCapture) {
       OpenMPAtomicCompareCaptureChecker Checker(*this);
       if (!Checker.checkStmt(Body, ErrorInfo)) {
         Diag(ErrorInfo.ErrorLoc, diag::err_omp_atomic_compare_capture)
             << ErrorInfo.ErrorRange;
         Diag(ErrorInfo.NoteLoc, diag::note_omp_atomic_compare)
             << ErrorInfo.Error << ErrorInfo.NoteRange;
         return StmtError();
       }
-      // TODO: We don't set X, D, E, etc. here because in code gen we will emit
-      // error directly.
+      X = Checker.getX();
+      E = Checker.getE();
+      D = Checker.getD();
+      CE = Checker.getCond();
+      V = Checker.getV();
+      R = Checker.getR();
+      // We reuse IsXLHSInRHSPart to tell if it is in the form 'x ordop expr'.
+      IsXLHSInRHSPart = Checker.isXBinopExpr();
+      IsFailOnly = Checker.isFailOnly();
+      IsPostfixUpdate = Checker.isPostfixUpdate();
     } else {
       OpenMPAtomicCompareChecker::ErrorInfoTy ErrorInfo;
       OpenMPAtomicCompareChecker Checker(*this);
       if (!Checker.checkStmt(Body, ErrorInfo)) {
         Diag(ErrorInfo.ErrorLoc, diag::err_omp_atomic_compare)
             << ErrorInfo.ErrorRange;
         Diag(ErrorInfo.NoteLoc, diag::note_omp_atomic_compare)
-            << ErrorInfo.Error << ErrorInfo.NoteRange;
         return StmtError();
       }
       X = Checker.getX();
       E = Checker.getE();
       D = Checker.getD();
       CE = Checker.getCond();
       // We reuse IsXLHSInRHSPart to tell if it is in the form 'x ordop expr'.
       IsXLHSInRHSPart = Checker.isXBinopExpr();
     }
   }
 
   setFunctionHasBranchProtectedScope();
 
   return OMPAtomicDirective::Create(Context, StartLoc, EndLoc, Clauses, AStmt,
-                                    X, V, E, UE, D, CE, IsXLHSInRHSPart,
-                                    IsPostfixUpdate);
+                                    X, V, R, E, UE, D, CE, IsXLHSInRHSPart,
+                                    IsPostfixUpdate, IsFailOnly);
 }
 
 StmtResult Sema::ActOnOpenMPTargetDirective(ArrayRef<OMPClause *> Clauses,
                                             Stmt *AStmt,
                                             SourceLocation StartLoc,
                                             SourceLocation EndLoc) {
   if (!AStmt)
     return StmtError();

clang/lib/Serialization/ASTReaderStmt.cpp

void ASTStmtReader::VisitOMPOrderedDirective(OMPOrderedDirective *D) {
   VisitOMPExecutableDirective(D);
 }
 
 void ASTStmtReader::VisitOMPAtomicDirective(OMPAtomicDirective *D) {
   VisitStmt(D);
   VisitOMPExecutableDirective(D);
   D->IsXLHSInRHSPart = Record.readBool();
   D->IsPostfixUpdate = Record.readBool();
+  D->IsFailOnly = Record.readBool();
 }
 
 void ASTStmtReader::VisitOMPTargetDirective(OMPTargetDirective *D) {
   VisitStmt(D);
   VisitOMPExecutableDirective(D);
 }
 
 void ASTStmtReader::VisitOMPTargetDataDirective(OMPTargetDataDirective *D) {

clang/lib/Serialization/ASTWriterStmt.cpp

void ASTStmtWriter::VisitOMPTaskDirective(OMPTaskDirective *D) {
   Code = serialization::STMT_OMP_TASK_DIRECTIVE;
 }
 
 void ASTStmtWriter::VisitOMPAtomicDirective(OMPAtomicDirective *D) {
   VisitStmt(D);
   VisitOMPExecutableDirective(D);
   Record.writeBool(D->isXLHSInRHSPart());
   Record.writeBool(D->isPostfixUpdate());
+  Record.writeBool(D->isFailOnly());
   Code = serialization::STMT_OMP_ATOMIC_DIRECTIVE;
 }
 
 void ASTStmtWriter::VisitOMPTargetDirective(OMPTargetDirective *D) {
   VisitStmt(D);
   VisitOMPExecutableDirective(D);
   Code = serialization::STMT_OMP_TARGET_DIRECTIVE;
 }

clang/test/OpenMP/atomic_compare_codegen.cpp

 #pragma omp atomic compare seq_cst
   if (ullx == ulle)
     ullx = ulld;
 #pragma omp atomic compare seq_cst
   if (ulle == ullx)
     ullx = ulld;
 }
 
+void bar() {
+  char cx, cv, cr, ce, cd;
+  unsigned char ucx, ucv, ucr, uce, ucd;
+  short sx, sv, sr, se, sd;
+  unsigned short usx, usv, usr, use, usd;
+  int ix, iv, ir, ie, id;
+  unsigned int uix, uiv, uir, uie, uid;
+  long lx, lv, lr, le, ld;
+  unsigned long ulx, ulv, ulr, ule, uld;
+  long long llx, llv, llr, lle, lld;
+  unsigned long long ullx, ullv, ullr, ulle, ulld;
+
+#pragma omp atomic compare capture
+  {
+    cv = cx;
+    if (ce > cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cv = cx;
+    if (cx > ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cv = cx;
+    if (ce < cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cv = cx;
+    if (cx < ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cv = cx;
+    if (cx == ce) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cv = cx;
+    if (ce == cx) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (ce > cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (cx > ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ce < cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (cx < ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (cx == ce) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ce == cx) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  if (cx == ce) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  if (ce == cx) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    cv = cx;
+    if (ce > cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cv = cx;
+    if (cx > ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cv = cx;
+    if (ce < cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cv = cx;
+    if (cx < ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cv = cx;
+    if (cx == ce) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cv = cx;
+    if (ce == cx) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ce > cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (cx > ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ce < cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (cx < ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (cx == ce) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ce == cx) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (cx == ce) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ce == cx) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    cv = cx;
+    if (ce > cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cv = cx;
+    if (cx > ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cv = cx;
+    if (ce < cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cv = cx;
+    if (cx < ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cv = cx;
+    if (cx == ce) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cv = cx;
+    if (ce == cx) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ce > cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (cx > ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ce < cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (cx < ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (cx == ce) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ce == cx) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  if (cx == ce) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  if (ce == cx) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    cv = cx;
+    if (ce > cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cv = cx;
+    if (cx > ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cv = cx;
+    if (ce < cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cv = cx;
+    if (cx < ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cv = cx;
+    if (cx == ce) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cv = cx;
+    if (ce == cx) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ce > cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (cx > ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ce < cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (cx < ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (cx == ce) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ce == cx) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (cx == ce) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ce == cx) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    cv = cx;
+    if (ce > cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cv = cx;
+    if (cx > ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cv = cx;
+    if (ce < cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cv = cx;
+    if (cx < ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cv = cx;
+    if (cx == ce) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cv = cx;
+    if (ce == cx) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ce > cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (cx > ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ce < cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (cx < ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (cx == ce) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ce == cx) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  if (cx == ce) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  if (ce == cx) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture release
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    cv = cx;
+    if (ce > cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cv = cx;
+    if (cx > ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cv = cx;
+    if (ce < cx) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cv = cx;
+    if (cx < ce) {
+      cx = ce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cv = cx;
+    if (cx == ce) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cv = cx;
+    if (ce == cx) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ce > cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (cx > ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ce < cx) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (cx < ce) {
+      cx = ce;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (cx == ce) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ce == cx) {
+      cx = cd;
+    }
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (cx == ce) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ce == cx) {
+    cx = cd;
+  } else {
+    cv = cx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cr = cx == ce;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    cr = ce == cx;
+    if (cr) {
+      cx = cd;
+    } else {
+      cv = cx;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    ucv = ucx;
+    if (uce > ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucv = ucx;
+    if (ucx > uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucv = ucx;
+    if (uce < ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucv = ucx;
+    if (ucx < uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucv = ucx;
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucv = ucx;
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (uce > ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ucx > uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (uce < ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ucx < uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  if (ucx == uce) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  if (uce == ucx) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    ucv = ucx;
+    if (uce > ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucv = ucx;
+    if (ucx > uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucv = ucx;
+    if (uce < ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucv = ucx;
+    if (ucx < uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucv = ucx;
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucv = ucx;
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uce > ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ucx > uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uce < ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ucx < uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ucx == uce) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (uce == ucx) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    ucv = ucx;
+    if (uce > ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucv = ucx;
+    if (ucx > uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucv = ucx;
+    if (uce < ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucv = ucx;
+    if (ucx < uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucv = ucx;
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucv = ucx;
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uce > ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ucx > uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uce < ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ucx < uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  if (ucx == uce) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  if (uce == ucx) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    ucv = ucx;
+    if (uce > ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucv = ucx;
+    if (ucx > uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucv = ucx;
+    if (uce < ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucv = ucx;
+    if (ucx < uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucv = ucx;
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucv = ucx;
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uce > ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ucx > uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uce < ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ucx < uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ucx == uce) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (uce == ucx) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    ucv = ucx;
+    if (uce > ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucv = ucx;
+    if (ucx > uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucv = ucx;
+    if (uce < ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucv = ucx;
+    if (ucx < uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucv = ucx;
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucv = ucx;
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uce > ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ucx > uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uce < ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ucx < uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  if (ucx == uce) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  if (uce == ucx) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture release
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    ucv = ucx;
+    if (uce > ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucv = ucx;
+    if (ucx > uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucv = ucx;
+    if (uce < ucx) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucv = ucx;
+    if (ucx < uce) {
+      ucx = uce;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucv = ucx;
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucv = ucx;
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uce > ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ucx > uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uce < ucx) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ucx < uce) {
+      ucx = uce;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ucx == uce) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uce == ucx) {
+      ucx = ucd;
+    }
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ucx == uce) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (uce == ucx) {
+    ucx = ucd;
+  } else {
+    ucv = ucx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucr = ucx == uce;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ucr = uce == ucx;
+    if (ucr) {
+      ucx = ucd;
+    } else {
+      ucv = ucx;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    sv = sx;
+    if (se > sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sv = sx;
+    if (sx > se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sv = sx;
+    if (se < sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sv = sx;
+    if (sx < se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sv = sx;
+    if (sx == se) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sv = sx;
+    if (se == sx) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (se > sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (sx > se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (se < sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (sx < se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (sx == se) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (se == sx) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  if (sx == se) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  if (se == sx) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    sv = sx;
+    if (se > sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sv = sx;
+    if (sx > se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sv = sx;
+    if (se < sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sv = sx;
+    if (sx < se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sv = sx;
+    if (sx == se) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sv = sx;
+    if (se == sx) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (se > sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (sx > se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (se < sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (sx < se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (sx == se) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (se == sx) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (sx == se) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (se == sx) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    sv = sx;
+    if (se > sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sv = sx;
+    if (sx > se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sv = sx;
+    if (se < sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sv = sx;
+    if (sx < se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sv = sx;
+    if (sx == se) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sv = sx;
+    if (se == sx) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (se > sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (sx > se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (se < sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (sx < se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (sx == se) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (se == sx) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  if (sx == se) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  if (se == sx) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    sv = sx;
+    if (se > sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sv = sx;
+    if (sx > se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sv = sx;
+    if (se < sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sv = sx;
+    if (sx < se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sv = sx;
+    if (sx == se) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sv = sx;
+    if (se == sx) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (se > sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (sx > se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (se < sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (sx < se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (sx == se) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (se == sx) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (sx == se) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (se == sx) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    sv = sx;
+    if (se > sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sv = sx;
+    if (sx > se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sv = sx;
+    if (se < sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sv = sx;
+    if (sx < se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sv = sx;
+    if (sx == se) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sv = sx;
+    if (se == sx) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (se > sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (sx > se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (se < sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (sx < se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (sx == se) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (se == sx) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  if (sx == se) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  if (se == sx) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture release
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    sv = sx;
+    if (se > sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sv = sx;
+    if (sx > se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sv = sx;
+    if (se < sx) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sv = sx;
+    if (sx < se) {
+      sx = se;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sv = sx;
+    if (sx == se) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sv = sx;
+    if (se == sx) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (se > sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (sx > se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (se < sx) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (sx < se) {
+      sx = se;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (sx == se) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (se == sx) {
+      sx = sd;
+    }
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (sx == se) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (se == sx) {
+    sx = sd;
+  } else {
+    sv = sx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sr = sx == se;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    sr = se == sx;
+    if (sr) {
+      sx = sd;
+    } else {
+      sv = sx;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    usv = usx;
+    if (use > usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usv = usx;
+    if (usx > use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usv = usx;
+    if (use < usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usv = usx;
+    if (usx < use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usv = usx;
+    if (usx == use) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usv = usx;
+    if (use == usx) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (use > usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (usx > use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (use < usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (usx < use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (usx == use) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (use == usx) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  if (usx == use) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  if (use == usx) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    usv = usx;
+    if (use > usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usv = usx;
+    if (usx > use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usv = usx;
+    if (use < usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usv = usx;
+    if (usx < use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usv = usx;
+    if (usx == use) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usv = usx;
+    if (use == usx) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (use > usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (usx > use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (use < usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (usx < use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (usx == use) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (use == usx) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (usx == use) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (use == usx) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    usv = usx;
+    if (use > usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usv = usx;
+    if (usx > use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usv = usx;
+    if (use < usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usv = usx;
+    if (usx < use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usv = usx;
+    if (usx == use) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usv = usx;
+    if (use == usx) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (use > usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (usx > use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (use < usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (usx < use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (usx == use) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (use == usx) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  if (usx == use) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  if (use == usx) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    usv = usx;
+    if (use > usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usv = usx;
+    if (usx > use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usv = usx;
+    if (use < usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usv = usx;
+    if (usx < use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usv = usx;
+    if (usx == use) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usv = usx;
+    if (use == usx) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (use > usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (usx > use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (use < usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (usx < use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (usx == use) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (use == usx) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (usx == use) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (use == usx) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    usv = usx;
+    if (use > usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usv = usx;
+    if (usx > use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usv = usx;
+    if (use < usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usv = usx;
+    if (usx < use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usv = usx;
+    if (usx == use) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usv = usx;
+    if (use == usx) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (use > usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (usx > use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (use < usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (usx < use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (usx == use) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (use == usx) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  if (usx == use) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  if (use == usx) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture release
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    usv = usx;
+    if (use > usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usv = usx;
+    if (usx > use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usv = usx;
+    if (use < usx) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usv = usx;
+    if (usx < use) {
+      usx = use;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usv = usx;
+    if (usx == use) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usv = usx;
+    if (use == usx) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (use > usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (usx > use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (use < usx) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (usx < use) {
+      usx = use;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (usx == use) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (use == usx) {
+      usx = usd;
+    }
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (usx == use) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (use == usx) {
+    usx = usd;
+  } else {
+    usv = usx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usr = usx == use;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    usr = use == usx;
+    if (usr) {
+      usx = usd;
+    } else {
+      usv = usx;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    iv = ix;
+    if (ie > ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    iv = ix;
+    if (ix > ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    iv = ix;
+    if (ie < ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    iv = ix;
+    if (ix < ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    iv = ix;
+    if (ix == ie) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    iv = ix;
+    if (ie == ix) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (ie > ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ix > ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ie < ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ix < ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ix == ie) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ie == ix) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  if (ix == ie) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  if (ie == ix) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    iv = ix;
+    if (ie > ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    iv = ix;
+    if (ix > ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    iv = ix;
+    if (ie < ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    iv = ix;
+    if (ix < ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    iv = ix;
+    if (ix == ie) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    iv = ix;
+    if (ie == ix) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ie > ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ix > ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ie < ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ix < ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ix == ie) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ie == ix) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ix == ie) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ie == ix) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    iv = ix;
+    if (ie > ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    iv = ix;
+    if (ix > ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    iv = ix;
+    if (ie < ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    iv = ix;
+    if (ix < ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    iv = ix;
+    if (ix == ie) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    iv = ix;
+    if (ie == ix) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ie > ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ix > ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ie < ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ix < ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ix == ie) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ie == ix) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  if (ix == ie) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  if (ie == ix) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    iv = ix;
+    if (ie > ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    iv = ix;
+    if (ix > ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    iv = ix;
+    if (ie < ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    iv = ix;
+    if (ix < ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    iv = ix;
+    if (ix == ie) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    iv = ix;
+    if (ie == ix) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ie > ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ix > ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ie < ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ix < ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ix == ie) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ie == ix) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ix == ie) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ie == ix) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    iv = ix;
+    if (ie > ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    iv = ix;
+    if (ix > ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    iv = ix;
+    if (ie < ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    iv = ix;
+    if (ix < ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    iv = ix;
+    if (ix == ie) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    iv = ix;
+    if (ie == ix) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ie > ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ix > ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ie < ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ix < ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ix == ie) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ie == ix) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  if (ix == ie) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  if (ie == ix) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture release
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    iv = ix;
+    if (ie > ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    iv = ix;
+    if (ix > ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    iv = ix;
+    if (ie < ix) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    iv = ix;
+    if (ix < ie) {
+      ix = ie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    iv = ix;
+    if (ix == ie) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    iv = ix;
+    if (ie == ix) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ie > ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ix > ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ie < ix) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ix < ie) {
+      ix = ie;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ix == ie) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ie == ix) {
+      ix = id;
+    }
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ix == ie) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ie == ix) {
+    ix = id;
+  } else {
+    iv = ix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ir = ix == ie;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ir = ie == ix;
+    if (ir) {
+      ix = id;
+    } else {
+      iv = ix;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    uiv = uix;
+    if (uie > uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uiv = uix;
+    if (uix > uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uiv = uix;
+    if (uie < uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uiv = uix;
+    if (uix < uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uiv = uix;
+    if (uix == uie) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uiv = uix;
+    if (uie == uix) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (uie > uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (uix > uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (uie < uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (uix < uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (uix == uie) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  {
+    if (uie == uix) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  if (uix == uie) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  if (uie == uix) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    uiv = uix;
+    if (uie > uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uiv = uix;
+    if (uix > uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uiv = uix;
+    if (uie < uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uiv = uix;
+    if (uix < uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uiv = uix;
+    if (uix == uie) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uiv = uix;
+    if (uie == uix) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uie > uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uix > uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uie < uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uix < uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uix == uie) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (uie == uix) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (uix == uie) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (uie == uix) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    uiv = uix;
+    if (uie > uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uiv = uix;
+    if (uix > uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uiv = uix;
+    if (uie < uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uiv = uix;
+    if (uix < uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uiv = uix;
+    if (uix == uie) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uiv = uix;
+    if (uie == uix) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uie > uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uix > uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uie < uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uix < uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uix == uie) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (uie == uix) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  if (uix == uie) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  if (uie == uix) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    uiv = uix;
+    if (uie > uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uiv = uix;
+    if (uix > uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uiv = uix;
+    if (uie < uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uiv = uix;
+    if (uix < uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uiv = uix;
+    if (uix == uie) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uiv = uix;
+    if (uie == uix) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uie > uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uix > uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uie < uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uix < uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uix == uie) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (uie == uix) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  if (uix == uie) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  if (uie == uix) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    uiv = uix;
+    if (uie > uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uiv = uix;
+    if (uix > uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uiv = uix;
+    if (uie < uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uiv = uix;
+    if (uix < uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uiv = uix;
+    if (uix == uie) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uiv = uix;
+    if (uie == uix) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uie > uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uix > uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uie < uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uix < uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uix == uie) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (uie == uix) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  if (uix == uie) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  if (uie == uix) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture release
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    uiv = uix;
+    if (uie > uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uiv = uix;
+    if (uix > uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uiv = uix;
+    if (uie < uix) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uiv = uix;
+    if (uix < uie) {
+      uix = uie;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uiv = uix;
+    if (uix == uie) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uiv = uix;
+    if (uie == uix) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uie > uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uix > uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uie < uix) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uix < uie) {
+      uix = uie;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uix == uie) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (uie == uix) {
+      uix = uid;
+    }
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (uix == uie) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (uie == uix) {
+    uix = uid;
+  } else {
+    uiv = uix;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uir = uix == uie;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    uir = uie == uix;
+    if (uir) {
+      uix = uid;
+    } else {
+      uiv = uix;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    lv = lx;
+    if (le > lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lv = lx;
+    if (lx > le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lv = lx;
+    if (le < lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lv = lx;
+    if (lx < le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lv = lx;
+    if (lx == le) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lv = lx;
+    if (le == lx) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (le > lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (lx > le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (le < lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (lx < le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (lx == le) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (le == lx) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  if (lx == le) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  if (le == lx) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    lv = lx;
+    if (le > lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lv = lx;
+    if (lx > le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lv = lx;
+    if (le < lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lv = lx;
+    if (lx < le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lv = lx;
+    if (lx == le) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lv = lx;
+    if (le == lx) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (le > lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (lx > le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (le < lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (lx < le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (lx == le) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (le == lx) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (lx == le) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (le == lx) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    lv = lx;
+    if (le > lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lv = lx;
+    if (lx > le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lv = lx;
+    if (le < lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lv = lx;
+    if (lx < le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lv = lx;
+    if (lx == le) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lv = lx;
+    if (le == lx) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (le > lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (lx > le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (le < lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (lx < le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (lx == le) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (le == lx) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  if (lx == le) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  if (le == lx) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    lv = lx;
+    if (le > lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lv = lx;
+    if (lx > le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lv = lx;
+    if (le < lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lv = lx;
+    if (lx < le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lv = lx;
+    if (lx == le) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lv = lx;
+    if (le == lx) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (le > lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (lx > le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (le < lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (lx < le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (lx == le) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (le == lx) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (lx == le) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (le == lx) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    lv = lx;
+    if (le > lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lv = lx;
+    if (lx > le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lv = lx;
+    if (le < lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lv = lx;
+    if (lx < le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lv = lx;
+    if (lx == le) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lv = lx;
+    if (le == lx) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (le > lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (lx > le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (le < lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (lx < le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (lx == le) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (le == lx) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  if (lx == le) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  if (le == lx) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture release
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    lv = lx;
+    if (le > lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lv = lx;
+    if (lx > le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lv = lx;
+    if (le < lx) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lv = lx;
+    if (lx < le) {
+      lx = le;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lv = lx;
+    if (lx == le) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lv = lx;
+    if (le == lx) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (le > lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (lx > le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (le < lx) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (lx < le) {
+      lx = le;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (lx == le) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (le == lx) {
+      lx = ld;
+    }
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (lx == le) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (le == lx) {
+    lx = ld;
+  } else {
+    lv = lx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lr = lx == le;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    lr = le == lx;
+    if (lr) {
+      lx = ld;
+    } else {
+      lv = lx;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    ulv = ulx;
+    if (ule > ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulv = ulx;
+    if (ulx > ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulv = ulx;
+    if (ule < ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulv = ulx;
+    if (ulx < ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulv = ulx;
+    if (ulx == ule) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulv = ulx;
+    if (ule == ulx) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (ule > ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ulx > ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ule < ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ulx < ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ulx == ule) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ule == ulx) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  if (ulx == ule) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  if (ule == ulx) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    ulv = ulx;
+    if (ule > ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulv = ulx;
+    if (ulx > ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulv = ulx;
+    if (ule < ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulv = ulx;
+    if (ulx < ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulv = ulx;
+    if (ulx == ule) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulv = ulx;
+    if (ule == ulx) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ule > ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ulx > ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ule < ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ulx < ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ulx == ule) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ule == ulx) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ulx == ule) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ule == ulx) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    ulv = ulx;
+    if (ule > ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulv = ulx;
+    if (ulx > ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulv = ulx;
+    if (ule < ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulv = ulx;
+    if (ulx < ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulv = ulx;
+    if (ulx == ule) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulv = ulx;
+    if (ule == ulx) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ule > ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ulx > ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ule < ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ulx < ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ulx == ule) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ule == ulx) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  if (ulx == ule) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  if (ule == ulx) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    ulv = ulx;
+    if (ule > ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulv = ulx;
+    if (ulx > ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulv = ulx;
+    if (ule < ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulv = ulx;
+    if (ulx < ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulv = ulx;
+    if (ulx == ule) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulv = ulx;
+    if (ule == ulx) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ule > ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ulx > ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ule < ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ulx < ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ulx == ule) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ule == ulx) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ulx == ule) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ule == ulx) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    ulv = ulx;
+    if (ule > ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulv = ulx;
+    if (ulx > ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulv = ulx;
+    if (ule < ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulv = ulx;
+    if (ulx < ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulv = ulx;
+    if (ulx == ule) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulv = ulx;
+    if (ule == ulx) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ule > ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ulx > ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ule < ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ulx < ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ulx == ule) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ule == ulx) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  if (ulx == ule) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  if (ule == ulx) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture release
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    ulv = ulx;
+    if (ule > ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulv = ulx;
+    if (ulx > ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulv = ulx;
+    if (ule < ulx) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulv = ulx;
+    if (ulx < ule) {
+      ulx = ule;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulv = ulx;
+    if (ulx == ule) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulv = ulx;
+    if (ule == ulx) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ule > ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ulx > ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ule < ulx) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ulx < ule) {
+      ulx = ule;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ulx == ule) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ule == ulx) {
+      ulx = uld;
+    }
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ulx == ule) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ule == ulx) {
+    ulx = uld;
+  } else {
+    ulv = ulx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulr = ulx == ule;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ulr = ule == ulx;
+    if (ulr) {
+      ulx = uld;
+    } else {
+      ulv = ulx;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    llv = llx;
+    if (lle > llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llv = llx;
+    if (llx > lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llv = llx;
+    if (lle < llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llv = llx;
+    if (llx < lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llv = llx;
+    if (llx == lle) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llv = llx;
+    if (lle == llx) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (lle > llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (llx > lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (lle < llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (llx < lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (llx == lle) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (lle == llx) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  if (llx == lle) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  if (lle == llx) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    llv = llx;
+    if (lle > llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llv = llx;
+    if (llx > lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llv = llx;
+    if (lle < llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llv = llx;
+    if (llx < lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llv = llx;
+    if (llx == lle) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llv = llx;
+    if (lle == llx) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (lle > llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (llx > lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (lle < llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (llx < lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (llx == lle) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (lle == llx) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (llx == lle) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (lle == llx) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    llv = llx;
+    if (lle > llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llv = llx;
+    if (llx > lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llv = llx;
+    if (lle < llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llv = llx;
+    if (llx < lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llv = llx;
+    if (llx == lle) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llv = llx;
+    if (lle == llx) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (lle > llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (llx > lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (lle < llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (llx < lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (llx == lle) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (lle == llx) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  if (llx == lle) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  if (lle == llx) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    llv = llx;
+    if (lle > llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llv = llx;
+    if (llx > lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llv = llx;
+    if (lle < llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llv = llx;
+    if (llx < lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llv = llx;
+    if (llx == lle) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llv = llx;
+    if (lle == llx) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (lle > llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (llx > lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (lle < llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (llx < lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (llx == lle) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (lle == llx) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (llx == lle) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (lle == llx) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    llv = llx;
+    if (lle > llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llv = llx;
+    if (llx > lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llv = llx;
+    if (lle < llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llv = llx;
+    if (llx < lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llv = llx;
+    if (llx == lle) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llv = llx;
+    if (lle == llx) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (lle > llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (llx > lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (lle < llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (llx < lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (llx == lle) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (lle == llx) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  if (llx == lle) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  if (lle == llx) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture release
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    llv = llx;
+    if (lle > llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llv = llx;
+    if (llx > lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llv = llx;
+    if (lle < llx) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llv = llx;
+    if (llx < lle) {
+      llx = lle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llv = llx;
+    if (llx == lle) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llv = llx;
+    if (lle == llx) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (lle > llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (llx > lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (lle < llx) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (llx < lle) {
+      llx = lle;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (llx == lle) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (lle == llx) {
+      llx = lld;
+    }
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (llx == lle) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (lle == llx) {
+    llx = lld;
+  } else {
+    llv = llx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llr = llx == lle;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    llr = lle == llx;
+    if (llr) {
+      llx = lld;
+    } else {
+      llv = llx;
+    }
+  }
+
+#pragma omp atomic compare capture
+  {
+    ullv = ullx;
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullv = ullx;
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullv = ullx;
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullv = ullx;
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullv = ullx;
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullv = ullx;
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  {
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  if (ullx == ulle) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  if (ulle == ullx) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+#pragma omp atomic compare capture
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+
+#pragma omp atomic compare capture acq_rel
+  {
+    ullv = ullx;
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullv = ullx;
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullv = ullx;
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullv = ullx;
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullv = ullx;
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullv = ullx;
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ullx == ulle) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  if (ulle == ullx) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+#pragma omp atomic compare capture acq_rel
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+
+#pragma omp atomic compare capture acquire
+  {
+    ullv = ullx;
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullv = ullx;
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullv = ullx;
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullv = ullx;
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullv = ullx;
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullv = ullx;
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  if (ullx == ulle) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  if (ulle == ullx) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+#pragma omp atomic compare capture acquire
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+
+#pragma omp atomic compare capture relaxed
+  {
+    ullv = ullx;
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullv = ullx;
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullv = ullx;
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullv = ullx;
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullv = ullx;
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullv = ullx;
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ullx == ulle) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  if (ulle == ullx) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+#pragma omp atomic compare capture relaxed
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+
+#pragma omp atomic compare capture release
+  {
+    ullv = ullx;
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullv = ullx;
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullv = ullx;
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullv = ullx;
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullv = ullx;
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullv = ullx;
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  {
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  if (ullx == ulle) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  if (ulle == ullx) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture release
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+#pragma omp atomic compare capture release
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+
+#pragma omp atomic compare capture seq_cst
+  {
+    ullv = ullx;
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullv = ullx;
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullv = ullx;
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullv = ullx;
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullv = ullx;
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullv = ullx;
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ulle > ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ullx > ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ulle < ullx) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ullx < ulle) {
+      ullx = ulle;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ullx == ulle) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    if (ulle == ullx) {
+      ullx = ulld;
+    }
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ullx == ulle) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  if (ulle == ullx) {
+    ullx = ulld;
+  } else {
+    ullv = ullx;
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullr = ullx == ulle;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+#pragma omp atomic compare capture seq_cst
+  {
+    ullr = ulle == ullx;
+    if (ullr) {
+      ullx = ulld;
+    } else {
+      ullv = ullx;
+    }
+  }
+}
+
 #endif
 // CHECK-LABEL: @foo(
 // CHECK-NEXT:  entry:
 // CHECK-NEXT:    [[CX:%.*]] = alloca i8, align 1
 // CHECK-NEXT:    [[CE:%.*]] = alloca i8, align 1
 // CHECK-NEXT:    [[CD:%.*]] = alloca i8, align 1
 // CHECK-NEXT:    [[UCX:%.*]] = alloca i8, align 1
 // CHECK-NEXT:    [[UCE:%.*]] = alloca i8, align 1

 // CHECK-NEXT:    [[TMP1675:%.*]] = load i64, i64* [[ULLD]], align 8
 // CHECK-NEXT:    [[TMP1676:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP1674]], i64 [[TMP1675]] seq_cst seq_cst, align 8
 // CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
 // CHECK-NEXT:    [[TMP1677:%.*]] = load i64, i64* [[ULLE]], align 8
 // CHECK-NEXT:    [[TMP1678:%.*]] = load i64, i64* [[ULLD]], align 8
 // CHECK-NEXT:    [[TMP1679:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP1677]], i64 [[TMP1678]] seq_cst seq_cst, align 8
 // CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
 // CHECK-NEXT:    ret void
+//
+//
+// CHECK-LABEL: define {{[^@]+}}@bar
+// CHECK-SAME: () #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[CX:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[CV:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[CR:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[CE:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[CD:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[UCX:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[UCV:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[UCR:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[UCE:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[UCD:%.*]] = alloca i8, align 1
+// CHECK-NEXT:    [[SX:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[SV:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[SR:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[SE:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[SD:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[USX:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[USV:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[USR:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[USE:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[USD:%.*]] = alloca i16, align 2
+// CHECK-NEXT:    [[IX:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[IV:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[IR:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[IE:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[ID:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[UIX:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[UIV:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[UIR:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[UIE:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[UID:%.*]] = alloca i32, align 4
+// CHECK-NEXT:    [[LX:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LV:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LR:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LE:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LD:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULX:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULV:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULR:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULE:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULD:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LLX:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LLV:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LLR:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LLE:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[LLD:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULLX:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULLV:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULLR:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULLE:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[ULLD:%.*]] = alloca i64, align 8
+// CHECK-NEXT:    [[TMP0:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP1:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP0]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP1]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP3:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP2]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP3]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP5:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP4]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP5]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP6:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP7:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP6]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP7]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP8:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP9:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP10:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP8]], i8 [[TMP9]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP11:%.*]] = extractvalue { i8, i1 } [[TMP10]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP11]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP12:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP13:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP14:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP12]], i8 [[TMP13]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP15:%.*]] = extractvalue { i8, i1 } [[TMP14]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP15]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP16:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP17:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP16]] monotonic, align 1
+// CHECK-NEXT:    [[TMP18:%.*]] = icmp ugt i8 [[TMP17]], [[TMP16]]
+// CHECK-NEXT:    [[TMP19:%.*]] = select i1 [[TMP18]], i8 [[TMP16]], i8 [[TMP17]]
+// CHECK-NEXT:    store volatile i8 [[TMP19]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP20:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP21:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP20]] monotonic, align 1
+// CHECK-NEXT:    [[TMP22:%.*]] = icmp ult i8 [[TMP21]], [[TMP20]]
+// CHECK-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i8 [[TMP20]], i8 [[TMP21]]
+// CHECK-NEXT:    store volatile i8 [[TMP23]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP24:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP25:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP24]] monotonic, align 1
+// CHECK-NEXT:    [[TMP26:%.*]] = icmp ult i8 [[TMP25]], [[TMP24]]
+// CHECK-NEXT:    [[TMP27:%.*]] = select i1 [[TMP26]], i8 [[TMP24]], i8 [[TMP25]]
+// CHECK-NEXT:    store volatile i8 [[TMP27]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP28:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP29:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP28]] monotonic, align 1
+// CHECK-NEXT:    [[TMP30:%.*]] = icmp ugt i8 [[TMP29]], [[TMP28]]
+// CHECK-NEXT:    [[TMP31:%.*]] = select i1 [[TMP30]], i8 [[TMP28]], i8 [[TMP29]]
+// CHECK-NEXT:    store volatile i8 [[TMP31]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP32:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP33:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP34:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP32]], i8 [[TMP33]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP35:%.*]] = extractvalue { i8, i1 } [[TMP34]], 0
+// CHECK-NEXT:    [[TMP36:%.*]] = extractvalue { i8, i1 } [[TMP34]], 1
+// CHECK-NEXT:    [[TMP37:%.*]] = select i1 [[TMP36]], i8 [[TMP32]], i8 [[TMP35]]
+// CHECK-NEXT:    store volatile i8 [[TMP37]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP38:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP39:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP40:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP38]], i8 [[TMP39]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP41:%.*]] = extractvalue { i8, i1 } [[TMP40]], 0
+// CHECK-NEXT:    [[TMP42:%.*]] = extractvalue { i8, i1 } [[TMP40]], 1
+// CHECK-NEXT:    [[TMP43:%.*]] = select i1 [[TMP42]], i8 [[TMP38]], i8 [[TMP41]]
+// CHECK-NEXT:    store volatile i8 [[TMP43]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP44:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP45:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP46:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP44]], i8 [[TMP45]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP47:%.*]] = extractvalue { i8, i1 } [[TMP46]], 0
+// CHECK-NEXT:    [[TMP48:%.*]] = extractvalue { i8, i1 } [[TMP46]], 1
+// CHECK-NEXT:    br i1 [[TMP48]], label [[CX_ATOMIC_EXIT:%.*]], label [[CX_ATOMIC_CONT:%.*]]
+// CHECK:       cx.atomic.cont:
+// CHECK-NEXT:    store i8 [[TMP47]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT]]
+// CHECK:       cx.atomic.exit:
+// CHECK-NEXT:    [[TMP49:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP50:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP51:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP49]], i8 [[TMP50]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP52:%.*]] = extractvalue { i8, i1 } [[TMP51]], 0
+// CHECK-NEXT:    [[TMP53:%.*]] = extractvalue { i8, i1 } [[TMP51]], 1
+// CHECK-NEXT:    br i1 [[TMP53]], label [[CX_ATOMIC_EXIT1:%.*]], label [[CX_ATOMIC_CONT2:%.*]]
+// CHECK:       cx.atomic.cont2:
+// CHECK-NEXT:    store i8 [[TMP52]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT1]]
+// CHECK:       cx.atomic.exit1:
+// CHECK-NEXT:    [[TMP54:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP55:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP56:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP54]], i8 [[TMP55]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP57:%.*]] = extractvalue { i8, i1 } [[TMP56]], 1
+// CHECK-NEXT:    [[TMP58:%.*]] = zext i1 [[TMP57]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP58]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP59:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP60:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP61:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP59]], i8 [[TMP60]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP62:%.*]] = extractvalue { i8, i1 } [[TMP61]], 1
+// CHECK-NEXT:    [[TMP63:%.*]] = zext i1 [[TMP62]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP63]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP64:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP65:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP66:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP64]], i8 [[TMP65]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP67:%.*]] = extractvalue { i8, i1 } [[TMP66]], 0
+// CHECK-NEXT:    [[TMP68:%.*]] = extractvalue { i8, i1 } [[TMP66]], 1
+// CHECK-NEXT:    br i1 [[TMP68]], label [[CX_ATOMIC_EXIT3:%.*]], label [[CX_ATOMIC_CONT4:%.*]]
+// CHECK:       cx.atomic.cont4:
+// CHECK-NEXT:    store i8 [[TMP67]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT3]]
+// CHECK:       cx.atomic.exit3:
+// CHECK-NEXT:    [[TMP69:%.*]] = extractvalue { i8, i1 } [[TMP66]], 1
+// CHECK-NEXT:    [[TMP70:%.*]] = zext i1 [[TMP69]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP70]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP71:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP72:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP73:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP71]], i8 [[TMP72]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP74:%.*]] = extractvalue { i8, i1 } [[TMP73]], 0
+// CHECK-NEXT:    [[TMP75:%.*]] = extractvalue { i8, i1 } [[TMP73]], 1
+// CHECK-NEXT:    br i1 [[TMP75]], label [[CX_ATOMIC_EXIT5:%.*]], label [[CX_ATOMIC_CONT6:%.*]]
+// CHECK:       cx.atomic.cont6:
+// CHECK-NEXT:    store i8 [[TMP74]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT5]]
+// CHECK:       cx.atomic.exit5:
+// CHECK-NEXT:    [[TMP76:%.*]] = extractvalue { i8, i1 } [[TMP73]], 1
+// CHECK-NEXT:    [[TMP77:%.*]] = zext i1 [[TMP76]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP77]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP78:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP79:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP78]] acq_rel, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP79]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP80:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP81:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP80]] acq_rel, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP81]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP82:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP83:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP82]] acq_rel, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP83]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP84:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP85:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP84]] acq_rel, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP85]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP86:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP87:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP88:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP86]], i8 [[TMP87]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP89:%.*]] = extractvalue { i8, i1 } [[TMP88]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP89]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP90:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP91:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP92:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP90]], i8 [[TMP91]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP93:%.*]] = extractvalue { i8, i1 } [[TMP92]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP93]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP94:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP95:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP94]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP96:%.*]] = icmp ugt i8 [[TMP95]], [[TMP94]]
+// CHECK-NEXT:    [[TMP97:%.*]] = select i1 [[TMP96]], i8 [[TMP94]], i8 [[TMP95]]
+// CHECK-NEXT:    store volatile i8 [[TMP97]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP98:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP99:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP98]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP100:%.*]] = icmp ult i8 [[TMP99]], [[TMP98]]
+// CHECK-NEXT:    [[TMP101:%.*]] = select i1 [[TMP100]], i8 [[TMP98]], i8 [[TMP99]]
+// CHECK-NEXT:    store volatile i8 [[TMP101]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP102:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP103:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP102]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP104:%.*]] = icmp ult i8 [[TMP103]], [[TMP102]]
+// CHECK-NEXT:    [[TMP105:%.*]] = select i1 [[TMP104]], i8 [[TMP102]], i8 [[TMP103]]
+// CHECK-NEXT:    store volatile i8 [[TMP105]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP106:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP107:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP106]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP108:%.*]] = icmp ugt i8 [[TMP107]], [[TMP106]]
+// CHECK-NEXT:    [[TMP109:%.*]] = select i1 [[TMP108]], i8 [[TMP106]], i8 [[TMP107]]
+// CHECK-NEXT:    store volatile i8 [[TMP109]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP110:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP111:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP112:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP110]], i8 [[TMP111]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP113:%.*]] = extractvalue { i8, i1 } [[TMP112]], 0
+// CHECK-NEXT:    [[TMP114:%.*]] = extractvalue { i8, i1 } [[TMP112]], 1
+// CHECK-NEXT:    [[TMP115:%.*]] = select i1 [[TMP114]], i8 [[TMP110]], i8 [[TMP113]]
+// CHECK-NEXT:    store volatile i8 [[TMP115]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP116:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP117:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP118:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP116]], i8 [[TMP117]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP119:%.*]] = extractvalue { i8, i1 } [[TMP118]], 0
+// CHECK-NEXT:    [[TMP120:%.*]] = extractvalue { i8, i1 } [[TMP118]], 1
+// CHECK-NEXT:    [[TMP121:%.*]] = select i1 [[TMP120]], i8 [[TMP116]], i8 [[TMP119]]
+// CHECK-NEXT:    store volatile i8 [[TMP121]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP122:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP123:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP124:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP122]], i8 [[TMP123]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP125:%.*]] = extractvalue { i8, i1 } [[TMP124]], 0
+// CHECK-NEXT:    [[TMP126:%.*]] = extractvalue { i8, i1 } [[TMP124]], 1
+// CHECK-NEXT:    br i1 [[TMP126]], label [[CX_ATOMIC_EXIT7:%.*]], label [[CX_ATOMIC_CONT8:%.*]]
+// CHECK:       cx.atomic.cont8:
+// CHECK-NEXT:    store i8 [[TMP125]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT7]]
+// CHECK:       cx.atomic.exit7:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP127:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP128:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP129:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP127]], i8 [[TMP128]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP130:%.*]] = extractvalue { i8, i1 } [[TMP129]], 0
+// CHECK-NEXT:    [[TMP131:%.*]] = extractvalue { i8, i1 } [[TMP129]], 1
+// CHECK-NEXT:    br i1 [[TMP131]], label [[CX_ATOMIC_EXIT9:%.*]], label [[CX_ATOMIC_CONT10:%.*]]
+// CHECK:       cx.atomic.cont10:
+// CHECK-NEXT:    store i8 [[TMP130]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT9]]
+// CHECK:       cx.atomic.exit9:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP132:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP133:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP134:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP132]], i8 [[TMP133]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP135:%.*]] = extractvalue { i8, i1 } [[TMP134]], 1
+// CHECK-NEXT:    [[TMP136:%.*]] = zext i1 [[TMP135]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP136]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP137:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP138:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP139:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP137]], i8 [[TMP138]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP140:%.*]] = extractvalue { i8, i1 } [[TMP139]], 1
+// CHECK-NEXT:    [[TMP141:%.*]] = zext i1 [[TMP140]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP141]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP142:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP143:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP144:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP142]], i8 [[TMP143]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP145:%.*]] = extractvalue { i8, i1 } [[TMP144]], 0
+// CHECK-NEXT:    [[TMP146:%.*]] = extractvalue { i8, i1 } [[TMP144]], 1
+// CHECK-NEXT:    br i1 [[TMP146]], label [[CX_ATOMIC_EXIT11:%.*]], label [[CX_ATOMIC_CONT12:%.*]]
+// CHECK:       cx.atomic.cont12:
+// CHECK-NEXT:    store i8 [[TMP145]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT11]]
+// CHECK:       cx.atomic.exit11:
+// CHECK-NEXT:    [[TMP147:%.*]] = extractvalue { i8, i1 } [[TMP144]], 1
+// CHECK-NEXT:    [[TMP148:%.*]] = zext i1 [[TMP147]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP148]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP149:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP150:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP151:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP149]], i8 [[TMP150]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP152:%.*]] = extractvalue { i8, i1 } [[TMP151]], 0
+// CHECK-NEXT:    [[TMP153:%.*]] = extractvalue { i8, i1 } [[TMP151]], 1
+// CHECK-NEXT:    br i1 [[TMP153]], label [[CX_ATOMIC_EXIT13:%.*]], label [[CX_ATOMIC_CONT14:%.*]]
+// CHECK:       cx.atomic.cont14:
+// CHECK-NEXT:    store i8 [[TMP152]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT13]]
+// CHECK:       cx.atomic.exit13:
+// CHECK-NEXT:    [[TMP154:%.*]] = extractvalue { i8, i1 } [[TMP151]], 1
+// CHECK-NEXT:    [[TMP155:%.*]] = zext i1 [[TMP154]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP155]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP156:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP157:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP156]] acquire, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP157]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP158:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP159:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP158]] acquire, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP159]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP160:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP161:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP160]] acquire, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP161]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP162:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP163:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP162]] acquire, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP163]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP164:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP165:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP166:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP164]], i8 [[TMP165]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP167:%.*]] = extractvalue { i8, i1 } [[TMP166]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP167]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP168:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP169:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP170:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP168]], i8 [[TMP169]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP171:%.*]] = extractvalue { i8, i1 } [[TMP170]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP171]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP172:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP173:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP172]] acquire, align 1
+// CHECK-NEXT:    [[TMP174:%.*]] = icmp ugt i8 [[TMP173]], [[TMP172]]
+// CHECK-NEXT:    [[TMP175:%.*]] = select i1 [[TMP174]], i8 [[TMP172]], i8 [[TMP173]]
+// CHECK-NEXT:    store volatile i8 [[TMP175]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP176:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP177:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP176]] acquire, align 1
+// CHECK-NEXT:    [[TMP178:%.*]] = icmp ult i8 [[TMP177]], [[TMP176]]
+// CHECK-NEXT:    [[TMP179:%.*]] = select i1 [[TMP178]], i8 [[TMP176]], i8 [[TMP177]]
+// CHECK-NEXT:    store volatile i8 [[TMP179]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP180:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP181:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP180]] acquire, align 1
+// CHECK-NEXT:    [[TMP182:%.*]] = icmp ult i8 [[TMP181]], [[TMP180]]
+// CHECK-NEXT:    [[TMP183:%.*]] = select i1 [[TMP182]], i8 [[TMP180]], i8 [[TMP181]]
+// CHECK-NEXT:    store volatile i8 [[TMP183]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP184:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP185:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP184]] acquire, align 1
+// CHECK-NEXT:    [[TMP186:%.*]] = icmp ugt i8 [[TMP185]], [[TMP184]]
+// CHECK-NEXT:    [[TMP187:%.*]] = select i1 [[TMP186]], i8 [[TMP184]], i8 [[TMP185]]
+// CHECK-NEXT:    store volatile i8 [[TMP187]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP188:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP189:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP190:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP188]], i8 [[TMP189]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP191:%.*]] = extractvalue { i8, i1 } [[TMP190]], 0
+// CHECK-NEXT:    [[TMP192:%.*]] = extractvalue { i8, i1 } [[TMP190]], 1
+// CHECK-NEXT:    [[TMP193:%.*]] = select i1 [[TMP192]], i8 [[TMP188]], i8 [[TMP191]]
+// CHECK-NEXT:    store volatile i8 [[TMP193]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP194:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP195:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP196:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP194]], i8 [[TMP195]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP197:%.*]] = extractvalue { i8, i1 } [[TMP196]], 0
+// CHECK-NEXT:    [[TMP198:%.*]] = extractvalue { i8, i1 } [[TMP196]], 1
+// CHECK-NEXT:    [[TMP199:%.*]] = select i1 [[TMP198]], i8 [[TMP194]], i8 [[TMP197]]
+// CHECK-NEXT:    store volatile i8 [[TMP199]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP200:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP201:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP202:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP200]], i8 [[TMP201]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP203:%.*]] = extractvalue { i8, i1 } [[TMP202]], 0
+// CHECK-NEXT:    [[TMP204:%.*]] = extractvalue { i8, i1 } [[TMP202]], 1
+// CHECK-NEXT:    br i1 [[TMP204]], label [[CX_ATOMIC_EXIT15:%.*]], label [[CX_ATOMIC_CONT16:%.*]]
+// CHECK:       cx.atomic.cont16:
+// CHECK-NEXT:    store i8 [[TMP203]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT15]]
+// CHECK:       cx.atomic.exit15:
+// CHECK-NEXT:    [[TMP205:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP206:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP207:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP205]], i8 [[TMP206]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP208:%.*]] = extractvalue { i8, i1 } [[TMP207]], 0
+// CHECK-NEXT:    [[TMP209:%.*]] = extractvalue { i8, i1 } [[TMP207]], 1
+// CHECK-NEXT:    br i1 [[TMP209]], label [[CX_ATOMIC_EXIT17:%.*]], label [[CX_ATOMIC_CONT18:%.*]]
+// CHECK:       cx.atomic.cont18:
+// CHECK-NEXT:    store i8 [[TMP208]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT17]]
+// CHECK:       cx.atomic.exit17:
+// CHECK-NEXT:    [[TMP210:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP211:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP212:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP210]], i8 [[TMP211]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP213:%.*]] = extractvalue { i8, i1 } [[TMP212]], 1
+// CHECK-NEXT:    [[TMP214:%.*]] = zext i1 [[TMP213]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP214]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP215:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP216:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP217:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP215]], i8 [[TMP216]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP218:%.*]] = extractvalue { i8, i1 } [[TMP217]], 1
+// CHECK-NEXT:    [[TMP219:%.*]] = zext i1 [[TMP218]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP219]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP220:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP221:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP222:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP220]], i8 [[TMP221]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP223:%.*]] = extractvalue { i8, i1 } [[TMP222]], 0
+// CHECK-NEXT:    [[TMP224:%.*]] = extractvalue { i8, i1 } [[TMP222]], 1
+// CHECK-NEXT:    br i1 [[TMP224]], label [[CX_ATOMIC_EXIT19:%.*]], label [[CX_ATOMIC_CONT20:%.*]]
+// CHECK:       cx.atomic.cont20:
+// CHECK-NEXT:    store i8 [[TMP223]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT19]]
+// CHECK:       cx.atomic.exit19:
+// CHECK-NEXT:    [[TMP225:%.*]] = extractvalue { i8, i1 } [[TMP222]], 1
+// CHECK-NEXT:    [[TMP226:%.*]] = zext i1 [[TMP225]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP226]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP227:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP228:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP229:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP227]], i8 [[TMP228]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP230:%.*]] = extractvalue { i8, i1 } [[TMP229]], 0
+// CHECK-NEXT:    [[TMP231:%.*]] = extractvalue { i8, i1 } [[TMP229]], 1
+// CHECK-NEXT:    br i1 [[TMP231]], label [[CX_ATOMIC_EXIT21:%.*]], label [[CX_ATOMIC_CONT22:%.*]]
+// CHECK:       cx.atomic.cont22:
+// CHECK-NEXT:    store i8 [[TMP230]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT21]]
+// CHECK:       cx.atomic.exit21:
+// CHECK-NEXT:    [[TMP232:%.*]] = extractvalue { i8, i1 } [[TMP229]], 1
+// CHECK-NEXT:    [[TMP233:%.*]] = zext i1 [[TMP232]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP233]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP234:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP235:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP234]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP235]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP236:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP237:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP236]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP237]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP238:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP239:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP238]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP239]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP240:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP241:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP240]] monotonic, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP241]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP242:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP243:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP244:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP242]], i8 [[TMP243]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP245:%.*]] = extractvalue { i8, i1 } [[TMP244]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP245]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP246:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP247:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP248:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP246]], i8 [[TMP247]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP249:%.*]] = extractvalue { i8, i1 } [[TMP248]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP249]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP250:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP251:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP250]] monotonic, align 1
+// CHECK-NEXT:    [[TMP252:%.*]] = icmp ugt i8 [[TMP251]], [[TMP250]]
+// CHECK-NEXT:    [[TMP253:%.*]] = select i1 [[TMP252]], i8 [[TMP250]], i8 [[TMP251]]
+// CHECK-NEXT:    store volatile i8 [[TMP253]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP254:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP255:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP254]] monotonic, align 1
+// CHECK-NEXT:    [[TMP256:%.*]] = icmp ult i8 [[TMP255]], [[TMP254]]
+// CHECK-NEXT:    [[TMP257:%.*]] = select i1 [[TMP256]], i8 [[TMP254]], i8 [[TMP255]]
+// CHECK-NEXT:    store volatile i8 [[TMP257]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP258:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP259:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP258]] monotonic, align 1
+// CHECK-NEXT:    [[TMP260:%.*]] = icmp ult i8 [[TMP259]], [[TMP258]]
+// CHECK-NEXT:    [[TMP261:%.*]] = select i1 [[TMP260]], i8 [[TMP258]], i8 [[TMP259]]
+// CHECK-NEXT:    store volatile i8 [[TMP261]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP262:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP263:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP262]] monotonic, align 1
+// CHECK-NEXT:    [[TMP264:%.*]] = icmp ugt i8 [[TMP263]], [[TMP262]]
+// CHECK-NEXT:    [[TMP265:%.*]] = select i1 [[TMP264]], i8 [[TMP262]], i8 [[TMP263]]
+// CHECK-NEXT:    store volatile i8 [[TMP265]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP266:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP267:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP268:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP266]], i8 [[TMP267]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP269:%.*]] = extractvalue { i8, i1 } [[TMP268]], 0
+// CHECK-NEXT:    [[TMP270:%.*]] = extractvalue { i8, i1 } [[TMP268]], 1
+// CHECK-NEXT:    [[TMP271:%.*]] = select i1 [[TMP270]], i8 [[TMP266]], i8 [[TMP269]]
+// CHECK-NEXT:    store volatile i8 [[TMP271]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP272:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP273:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP274:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP272]], i8 [[TMP273]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP275:%.*]] = extractvalue { i8, i1 } [[TMP274]], 0
+// CHECK-NEXT:    [[TMP276:%.*]] = extractvalue { i8, i1 } [[TMP274]], 1
+// CHECK-NEXT:    [[TMP277:%.*]] = select i1 [[TMP276]], i8 [[TMP272]], i8 [[TMP275]]
+// CHECK-NEXT:    store volatile i8 [[TMP277]], i8* [[CV]], align 1
+// CHECK-NEXT:    [[TMP278:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP279:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP280:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP278]], i8 [[TMP279]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP281:%.*]] = extractvalue { i8, i1 } [[TMP280]], 0
+// CHECK-NEXT:    [[TMP282:%.*]] = extractvalue { i8, i1 } [[TMP280]], 1
+// CHECK-NEXT:    br i1 [[TMP282]], label [[CX_ATOMIC_EXIT23:%.*]], label [[CX_ATOMIC_CONT24:%.*]]
+// CHECK:       cx.atomic.cont24:
+// CHECK-NEXT:    store i8 [[TMP281]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT23]]
+// CHECK:       cx.atomic.exit23:
+// CHECK-NEXT:    [[TMP283:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP284:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP285:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP283]], i8 [[TMP284]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP286:%.*]] = extractvalue { i8, i1 } [[TMP285]], 0
+// CHECK-NEXT:    [[TMP287:%.*]] = extractvalue { i8, i1 } [[TMP285]], 1
+// CHECK-NEXT:    br i1 [[TMP287]], label [[CX_ATOMIC_EXIT25:%.*]], label [[CX_ATOMIC_CONT26:%.*]]
+// CHECK:       cx.atomic.cont26:
+// CHECK-NEXT:    store i8 [[TMP286]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT25]]
+// CHECK:       cx.atomic.exit25:
+// CHECK-NEXT:    [[TMP288:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP289:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP290:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP288]], i8 [[TMP289]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP291:%.*]] = extractvalue { i8, i1 } [[TMP290]], 1
+// CHECK-NEXT:    [[TMP292:%.*]] = zext i1 [[TMP291]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP292]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP293:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP294:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP295:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP293]], i8 [[TMP294]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP296:%.*]] = extractvalue { i8, i1 } [[TMP295]], 1
+// CHECK-NEXT:    [[TMP297:%.*]] = zext i1 [[TMP296]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP297]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP298:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP299:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP300:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP298]], i8 [[TMP299]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP301:%.*]] = extractvalue { i8, i1 } [[TMP300]], 0
+// CHECK-NEXT:    [[TMP302:%.*]] = extractvalue { i8, i1 } [[TMP300]], 1
+// CHECK-NEXT:    br i1 [[TMP302]], label [[CX_ATOMIC_EXIT27:%.*]], label [[CX_ATOMIC_CONT28:%.*]]
+// CHECK:       cx.atomic.cont28:
+// CHECK-NEXT:    store i8 [[TMP301]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT27]]
+// CHECK:       cx.atomic.exit27:
+// CHECK-NEXT:    [[TMP303:%.*]] = extractvalue { i8, i1 } [[TMP300]], 1
+// CHECK-NEXT:    [[TMP304:%.*]] = zext i1 [[TMP303]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP304]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP305:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP306:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP307:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP305]], i8 [[TMP306]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP308:%.*]] = extractvalue { i8, i1 } [[TMP307]], 0
+// CHECK-NEXT:    [[TMP309:%.*]] = extractvalue { i8, i1 } [[TMP307]], 1
+// CHECK-NEXT:    br i1 [[TMP309]], label [[CX_ATOMIC_EXIT29:%.*]], label [[CX_ATOMIC_CONT30:%.*]]
+// CHECK:       cx.atomic.cont30:
+// CHECK-NEXT:    store i8 [[TMP308]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT29]]
+// CHECK:       cx.atomic.exit29:
+// CHECK-NEXT:    [[TMP310:%.*]] = extractvalue { i8, i1 } [[TMP307]], 1
+// CHECK-NEXT:    [[TMP311:%.*]] = zext i1 [[TMP310]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP311]], i8* [[CR]], align 1
+// CHECK-NEXT:    [[TMP312:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP313:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP312]] release, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP313]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP314:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP315:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP314]] release, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP315]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP316:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP317:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP316]] release, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP317]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP318:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP319:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP318]] release, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP319]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP320:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP321:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP322:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP320]], i8 [[TMP321]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP323:%.*]] = extractvalue { i8, i1 } [[TMP322]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP323]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP324:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP325:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP326:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP324]], i8 [[TMP325]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP327:%.*]] = extractvalue { i8, i1 } [[TMP326]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP327]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP328:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP329:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP328]] release, align 1
+// CHECK-NEXT:    [[TMP330:%.*]] = icmp ugt i8 [[TMP329]], [[TMP328]]
+// CHECK-NEXT:    [[TMP331:%.*]] = select i1 [[TMP330]], i8 [[TMP328]], i8 [[TMP329]]
+// CHECK-NEXT:    store volatile i8 [[TMP331]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP332:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP333:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP332]] release, align 1
+// CHECK-NEXT:    [[TMP334:%.*]] = icmp ult i8 [[TMP333]], [[TMP332]]
+// CHECK-NEXT:    [[TMP335:%.*]] = select i1 [[TMP334]], i8 [[TMP332]], i8 [[TMP333]]
+// CHECK-NEXT:    store volatile i8 [[TMP335]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP336:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP337:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP336]] release, align 1
+// CHECK-NEXT:    [[TMP338:%.*]] = icmp ult i8 [[TMP337]], [[TMP336]]
+// CHECK-NEXT:    [[TMP339:%.*]] = select i1 [[TMP338]], i8 [[TMP336]], i8 [[TMP337]]
+// CHECK-NEXT:    store volatile i8 [[TMP339]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP340:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP341:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP340]] release, align 1
+// CHECK-NEXT:    [[TMP342:%.*]] = icmp ugt i8 [[TMP341]], [[TMP340]]
+// CHECK-NEXT:    [[TMP343:%.*]] = select i1 [[TMP342]], i8 [[TMP340]], i8 [[TMP341]]
+// CHECK-NEXT:    store volatile i8 [[TMP343]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP344:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP345:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP346:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP344]], i8 [[TMP345]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP347:%.*]] = extractvalue { i8, i1 } [[TMP346]], 0
+// CHECK-NEXT:    [[TMP348:%.*]] = extractvalue { i8, i1 } [[TMP346]], 1
+// CHECK-NEXT:    [[TMP349:%.*]] = select i1 [[TMP348]], i8 [[TMP344]], i8 [[TMP347]]
+// CHECK-NEXT:    store volatile i8 [[TMP349]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP350:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP351:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP352:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP350]], i8 [[TMP351]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP353:%.*]] = extractvalue { i8, i1 } [[TMP352]], 0
+// CHECK-NEXT:    [[TMP354:%.*]] = extractvalue { i8, i1 } [[TMP352]], 1
+// CHECK-NEXT:    [[TMP355:%.*]] = select i1 [[TMP354]], i8 [[TMP350]], i8 [[TMP353]]
+// CHECK-NEXT:    store volatile i8 [[TMP355]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP356:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP357:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP358:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP356]], i8 [[TMP357]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP359:%.*]] = extractvalue { i8, i1 } [[TMP358]], 0
+// CHECK-NEXT:    [[TMP360:%.*]] = extractvalue { i8, i1 } [[TMP358]], 1
+// CHECK-NEXT:    br i1 [[TMP360]], label [[CX_ATOMIC_EXIT31:%.*]], label [[CX_ATOMIC_CONT32:%.*]]
+// CHECK:       cx.atomic.cont32:
+// CHECK-NEXT:    store i8 [[TMP359]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT31]]
+// CHECK:       cx.atomic.exit31:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP361:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP362:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP363:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP361]], i8 [[TMP362]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP364:%.*]] = extractvalue { i8, i1 } [[TMP363]], 0
+// CHECK-NEXT:    [[TMP365:%.*]] = extractvalue { i8, i1 } [[TMP363]], 1
+// CHECK-NEXT:    br i1 [[TMP365]], label [[CX_ATOMIC_EXIT33:%.*]], label [[CX_ATOMIC_CONT34:%.*]]
+// CHECK:       cx.atomic.cont34:
+// CHECK-NEXT:    store i8 [[TMP364]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT33]]
+// CHECK:       cx.atomic.exit33:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP366:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP367:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP368:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP366]], i8 [[TMP367]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP369:%.*]] = extractvalue { i8, i1 } [[TMP368]], 1
+// CHECK-NEXT:    [[TMP370:%.*]] = zext i1 [[TMP369]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP370]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP371:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP372:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP373:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP371]], i8 [[TMP372]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP374:%.*]] = extractvalue { i8, i1 } [[TMP373]], 1
+// CHECK-NEXT:    [[TMP375:%.*]] = zext i1 [[TMP374]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP375]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP376:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP377:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP378:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP376]], i8 [[TMP377]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP379:%.*]] = extractvalue { i8, i1 } [[TMP378]], 0
+// CHECK-NEXT:    [[TMP380:%.*]] = extractvalue { i8, i1 } [[TMP378]], 1
+// CHECK-NEXT:    br i1 [[TMP380]], label [[CX_ATOMIC_EXIT35:%.*]], label [[CX_ATOMIC_CONT36:%.*]]
+// CHECK:       cx.atomic.cont36:
+// CHECK-NEXT:    store i8 [[TMP379]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT35]]
+// CHECK:       cx.atomic.exit35:
+// CHECK-NEXT:    [[TMP381:%.*]] = extractvalue { i8, i1 } [[TMP378]], 1
+// CHECK-NEXT:    [[TMP382:%.*]] = zext i1 [[TMP381]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP382]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP383:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP384:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP385:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP383]], i8 [[TMP384]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP386:%.*]] = extractvalue { i8, i1 } [[TMP385]], 0
+// CHECK-NEXT:    [[TMP387:%.*]] = extractvalue { i8, i1 } [[TMP385]], 1
+// CHECK-NEXT:    br i1 [[TMP387]], label [[CX_ATOMIC_EXIT37:%.*]], label [[CX_ATOMIC_CONT38:%.*]]
+// CHECK:       cx.atomic.cont38:
+// CHECK-NEXT:    store i8 [[TMP386]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT37]]
+// CHECK:       cx.atomic.exit37:
+// CHECK-NEXT:    [[TMP388:%.*]] = extractvalue { i8, i1 } [[TMP385]], 1
+// CHECK-NEXT:    [[TMP389:%.*]] = zext i1 [[TMP388]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP389]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP390:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP391:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP390]] seq_cst, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP391]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP392:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP393:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP392]] seq_cst, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP393]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP394:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP395:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP394]] seq_cst, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP395]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP396:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP397:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP396]] seq_cst, align 1
+// CHECK-NEXT:    store volatile i8 [[TMP397]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP398:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP399:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP400:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP398]], i8 [[TMP399]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP401:%.*]] = extractvalue { i8, i1 } [[TMP400]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP401]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP402:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP403:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP404:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP402]], i8 [[TMP403]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP405:%.*]] = extractvalue { i8, i1 } [[TMP404]], 0
+// CHECK-NEXT:    store volatile i8 [[TMP405]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP406:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP407:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP406]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP408:%.*]] = icmp ugt i8 [[TMP407]], [[TMP406]]
+// CHECK-NEXT:    [[TMP409:%.*]] = select i1 [[TMP408]], i8 [[TMP406]], i8 [[TMP407]]
+// CHECK-NEXT:    store volatile i8 [[TMP409]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP410:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP411:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP410]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP412:%.*]] = icmp ult i8 [[TMP411]], [[TMP410]]
+// CHECK-NEXT:    [[TMP413:%.*]] = select i1 [[TMP412]], i8 [[TMP410]], i8 [[TMP411]]
+// CHECK-NEXT:    store volatile i8 [[TMP413]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP414:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP415:%.*]] = atomicrmw umin i8* [[CX]], i8 [[TMP414]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP416:%.*]] = icmp ult i8 [[TMP415]], [[TMP414]]
+// CHECK-NEXT:    [[TMP417:%.*]] = select i1 [[TMP416]], i8 [[TMP414]], i8 [[TMP415]]
+// CHECK-NEXT:    store volatile i8 [[TMP417]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP418:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP419:%.*]] = atomicrmw umax i8* [[CX]], i8 [[TMP418]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP420:%.*]] = icmp ugt i8 [[TMP419]], [[TMP418]]
+// CHECK-NEXT:    [[TMP421:%.*]] = select i1 [[TMP420]], i8 [[TMP418]], i8 [[TMP419]]
+// CHECK-NEXT:    store volatile i8 [[TMP421]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP422:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP423:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP424:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP422]], i8 [[TMP423]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP425:%.*]] = extractvalue { i8, i1 } [[TMP424]], 0
+// CHECK-NEXT:    [[TMP426:%.*]] = extractvalue { i8, i1 } [[TMP424]], 1
+// CHECK-NEXT:    [[TMP427:%.*]] = select i1 [[TMP426]], i8 [[TMP422]], i8 [[TMP425]]
+// CHECK-NEXT:    store volatile i8 [[TMP427]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP428:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP429:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP430:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP428]], i8 [[TMP429]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP431:%.*]] = extractvalue { i8, i1 } [[TMP430]], 0
+// CHECK-NEXT:    [[TMP432:%.*]] = extractvalue { i8, i1 } [[TMP430]], 1
+// CHECK-NEXT:    [[TMP433:%.*]] = select i1 [[TMP432]], i8 [[TMP428]], i8 [[TMP431]]
+// CHECK-NEXT:    store volatile i8 [[TMP433]], i8* [[CV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP434:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP435:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP436:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP434]], i8 [[TMP435]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP437:%.*]] = extractvalue { i8, i1 } [[TMP436]], 0
+// CHECK-NEXT:    [[TMP438:%.*]] = extractvalue { i8, i1 } [[TMP436]], 1
+// CHECK-NEXT:    br i1 [[TMP438]], label [[CX_ATOMIC_EXIT39:%.*]], label [[CX_ATOMIC_CONT40:%.*]]
+// CHECK:       cx.atomic.cont40:
+// CHECK-NEXT:    store i8 [[TMP437]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT39]]
+// CHECK:       cx.atomic.exit39:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP439:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP440:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP441:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP439]], i8 [[TMP440]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP442:%.*]] = extractvalue { i8, i1 } [[TMP441]], 0
+// CHECK-NEXT:    [[TMP443:%.*]] = extractvalue { i8, i1 } [[TMP441]], 1
+// CHECK-NEXT:    br i1 [[TMP443]], label [[CX_ATOMIC_EXIT41:%.*]], label [[CX_ATOMIC_CONT42:%.*]]
+// CHECK:       cx.atomic.cont42:
+// CHECK-NEXT:    store i8 [[TMP442]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT41]]
+// CHECK:       cx.atomic.exit41:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP444:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP445:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP446:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP444]], i8 [[TMP445]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP447:%.*]] = extractvalue { i8, i1 } [[TMP446]], 1
+// CHECK-NEXT:    [[TMP448:%.*]] = zext i1 [[TMP447]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP448]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP449:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP450:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP451:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP449]], i8 [[TMP450]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP452:%.*]] = extractvalue { i8, i1 } [[TMP451]], 1
+// CHECK-NEXT:    [[TMP453:%.*]] = zext i1 [[TMP452]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP453]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP454:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP455:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP456:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP454]], i8 [[TMP455]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP457:%.*]] = extractvalue { i8, i1 } [[TMP456]], 0
+// CHECK-NEXT:    [[TMP458:%.*]] = extractvalue { i8, i1 } [[TMP456]], 1
+// CHECK-NEXT:    br i1 [[TMP458]], label [[CX_ATOMIC_EXIT43:%.*]], label [[CX_ATOMIC_CONT44:%.*]]
+// CHECK:       cx.atomic.cont44:
+// CHECK-NEXT:    store i8 [[TMP457]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT43]]
+// CHECK:       cx.atomic.exit43:
+// CHECK-NEXT:    [[TMP459:%.*]] = extractvalue { i8, i1 } [[TMP456]], 1
+// CHECK-NEXT:    [[TMP460:%.*]] = zext i1 [[TMP459]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP460]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP461:%.*]] = load i8, i8* [[CE]], align 1
+// CHECK-NEXT:    [[TMP462:%.*]] = load i8, i8* [[CD]], align 1
+// CHECK-NEXT:    [[TMP463:%.*]] = cmpxchg i8* [[CX]], i8 [[TMP461]], i8 [[TMP462]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP464:%.*]] = extractvalue { i8, i1 } [[TMP463]], 0
+// CHECK-NEXT:    [[TMP465:%.*]] = extractvalue { i8, i1 } [[TMP463]], 1
+// CHECK-NEXT:    br i1 [[TMP465]], label [[CX_ATOMIC_EXIT45:%.*]], label [[CX_ATOMIC_CONT46:%.*]]
+// CHECK:       cx.atomic.cont46:
+// CHECK-NEXT:    store i8 [[TMP464]], i8* [[CV]], align 1
+// CHECK-NEXT:    br label [[CX_ATOMIC_EXIT45]]
+// CHECK:       cx.atomic.exit45:
+// CHECK-NEXT:    [[TMP466:%.*]] = extractvalue { i8, i1 } [[TMP463]], 1
+// CHECK-NEXT:    [[TMP467:%.*]] = zext i1 [[TMP466]] to i8
+// CHECK-NEXT:    store volatile i8 [[TMP467]], i8* [[CR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP468:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP469:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP468]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP469]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP470:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP471:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP470]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP471]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP472:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP473:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP472]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP473]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP474:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP475:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP474]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP475]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP476:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP477:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP478:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP476]], i8 [[TMP477]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP479:%.*]] = extractvalue { i8, i1 } [[TMP478]], 0
+// CHECK-NEXT:    store i8 [[TMP479]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP480:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP481:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP482:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP480]], i8 [[TMP481]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP483:%.*]] = extractvalue { i8, i1 } [[TMP482]], 0
+// CHECK-NEXT:    store i8 [[TMP483]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP484:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP485:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP484]] monotonic, align 1
+// CHECK-NEXT:    [[TMP486:%.*]] = icmp ugt i8 [[TMP485]], [[TMP484]]
+// CHECK-NEXT:    [[TMP487:%.*]] = select i1 [[TMP486]], i8 [[TMP484]], i8 [[TMP485]]
+// CHECK-NEXT:    store i8 [[TMP487]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP488:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP489:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP488]] monotonic, align 1
+// CHECK-NEXT:    [[TMP490:%.*]] = icmp ult i8 [[TMP489]], [[TMP488]]
+// CHECK-NEXT:    [[TMP491:%.*]] = select i1 [[TMP490]], i8 [[TMP488]], i8 [[TMP489]]
+// CHECK-NEXT:    store i8 [[TMP491]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP492:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP493:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP492]] monotonic, align 1
+// CHECK-NEXT:    [[TMP494:%.*]] = icmp ult i8 [[TMP493]], [[TMP492]]
+// CHECK-NEXT:    [[TMP495:%.*]] = select i1 [[TMP494]], i8 [[TMP492]], i8 [[TMP493]]
+// CHECK-NEXT:    store i8 [[TMP495]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP496:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP497:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP496]] monotonic, align 1
+// CHECK-NEXT:    [[TMP498:%.*]] = icmp ugt i8 [[TMP497]], [[TMP496]]
+// CHECK-NEXT:    [[TMP499:%.*]] = select i1 [[TMP498]], i8 [[TMP496]], i8 [[TMP497]]
+// CHECK-NEXT:    store i8 [[TMP499]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP500:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP501:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP502:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP500]], i8 [[TMP501]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP503:%.*]] = extractvalue { i8, i1 } [[TMP502]], 0
+// CHECK-NEXT:    [[TMP504:%.*]] = extractvalue { i8, i1 } [[TMP502]], 1
+// CHECK-NEXT:    [[TMP505:%.*]] = select i1 [[TMP504]], i8 [[TMP500]], i8 [[TMP503]]
+// CHECK-NEXT:    store i8 [[TMP505]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP506:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP507:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP508:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP506]], i8 [[TMP507]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP509:%.*]] = extractvalue { i8, i1 } [[TMP508]], 0
+// CHECK-NEXT:    [[TMP510:%.*]] = extractvalue { i8, i1 } [[TMP508]], 1
+// CHECK-NEXT:    [[TMP511:%.*]] = select i1 [[TMP510]], i8 [[TMP506]], i8 [[TMP509]]
+// CHECK-NEXT:    store i8 [[TMP511]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP512:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP513:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP514:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP512]], i8 [[TMP513]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP515:%.*]] = extractvalue { i8, i1 } [[TMP514]], 0
+// CHECK-NEXT:    [[TMP516:%.*]] = extractvalue { i8, i1 } [[TMP514]], 1
+// CHECK-NEXT:    br i1 [[TMP516]], label [[UCX_ATOMIC_EXIT:%.*]], label [[UCX_ATOMIC_CONT:%.*]]
+// CHECK:       ucx.atomic.cont:
+// CHECK-NEXT:    store i8 [[TMP515]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT]]
+// CHECK:       ucx.atomic.exit:
+// CHECK-NEXT:    [[TMP517:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP518:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP519:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP517]], i8 [[TMP518]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP520:%.*]] = extractvalue { i8, i1 } [[TMP519]], 0
+// CHECK-NEXT:    [[TMP521:%.*]] = extractvalue { i8, i1 } [[TMP519]], 1
+// CHECK-NEXT:    br i1 [[TMP521]], label [[UCX_ATOMIC_EXIT47:%.*]], label [[UCX_ATOMIC_CONT48:%.*]]
+// CHECK:       ucx.atomic.cont48:
+// CHECK-NEXT:    store i8 [[TMP520]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT47]]
+// CHECK:       ucx.atomic.exit47:
+// CHECK-NEXT:    [[TMP522:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP523:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP524:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP522]], i8 [[TMP523]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP525:%.*]] = extractvalue { i8, i1 } [[TMP524]], 1
+// CHECK-NEXT:    [[TMP526:%.*]] = zext i1 [[TMP525]] to i8
+// CHECK-NEXT:    store i8 [[TMP526]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP527:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP528:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP529:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP527]], i8 [[TMP528]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP530:%.*]] = extractvalue { i8, i1 } [[TMP529]], 1
+// CHECK-NEXT:    [[TMP531:%.*]] = zext i1 [[TMP530]] to i8
+// CHECK-NEXT:    store i8 [[TMP531]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP532:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP533:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP534:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP532]], i8 [[TMP533]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP535:%.*]] = extractvalue { i8, i1 } [[TMP534]], 0
+// CHECK-NEXT:    [[TMP536:%.*]] = extractvalue { i8, i1 } [[TMP534]], 1
+// CHECK-NEXT:    br i1 [[TMP536]], label [[UCX_ATOMIC_EXIT49:%.*]], label [[UCX_ATOMIC_CONT50:%.*]]
+// CHECK:       ucx.atomic.cont50:
+// CHECK-NEXT:    store i8 [[TMP535]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT49]]
+// CHECK:       ucx.atomic.exit49:
+// CHECK-NEXT:    [[TMP537:%.*]] = extractvalue { i8, i1 } [[TMP534]], 1
+// CHECK-NEXT:    [[TMP538:%.*]] = zext i1 [[TMP537]] to i8
+// CHECK-NEXT:    store i8 [[TMP538]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP539:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP540:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP541:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP539]], i8 [[TMP540]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP542:%.*]] = extractvalue { i8, i1 } [[TMP541]], 0
+// CHECK-NEXT:    [[TMP543:%.*]] = extractvalue { i8, i1 } [[TMP541]], 1
+// CHECK-NEXT:    br i1 [[TMP543]], label [[UCX_ATOMIC_EXIT51:%.*]], label [[UCX_ATOMIC_CONT52:%.*]]
+// CHECK:       ucx.atomic.cont52:
+// CHECK-NEXT:    store i8 [[TMP542]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT51]]
+// CHECK:       ucx.atomic.exit51:
+// CHECK-NEXT:    [[TMP544:%.*]] = extractvalue { i8, i1 } [[TMP541]], 1
+// CHECK-NEXT:    [[TMP545:%.*]] = zext i1 [[TMP544]] to i8
+// CHECK-NEXT:    store i8 [[TMP545]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP546:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP547:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP546]] acq_rel, align 1
+// CHECK-NEXT:    store i8 [[TMP547]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP548:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP549:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP548]] acq_rel, align 1
+// CHECK-NEXT:    store i8 [[TMP549]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP550:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP551:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP550]] acq_rel, align 1
+// CHECK-NEXT:    store i8 [[TMP551]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP552:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP553:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP552]] acq_rel, align 1
+// CHECK-NEXT:    store i8 [[TMP553]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP554:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP555:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP556:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP554]], i8 [[TMP555]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP557:%.*]] = extractvalue { i8, i1 } [[TMP556]], 0
+// CHECK-NEXT:    store i8 [[TMP557]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP558:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP559:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP560:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP558]], i8 [[TMP559]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP561:%.*]] = extractvalue { i8, i1 } [[TMP560]], 0
+// CHECK-NEXT:    store i8 [[TMP561]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP562:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP563:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP562]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP564:%.*]] = icmp ugt i8 [[TMP563]], [[TMP562]]
+// CHECK-NEXT:    [[TMP565:%.*]] = select i1 [[TMP564]], i8 [[TMP562]], i8 [[TMP563]]
+// CHECK-NEXT:    store i8 [[TMP565]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP566:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP567:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP566]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP568:%.*]] = icmp ult i8 [[TMP567]], [[TMP566]]
+// CHECK-NEXT:    [[TMP569:%.*]] = select i1 [[TMP568]], i8 [[TMP566]], i8 [[TMP567]]
+// CHECK-NEXT:    store i8 [[TMP569]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP570:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP571:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP570]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP572:%.*]] = icmp ult i8 [[TMP571]], [[TMP570]]
+// CHECK-NEXT:    [[TMP573:%.*]] = select i1 [[TMP572]], i8 [[TMP570]], i8 [[TMP571]]
+// CHECK-NEXT:    store i8 [[TMP573]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP574:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP575:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP574]] acq_rel, align 1
+// CHECK-NEXT:    [[TMP576:%.*]] = icmp ugt i8 [[TMP575]], [[TMP574]]
+// CHECK-NEXT:    [[TMP577:%.*]] = select i1 [[TMP576]], i8 [[TMP574]], i8 [[TMP575]]
+// CHECK-NEXT:    store i8 [[TMP577]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP578:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP579:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP580:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP578]], i8 [[TMP579]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP581:%.*]] = extractvalue { i8, i1 } [[TMP580]], 0
+// CHECK-NEXT:    [[TMP582:%.*]] = extractvalue { i8, i1 } [[TMP580]], 1
+// CHECK-NEXT:    [[TMP583:%.*]] = select i1 [[TMP582]], i8 [[TMP578]], i8 [[TMP581]]
+// CHECK-NEXT:    store i8 [[TMP583]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP584:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP585:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP586:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP584]], i8 [[TMP585]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP587:%.*]] = extractvalue { i8, i1 } [[TMP586]], 0
+// CHECK-NEXT:    [[TMP588:%.*]] = extractvalue { i8, i1 } [[TMP586]], 1
+// CHECK-NEXT:    [[TMP589:%.*]] = select i1 [[TMP588]], i8 [[TMP584]], i8 [[TMP587]]
+// CHECK-NEXT:    store i8 [[TMP589]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP590:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP591:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP592:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP590]], i8 [[TMP591]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP593:%.*]] = extractvalue { i8, i1 } [[TMP592]], 0
+// CHECK-NEXT:    [[TMP594:%.*]] = extractvalue { i8, i1 } [[TMP592]], 1
+// CHECK-NEXT:    br i1 [[TMP594]], label [[UCX_ATOMIC_EXIT53:%.*]], label [[UCX_ATOMIC_CONT54:%.*]]
+// CHECK:       ucx.atomic.cont54:
+// CHECK-NEXT:    store i8 [[TMP593]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT53]]
+// CHECK:       ucx.atomic.exit53:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP595:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP596:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP597:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP595]], i8 [[TMP596]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP598:%.*]] = extractvalue { i8, i1 } [[TMP597]], 0
+// CHECK-NEXT:    [[TMP599:%.*]] = extractvalue { i8, i1 } [[TMP597]], 1
+// CHECK-NEXT:    br i1 [[TMP599]], label [[UCX_ATOMIC_EXIT55:%.*]], label [[UCX_ATOMIC_CONT56:%.*]]
+// CHECK:       ucx.atomic.cont56:
+// CHECK-NEXT:    store i8 [[TMP598]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT55]]
+// CHECK:       ucx.atomic.exit55:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP600:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP601:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP602:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP600]], i8 [[TMP601]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP603:%.*]] = extractvalue { i8, i1 } [[TMP602]], 1
+// CHECK-NEXT:    [[TMP604:%.*]] = zext i1 [[TMP603]] to i8
+// CHECK-NEXT:    store i8 [[TMP604]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP605:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP606:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP607:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP605]], i8 [[TMP606]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP608:%.*]] = extractvalue { i8, i1 } [[TMP607]], 1
+// CHECK-NEXT:    [[TMP609:%.*]] = zext i1 [[TMP608]] to i8
+// CHECK-NEXT:    store i8 [[TMP609]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP610:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP611:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP612:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP610]], i8 [[TMP611]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP613:%.*]] = extractvalue { i8, i1 } [[TMP612]], 0
+// CHECK-NEXT:    [[TMP614:%.*]] = extractvalue { i8, i1 } [[TMP612]], 1
+// CHECK-NEXT:    br i1 [[TMP614]], label [[UCX_ATOMIC_EXIT57:%.*]], label [[UCX_ATOMIC_CONT58:%.*]]
+// CHECK:       ucx.atomic.cont58:
+// CHECK-NEXT:    store i8 [[TMP613]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT57]]
+// CHECK:       ucx.atomic.exit57:
+// CHECK-NEXT:    [[TMP615:%.*]] = extractvalue { i8, i1 } [[TMP612]], 1
+// CHECK-NEXT:    [[TMP616:%.*]] = zext i1 [[TMP615]] to i8
+// CHECK-NEXT:    store i8 [[TMP616]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP617:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP618:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP619:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP617]], i8 [[TMP618]] acq_rel acquire, align 1
+// CHECK-NEXT:    [[TMP620:%.*]] = extractvalue { i8, i1 } [[TMP619]], 0
+// CHECK-NEXT:    [[TMP621:%.*]] = extractvalue { i8, i1 } [[TMP619]], 1
+// CHECK-NEXT:    br i1 [[TMP621]], label [[UCX_ATOMIC_EXIT59:%.*]], label [[UCX_ATOMIC_CONT60:%.*]]
+// CHECK:       ucx.atomic.cont60:
+// CHECK-NEXT:    store i8 [[TMP620]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT59]]
+// CHECK:       ucx.atomic.exit59:
+// CHECK-NEXT:    [[TMP622:%.*]] = extractvalue { i8, i1 } [[TMP619]], 1
+// CHECK-NEXT:    [[TMP623:%.*]] = zext i1 [[TMP622]] to i8
+// CHECK-NEXT:    store i8 [[TMP623]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP624:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP625:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP624]] acquire, align 1
+// CHECK-NEXT:    store i8 [[TMP625]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP626:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP627:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP626]] acquire, align 1
+// CHECK-NEXT:    store i8 [[TMP627]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP628:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP629:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP628]] acquire, align 1
+// CHECK-NEXT:    store i8 [[TMP629]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP630:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP631:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP630]] acquire, align 1
+// CHECK-NEXT:    store i8 [[TMP631]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP632:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP633:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP634:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP632]], i8 [[TMP633]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP635:%.*]] = extractvalue { i8, i1 } [[TMP634]], 0
+// CHECK-NEXT:    store i8 [[TMP635]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP636:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP637:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP638:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP636]], i8 [[TMP637]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP639:%.*]] = extractvalue { i8, i1 } [[TMP638]], 0
+// CHECK-NEXT:    store i8 [[TMP639]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP640:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP641:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP640]] acquire, align 1
+// CHECK-NEXT:    [[TMP642:%.*]] = icmp ugt i8 [[TMP641]], [[TMP640]]
+// CHECK-NEXT:    [[TMP643:%.*]] = select i1 [[TMP642]], i8 [[TMP640]], i8 [[TMP641]]
+// CHECK-NEXT:    store i8 [[TMP643]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP644:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP645:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP644]] acquire, align 1
+// CHECK-NEXT:    [[TMP646:%.*]] = icmp ult i8 [[TMP645]], [[TMP644]]
+// CHECK-NEXT:    [[TMP647:%.*]] = select i1 [[TMP646]], i8 [[TMP644]], i8 [[TMP645]]
+// CHECK-NEXT:    store i8 [[TMP647]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP648:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP649:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP648]] acquire, align 1
+// CHECK-NEXT:    [[TMP650:%.*]] = icmp ult i8 [[TMP649]], [[TMP648]]
+// CHECK-NEXT:    [[TMP651:%.*]] = select i1 [[TMP650]], i8 [[TMP648]], i8 [[TMP649]]
+// CHECK-NEXT:    store i8 [[TMP651]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP652:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP653:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP652]] acquire, align 1
+// CHECK-NEXT:    [[TMP654:%.*]] = icmp ugt i8 [[TMP653]], [[TMP652]]
+// CHECK-NEXT:    [[TMP655:%.*]] = select i1 [[TMP654]], i8 [[TMP652]], i8 [[TMP653]]
+// CHECK-NEXT:    store i8 [[TMP655]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP656:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP657:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP658:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP656]], i8 [[TMP657]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP659:%.*]] = extractvalue { i8, i1 } [[TMP658]], 0
+// CHECK-NEXT:    [[TMP660:%.*]] = extractvalue { i8, i1 } [[TMP658]], 1
+// CHECK-NEXT:    [[TMP661:%.*]] = select i1 [[TMP660]], i8 [[TMP656]], i8 [[TMP659]]
+// CHECK-NEXT:    store i8 [[TMP661]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP662:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP663:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP664:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP662]], i8 [[TMP663]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP665:%.*]] = extractvalue { i8, i1 } [[TMP664]], 0
+// CHECK-NEXT:    [[TMP666:%.*]] = extractvalue { i8, i1 } [[TMP664]], 1
+// CHECK-NEXT:    [[TMP667:%.*]] = select i1 [[TMP666]], i8 [[TMP662]], i8 [[TMP665]]
+// CHECK-NEXT:    store i8 [[TMP667]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP668:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP669:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP670:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP668]], i8 [[TMP669]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP671:%.*]] = extractvalue { i8, i1 } [[TMP670]], 0
+// CHECK-NEXT:    [[TMP672:%.*]] = extractvalue { i8, i1 } [[TMP670]], 1
+// CHECK-NEXT:    br i1 [[TMP672]], label [[UCX_ATOMIC_EXIT61:%.*]], label [[UCX_ATOMIC_CONT62:%.*]]
+// CHECK:       ucx.atomic.cont62:
+// CHECK-NEXT:    store i8 [[TMP671]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT61]]
+// CHECK:       ucx.atomic.exit61:
+// CHECK-NEXT:    [[TMP673:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP674:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP675:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP673]], i8 [[TMP674]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP676:%.*]] = extractvalue { i8, i1 } [[TMP675]], 0
+// CHECK-NEXT:    [[TMP677:%.*]] = extractvalue { i8, i1 } [[TMP675]], 1
+// CHECK-NEXT:    br i1 [[TMP677]], label [[UCX_ATOMIC_EXIT63:%.*]], label [[UCX_ATOMIC_CONT64:%.*]]
+// CHECK:       ucx.atomic.cont64:
+// CHECK-NEXT:    store i8 [[TMP676]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT63]]
+// CHECK:       ucx.atomic.exit63:
+// CHECK-NEXT:    [[TMP678:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP679:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP680:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP678]], i8 [[TMP679]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP681:%.*]] = extractvalue { i8, i1 } [[TMP680]], 1
+// CHECK-NEXT:    [[TMP682:%.*]] = zext i1 [[TMP681]] to i8
+// CHECK-NEXT:    store i8 [[TMP682]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP683:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP684:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP685:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP683]], i8 [[TMP684]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP686:%.*]] = extractvalue { i8, i1 } [[TMP685]], 1
+// CHECK-NEXT:    [[TMP687:%.*]] = zext i1 [[TMP686]] to i8
+// CHECK-NEXT:    store i8 [[TMP687]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP688:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP689:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP690:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP688]], i8 [[TMP689]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP691:%.*]] = extractvalue { i8, i1 } [[TMP690]], 0
+// CHECK-NEXT:    [[TMP692:%.*]] = extractvalue { i8, i1 } [[TMP690]], 1
+// CHECK-NEXT:    br i1 [[TMP692]], label [[UCX_ATOMIC_EXIT65:%.*]], label [[UCX_ATOMIC_CONT66:%.*]]
+// CHECK:       ucx.atomic.cont66:
+// CHECK-NEXT:    store i8 [[TMP691]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT65]]
+// CHECK:       ucx.atomic.exit65:
+// CHECK-NEXT:    [[TMP693:%.*]] = extractvalue { i8, i1 } [[TMP690]], 1
+// CHECK-NEXT:    [[TMP694:%.*]] = zext i1 [[TMP693]] to i8
+// CHECK-NEXT:    store i8 [[TMP694]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP695:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP696:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP697:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP695]], i8 [[TMP696]] acquire acquire, align 1
+// CHECK-NEXT:    [[TMP698:%.*]] = extractvalue { i8, i1 } [[TMP697]], 0
+// CHECK-NEXT:    [[TMP699:%.*]] = extractvalue { i8, i1 } [[TMP697]], 1
+// CHECK-NEXT:    br i1 [[TMP699]], label [[UCX_ATOMIC_EXIT67:%.*]], label [[UCX_ATOMIC_CONT68:%.*]]
+// CHECK:       ucx.atomic.cont68:
+// CHECK-NEXT:    store i8 [[TMP698]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT67]]
+// CHECK:       ucx.atomic.exit67:
+// CHECK-NEXT:    [[TMP700:%.*]] = extractvalue { i8, i1 } [[TMP697]], 1
+// CHECK-NEXT:    [[TMP701:%.*]] = zext i1 [[TMP700]] to i8
+// CHECK-NEXT:    store i8 [[TMP701]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP702:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP703:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP702]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP703]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP704:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP705:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP704]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP705]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP706:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP707:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP706]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP707]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP708:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP709:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP708]] monotonic, align 1
+// CHECK-NEXT:    store i8 [[TMP709]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP710:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP711:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP712:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP710]], i8 [[TMP711]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP713:%.*]] = extractvalue { i8, i1 } [[TMP712]], 0
+// CHECK-NEXT:    store i8 [[TMP713]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP714:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP715:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP716:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP714]], i8 [[TMP715]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP717:%.*]] = extractvalue { i8, i1 } [[TMP716]], 0
+// CHECK-NEXT:    store i8 [[TMP717]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP718:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP719:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP718]] monotonic, align 1
+// CHECK-NEXT:    [[TMP720:%.*]] = icmp ugt i8 [[TMP719]], [[TMP718]]
+// CHECK-NEXT:    [[TMP721:%.*]] = select i1 [[TMP720]], i8 [[TMP718]], i8 [[TMP719]]
+// CHECK-NEXT:    store i8 [[TMP721]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP722:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP723:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP722]] monotonic, align 1
+// CHECK-NEXT:    [[TMP724:%.*]] = icmp ult i8 [[TMP723]], [[TMP722]]
+// CHECK-NEXT:    [[TMP725:%.*]] = select i1 [[TMP724]], i8 [[TMP722]], i8 [[TMP723]]
+// CHECK-NEXT:    store i8 [[TMP725]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP726:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP727:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP726]] monotonic, align 1
+// CHECK-NEXT:    [[TMP728:%.*]] = icmp ult i8 [[TMP727]], [[TMP726]]
+// CHECK-NEXT:    [[TMP729:%.*]] = select i1 [[TMP728]], i8 [[TMP726]], i8 [[TMP727]]
+// CHECK-NEXT:    store i8 [[TMP729]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP730:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP731:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP730]] monotonic, align 1
+// CHECK-NEXT:    [[TMP732:%.*]] = icmp ugt i8 [[TMP731]], [[TMP730]]
+// CHECK-NEXT:    [[TMP733:%.*]] = select i1 [[TMP732]], i8 [[TMP730]], i8 [[TMP731]]
+// CHECK-NEXT:    store i8 [[TMP733]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP734:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP735:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP736:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP734]], i8 [[TMP735]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP737:%.*]] = extractvalue { i8, i1 } [[TMP736]], 0
+// CHECK-NEXT:    [[TMP738:%.*]] = extractvalue { i8, i1 } [[TMP736]], 1
+// CHECK-NEXT:    [[TMP739:%.*]] = select i1 [[TMP738]], i8 [[TMP734]], i8 [[TMP737]]
+// CHECK-NEXT:    store i8 [[TMP739]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP740:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP741:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP742:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP740]], i8 [[TMP741]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP743:%.*]] = extractvalue { i8, i1 } [[TMP742]], 0
+// CHECK-NEXT:    [[TMP744:%.*]] = extractvalue { i8, i1 } [[TMP742]], 1
+// CHECK-NEXT:    [[TMP745:%.*]] = select i1 [[TMP744]], i8 [[TMP740]], i8 [[TMP743]]
+// CHECK-NEXT:    store i8 [[TMP745]], i8* [[UCV]], align 1
+// CHECK-NEXT:    [[TMP746:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP747:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP748:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP746]], i8 [[TMP747]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP749:%.*]] = extractvalue { i8, i1 } [[TMP748]], 0
+// CHECK-NEXT:    [[TMP750:%.*]] = extractvalue { i8, i1 } [[TMP748]], 1
+// CHECK-NEXT:    br i1 [[TMP750]], label [[UCX_ATOMIC_EXIT69:%.*]], label [[UCX_ATOMIC_CONT70:%.*]]
+// CHECK:       ucx.atomic.cont70:
+// CHECK-NEXT:    store i8 [[TMP749]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT69]]
+// CHECK:       ucx.atomic.exit69:
+// CHECK-NEXT:    [[TMP751:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP752:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP753:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP751]], i8 [[TMP752]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP754:%.*]] = extractvalue { i8, i1 } [[TMP753]], 0
+// CHECK-NEXT:    [[TMP755:%.*]] = extractvalue { i8, i1 } [[TMP753]], 1
+// CHECK-NEXT:    br i1 [[TMP755]], label [[UCX_ATOMIC_EXIT71:%.*]], label [[UCX_ATOMIC_CONT72:%.*]]
+// CHECK:       ucx.atomic.cont72:
+// CHECK-NEXT:    store i8 [[TMP754]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT71]]
+// CHECK:       ucx.atomic.exit71:
+// CHECK-NEXT:    [[TMP756:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP757:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP758:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP756]], i8 [[TMP757]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP759:%.*]] = extractvalue { i8, i1 } [[TMP758]], 1
+// CHECK-NEXT:    [[TMP760:%.*]] = zext i1 [[TMP759]] to i8
+// CHECK-NEXT:    store i8 [[TMP760]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP761:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP762:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP763:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP761]], i8 [[TMP762]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP764:%.*]] = extractvalue { i8, i1 } [[TMP763]], 1
+// CHECK-NEXT:    [[TMP765:%.*]] = zext i1 [[TMP764]] to i8
+// CHECK-NEXT:    store i8 [[TMP765]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP766:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP767:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP768:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP766]], i8 [[TMP767]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP769:%.*]] = extractvalue { i8, i1 } [[TMP768]], 0
+// CHECK-NEXT:    [[TMP770:%.*]] = extractvalue { i8, i1 } [[TMP768]], 1
+// CHECK-NEXT:    br i1 [[TMP770]], label [[UCX_ATOMIC_EXIT73:%.*]], label [[UCX_ATOMIC_CONT74:%.*]]
+// CHECK:       ucx.atomic.cont74:
+// CHECK-NEXT:    store i8 [[TMP769]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT73]]
+// CHECK:       ucx.atomic.exit73:
+// CHECK-NEXT:    [[TMP771:%.*]] = extractvalue { i8, i1 } [[TMP768]], 1
+// CHECK-NEXT:    [[TMP772:%.*]] = zext i1 [[TMP771]] to i8
+// CHECK-NEXT:    store i8 [[TMP772]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP773:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP774:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP775:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP773]], i8 [[TMP774]] monotonic monotonic, align 1
+// CHECK-NEXT:    [[TMP776:%.*]] = extractvalue { i8, i1 } [[TMP775]], 0
+// CHECK-NEXT:    [[TMP777:%.*]] = extractvalue { i8, i1 } [[TMP775]], 1
+// CHECK-NEXT:    br i1 [[TMP777]], label [[UCX_ATOMIC_EXIT75:%.*]], label [[UCX_ATOMIC_CONT76:%.*]]
+// CHECK:       ucx.atomic.cont76:
+// CHECK-NEXT:    store i8 [[TMP776]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT75]]
+// CHECK:       ucx.atomic.exit75:
+// CHECK-NEXT:    [[TMP778:%.*]] = extractvalue { i8, i1 } [[TMP775]], 1
+// CHECK-NEXT:    [[TMP779:%.*]] = zext i1 [[TMP778]] to i8
+// CHECK-NEXT:    store i8 [[TMP779]], i8* [[UCR]], align 1
+// CHECK-NEXT:    [[TMP780:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP781:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP780]] release, align 1
+// CHECK-NEXT:    store i8 [[TMP781]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP782:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP783:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP782]] release, align 1
+// CHECK-NEXT:    store i8 [[TMP783]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP784:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP785:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP784]] release, align 1
+// CHECK-NEXT:    store i8 [[TMP785]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP786:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP787:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP786]] release, align 1
+// CHECK-NEXT:    store i8 [[TMP787]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP788:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP789:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP790:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP788]], i8 [[TMP789]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP791:%.*]] = extractvalue { i8, i1 } [[TMP790]], 0
+// CHECK-NEXT:    store i8 [[TMP791]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP792:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP793:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP794:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP792]], i8 [[TMP793]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP795:%.*]] = extractvalue { i8, i1 } [[TMP794]], 0
+// CHECK-NEXT:    store i8 [[TMP795]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP796:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP797:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP796]] release, align 1
+// CHECK-NEXT:    [[TMP798:%.*]] = icmp ugt i8 [[TMP797]], [[TMP796]]
+// CHECK-NEXT:    [[TMP799:%.*]] = select i1 [[TMP798]], i8 [[TMP796]], i8 [[TMP797]]
+// CHECK-NEXT:    store i8 [[TMP799]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP800:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP801:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP800]] release, align 1
+// CHECK-NEXT:    [[TMP802:%.*]] = icmp ult i8 [[TMP801]], [[TMP800]]
+// CHECK-NEXT:    [[TMP803:%.*]] = select i1 [[TMP802]], i8 [[TMP800]], i8 [[TMP801]]
+// CHECK-NEXT:    store i8 [[TMP803]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP804:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP805:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP804]] release, align 1
+// CHECK-NEXT:    [[TMP806:%.*]] = icmp ult i8 [[TMP805]], [[TMP804]]
+// CHECK-NEXT:    [[TMP807:%.*]] = select i1 [[TMP806]], i8 [[TMP804]], i8 [[TMP805]]
+// CHECK-NEXT:    store i8 [[TMP807]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP808:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP809:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP808]] release, align 1
+// CHECK-NEXT:    [[TMP810:%.*]] = icmp ugt i8 [[TMP809]], [[TMP808]]
+// CHECK-NEXT:    [[TMP811:%.*]] = select i1 [[TMP810]], i8 [[TMP808]], i8 [[TMP809]]
+// CHECK-NEXT:    store i8 [[TMP811]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP812:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP813:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP814:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP812]], i8 [[TMP813]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP815:%.*]] = extractvalue { i8, i1 } [[TMP814]], 0
+// CHECK-NEXT:    [[TMP816:%.*]] = extractvalue { i8, i1 } [[TMP814]], 1
+// CHECK-NEXT:    [[TMP817:%.*]] = select i1 [[TMP816]], i8 [[TMP812]], i8 [[TMP815]]
+// CHECK-NEXT:    store i8 [[TMP817]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP818:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP819:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP820:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP818]], i8 [[TMP819]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP821:%.*]] = extractvalue { i8, i1 } [[TMP820]], 0
+// CHECK-NEXT:    [[TMP822:%.*]] = extractvalue { i8, i1 } [[TMP820]], 1
+// CHECK-NEXT:    [[TMP823:%.*]] = select i1 [[TMP822]], i8 [[TMP818]], i8 [[TMP821]]
+// CHECK-NEXT:    store i8 [[TMP823]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP824:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP825:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP826:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP824]], i8 [[TMP825]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP827:%.*]] = extractvalue { i8, i1 } [[TMP826]], 0
+// CHECK-NEXT:    [[TMP828:%.*]] = extractvalue { i8, i1 } [[TMP826]], 1
+// CHECK-NEXT:    br i1 [[TMP828]], label [[UCX_ATOMIC_EXIT77:%.*]], label [[UCX_ATOMIC_CONT78:%.*]]
+// CHECK:       ucx.atomic.cont78:
+// CHECK-NEXT:    store i8 [[TMP827]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT77]]
+// CHECK:       ucx.atomic.exit77:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP829:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP830:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP831:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP829]], i8 [[TMP830]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP832:%.*]] = extractvalue { i8, i1 } [[TMP831]], 0
+// CHECK-NEXT:    [[TMP833:%.*]] = extractvalue { i8, i1 } [[TMP831]], 1
+// CHECK-NEXT:    br i1 [[TMP833]], label [[UCX_ATOMIC_EXIT79:%.*]], label [[UCX_ATOMIC_CONT80:%.*]]
+// CHECK:       ucx.atomic.cont80:
+// CHECK-NEXT:    store i8 [[TMP832]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT79]]
+// CHECK:       ucx.atomic.exit79:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP834:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP835:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP836:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP834]], i8 [[TMP835]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP837:%.*]] = extractvalue { i8, i1 } [[TMP836]], 1
+// CHECK-NEXT:    [[TMP838:%.*]] = zext i1 [[TMP837]] to i8
+// CHECK-NEXT:    store i8 [[TMP838]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP839:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP840:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP841:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP839]], i8 [[TMP840]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP842:%.*]] = extractvalue { i8, i1 } [[TMP841]], 1
+// CHECK-NEXT:    [[TMP843:%.*]] = zext i1 [[TMP842]] to i8
+// CHECK-NEXT:    store i8 [[TMP843]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP844:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP845:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP846:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP844]], i8 [[TMP845]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP847:%.*]] = extractvalue { i8, i1 } [[TMP846]], 0
+// CHECK-NEXT:    [[TMP848:%.*]] = extractvalue { i8, i1 } [[TMP846]], 1
+// CHECK-NEXT:    br i1 [[TMP848]], label [[UCX_ATOMIC_EXIT81:%.*]], label [[UCX_ATOMIC_CONT82:%.*]]
+// CHECK:       ucx.atomic.cont82:
+// CHECK-NEXT:    store i8 [[TMP847]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT81]]
+// CHECK:       ucx.atomic.exit81:
+// CHECK-NEXT:    [[TMP849:%.*]] = extractvalue { i8, i1 } [[TMP846]], 1
+// CHECK-NEXT:    [[TMP850:%.*]] = zext i1 [[TMP849]] to i8
+// CHECK-NEXT:    store i8 [[TMP850]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP851:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP852:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP853:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP851]], i8 [[TMP852]] release monotonic, align 1
+// CHECK-NEXT:    [[TMP854:%.*]] = extractvalue { i8, i1 } [[TMP853]], 0
+// CHECK-NEXT:    [[TMP855:%.*]] = extractvalue { i8, i1 } [[TMP853]], 1
+// CHECK-NEXT:    br i1 [[TMP855]], label [[UCX_ATOMIC_EXIT83:%.*]], label [[UCX_ATOMIC_CONT84:%.*]]
+// CHECK:       ucx.atomic.cont84:
+// CHECK-NEXT:    store i8 [[TMP854]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT83]]
+// CHECK:       ucx.atomic.exit83:
+// CHECK-NEXT:    [[TMP856:%.*]] = extractvalue { i8, i1 } [[TMP853]], 1
+// CHECK-NEXT:    [[TMP857:%.*]] = zext i1 [[TMP856]] to i8
+// CHECK-NEXT:    store i8 [[TMP857]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP858:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP859:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP858]] seq_cst, align 1
+// CHECK-NEXT:    store i8 [[TMP859]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP860:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP861:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP860]] seq_cst, align 1
+// CHECK-NEXT:    store i8 [[TMP861]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP862:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP863:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP862]] seq_cst, align 1
+// CHECK-NEXT:    store i8 [[TMP863]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP864:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP865:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP864]] seq_cst, align 1
+// CHECK-NEXT:    store i8 [[TMP865]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP866:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP867:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP868:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP866]], i8 [[TMP867]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP869:%.*]] = extractvalue { i8, i1 } [[TMP868]], 0
+// CHECK-NEXT:    store i8 [[TMP869]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP870:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP871:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP872:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP870]], i8 [[TMP871]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP873:%.*]] = extractvalue { i8, i1 } [[TMP872]], 0
+// CHECK-NEXT:    store i8 [[TMP873]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP874:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP875:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP874]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP876:%.*]] = icmp ugt i8 [[TMP875]], [[TMP874]]
+// CHECK-NEXT:    [[TMP877:%.*]] = select i1 [[TMP876]], i8 [[TMP874]], i8 [[TMP875]]
+// CHECK-NEXT:    store i8 [[TMP877]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP878:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP879:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP878]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP880:%.*]] = icmp ult i8 [[TMP879]], [[TMP878]]
+// CHECK-NEXT:    [[TMP881:%.*]] = select i1 [[TMP880]], i8 [[TMP878]], i8 [[TMP879]]
+// CHECK-NEXT:    store i8 [[TMP881]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP882:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP883:%.*]] = atomicrmw umin i8* [[UCX]], i8 [[TMP882]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP884:%.*]] = icmp ult i8 [[TMP883]], [[TMP882]]
+// CHECK-NEXT:    [[TMP885:%.*]] = select i1 [[TMP884]], i8 [[TMP882]], i8 [[TMP883]]
+// CHECK-NEXT:    store i8 [[TMP885]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP886:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP887:%.*]] = atomicrmw umax i8* [[UCX]], i8 [[TMP886]] seq_cst, align 1
+// CHECK-NEXT:    [[TMP888:%.*]] = icmp ugt i8 [[TMP887]], [[TMP886]]
+// CHECK-NEXT:    [[TMP889:%.*]] = select i1 [[TMP888]], i8 [[TMP886]], i8 [[TMP887]]
+// CHECK-NEXT:    store i8 [[TMP889]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP890:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP891:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP892:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP890]], i8 [[TMP891]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP893:%.*]] = extractvalue { i8, i1 } [[TMP892]], 0
+// CHECK-NEXT:    [[TMP894:%.*]] = extractvalue { i8, i1 } [[TMP892]], 1
+// CHECK-NEXT:    [[TMP895:%.*]] = select i1 [[TMP894]], i8 [[TMP890]], i8 [[TMP893]]
+// CHECK-NEXT:    store i8 [[TMP895]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP896:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP897:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP898:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP896]], i8 [[TMP897]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP899:%.*]] = extractvalue { i8, i1 } [[TMP898]], 0
+// CHECK-NEXT:    [[TMP900:%.*]] = extractvalue { i8, i1 } [[TMP898]], 1
+// CHECK-NEXT:    [[TMP901:%.*]] = select i1 [[TMP900]], i8 [[TMP896]], i8 [[TMP899]]
+// CHECK-NEXT:    store i8 [[TMP901]], i8* [[UCV]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP902:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP903:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP904:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP902]], i8 [[TMP903]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP905:%.*]] = extractvalue { i8, i1 } [[TMP904]], 0
+// CHECK-NEXT:    [[TMP906:%.*]] = extractvalue { i8, i1 } [[TMP904]], 1
+// CHECK-NEXT:    br i1 [[TMP906]], label [[UCX_ATOMIC_EXIT85:%.*]], label [[UCX_ATOMIC_CONT86:%.*]]
+// CHECK:       ucx.atomic.cont86:
+// CHECK-NEXT:    store i8 [[TMP905]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT85]]
+// CHECK:       ucx.atomic.exit85:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP907:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP908:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP909:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP907]], i8 [[TMP908]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP910:%.*]] = extractvalue { i8, i1 } [[TMP909]], 0
+// CHECK-NEXT:    [[TMP911:%.*]] = extractvalue { i8, i1 } [[TMP909]], 1
+// CHECK-NEXT:    br i1 [[TMP911]], label [[UCX_ATOMIC_EXIT87:%.*]], label [[UCX_ATOMIC_CONT88:%.*]]
+// CHECK:       ucx.atomic.cont88:
+// CHECK-NEXT:    store i8 [[TMP910]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT87]]
+// CHECK:       ucx.atomic.exit87:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP912:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP913:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP914:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP912]], i8 [[TMP913]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP915:%.*]] = extractvalue { i8, i1 } [[TMP914]], 1
+// CHECK-NEXT:    [[TMP916:%.*]] = zext i1 [[TMP915]] to i8
+// CHECK-NEXT:    store i8 [[TMP916]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP917:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP918:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP919:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP917]], i8 [[TMP918]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP920:%.*]] = extractvalue { i8, i1 } [[TMP919]], 1
+// CHECK-NEXT:    [[TMP921:%.*]] = zext i1 [[TMP920]] to i8
+// CHECK-NEXT:    store i8 [[TMP921]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP922:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP923:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP924:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP922]], i8 [[TMP923]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP925:%.*]] = extractvalue { i8, i1 } [[TMP924]], 0
+// CHECK-NEXT:    [[TMP926:%.*]] = extractvalue { i8, i1 } [[TMP924]], 1
+// CHECK-NEXT:    br i1 [[TMP926]], label [[UCX_ATOMIC_EXIT89:%.*]], label [[UCX_ATOMIC_CONT90:%.*]]
+// CHECK:       ucx.atomic.cont90:
+// CHECK-NEXT:    store i8 [[TMP925]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT89]]
+// CHECK:       ucx.atomic.exit89:
+// CHECK-NEXT:    [[TMP927:%.*]] = extractvalue { i8, i1 } [[TMP924]], 1
+// CHECK-NEXT:    [[TMP928:%.*]] = zext i1 [[TMP927]] to i8
+// CHECK-NEXT:    store i8 [[TMP928]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP929:%.*]] = load i8, i8* [[UCE]], align 1
+// CHECK-NEXT:    [[TMP930:%.*]] = load i8, i8* [[UCD]], align 1
+// CHECK-NEXT:    [[TMP931:%.*]] = cmpxchg i8* [[UCX]], i8 [[TMP929]], i8 [[TMP930]] seq_cst seq_cst, align 1
+// CHECK-NEXT:    [[TMP932:%.*]] = extractvalue { i8, i1 } [[TMP931]], 0
+// CHECK-NEXT:    [[TMP933:%.*]] = extractvalue { i8, i1 } [[TMP931]], 1
+// CHECK-NEXT:    br i1 [[TMP933]], label [[UCX_ATOMIC_EXIT91:%.*]], label [[UCX_ATOMIC_CONT92:%.*]]
+// CHECK:       ucx.atomic.cont92:
+// CHECK-NEXT:    store i8 [[TMP932]], i8* [[UCV]], align 1
+// CHECK-NEXT:    br label [[UCX_ATOMIC_EXIT91]]
+// CHECK:       ucx.atomic.exit91:
+// CHECK-NEXT:    [[TMP934:%.*]] = extractvalue { i8, i1 } [[TMP931]], 1
+// CHECK-NEXT:    [[TMP935:%.*]] = zext i1 [[TMP934]] to i8
+// CHECK-NEXT:    store i8 [[TMP935]], i8* [[UCR]], align 1
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP936:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP937:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP936]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP937]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP938:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP939:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP938]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP939]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP940:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP941:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP940]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP941]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP942:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP943:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP942]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP943]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP944:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP945:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP946:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP944]], i16 [[TMP945]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP947:%.*]] = extractvalue { i16, i1 } [[TMP946]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP947]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP948:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP949:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP950:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP948]], i16 [[TMP949]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP951:%.*]] = extractvalue { i16, i1 } [[TMP950]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP951]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP952:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP953:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP952]] monotonic, align 2
+// CHECK-NEXT:    [[TMP954:%.*]] = icmp ugt i16 [[TMP953]], [[TMP952]]
+// CHECK-NEXT:    [[TMP955:%.*]] = select i1 [[TMP954]], i16 [[TMP952]], i16 [[TMP953]]
+// CHECK-NEXT:    store volatile i16 [[TMP955]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP956:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP957:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP956]] monotonic, align 2
+// CHECK-NEXT:    [[TMP958:%.*]] = icmp ult i16 [[TMP957]], [[TMP956]]
+// CHECK-NEXT:    [[TMP959:%.*]] = select i1 [[TMP958]], i16 [[TMP956]], i16 [[TMP957]]
+// CHECK-NEXT:    store volatile i16 [[TMP959]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP960:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP961:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP960]] monotonic, align 2
+// CHECK-NEXT:    [[TMP962:%.*]] = icmp ult i16 [[TMP961]], [[TMP960]]
+// CHECK-NEXT:    [[TMP963:%.*]] = select i1 [[TMP962]], i16 [[TMP960]], i16 [[TMP961]]
+// CHECK-NEXT:    store volatile i16 [[TMP963]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP964:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP965:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP964]] monotonic, align 2
+// CHECK-NEXT:    [[TMP966:%.*]] = icmp ugt i16 [[TMP965]], [[TMP964]]
+// CHECK-NEXT:    [[TMP967:%.*]] = select i1 [[TMP966]], i16 [[TMP964]], i16 [[TMP965]]
+// CHECK-NEXT:    store volatile i16 [[TMP967]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP968:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP969:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP970:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP968]], i16 [[TMP969]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP971:%.*]] = extractvalue { i16, i1 } [[TMP970]], 0
+// CHECK-NEXT:    [[TMP972:%.*]] = extractvalue { i16, i1 } [[TMP970]], 1
+// CHECK-NEXT:    [[TMP973:%.*]] = select i1 [[TMP972]], i16 [[TMP968]], i16 [[TMP971]]
+// CHECK-NEXT:    store volatile i16 [[TMP973]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP974:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP975:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP976:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP974]], i16 [[TMP975]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP977:%.*]] = extractvalue { i16, i1 } [[TMP976]], 0
+// CHECK-NEXT:    [[TMP978:%.*]] = extractvalue { i16, i1 } [[TMP976]], 1
+// CHECK-NEXT:    [[TMP979:%.*]] = select i1 [[TMP978]], i16 [[TMP974]], i16 [[TMP977]]
+// CHECK-NEXT:    store volatile i16 [[TMP979]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP980:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP981:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP982:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP980]], i16 [[TMP981]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP983:%.*]] = extractvalue { i16, i1 } [[TMP982]], 0
+// CHECK-NEXT:    [[TMP984:%.*]] = extractvalue { i16, i1 } [[TMP982]], 1
+// CHECK-NEXT:    br i1 [[TMP984]], label [[SX_ATOMIC_EXIT:%.*]], label [[SX_ATOMIC_CONT:%.*]]
+// CHECK:       sx.atomic.cont:
+// CHECK-NEXT:    store i16 [[TMP983]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT]]
+// CHECK:       sx.atomic.exit:
+// CHECK-NEXT:    [[TMP985:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP986:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP987:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP985]], i16 [[TMP986]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP988:%.*]] = extractvalue { i16, i1 } [[TMP987]], 0
+// CHECK-NEXT:    [[TMP989:%.*]] = extractvalue { i16, i1 } [[TMP987]], 1
+// CHECK-NEXT:    br i1 [[TMP989]], label [[SX_ATOMIC_EXIT93:%.*]], label [[SX_ATOMIC_CONT94:%.*]]
+// CHECK:       sx.atomic.cont94:
+// CHECK-NEXT:    store i16 [[TMP988]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT93]]
+// CHECK:       sx.atomic.exit93:
+// CHECK-NEXT:    [[TMP990:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP991:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP992:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP990]], i16 [[TMP991]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP993:%.*]] = extractvalue { i16, i1 } [[TMP992]], 1
+// CHECK-NEXT:    [[TMP994:%.*]] = zext i1 [[TMP993]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP994]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP995:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP996:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP997:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP995]], i16 [[TMP996]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP998:%.*]] = extractvalue { i16, i1 } [[TMP997]], 1
+// CHECK-NEXT:    [[TMP999:%.*]] = zext i1 [[TMP998]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP999]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1000:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1001:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1002:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1000]], i16 [[TMP1001]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1003:%.*]] = extractvalue { i16, i1 } [[TMP1002]], 0
+// CHECK-NEXT:    [[TMP1004:%.*]] = extractvalue { i16, i1 } [[TMP1002]], 1
+// CHECK-NEXT:    br i1 [[TMP1004]], label [[SX_ATOMIC_EXIT95:%.*]], label [[SX_ATOMIC_CONT96:%.*]]
+// CHECK:       sx.atomic.cont96:
+// CHECK-NEXT:    store i16 [[TMP1003]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT95]]
+// CHECK:       sx.atomic.exit95:
+// CHECK-NEXT:    [[TMP1005:%.*]] = extractvalue { i16, i1 } [[TMP1002]], 1
+// CHECK-NEXT:    [[TMP1006:%.*]] = zext i1 [[TMP1005]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1006]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1007:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1008:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1009:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1007]], i16 [[TMP1008]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1010:%.*]] = extractvalue { i16, i1 } [[TMP1009]], 0
+// CHECK-NEXT:    [[TMP1011:%.*]] = extractvalue { i16, i1 } [[TMP1009]], 1
+// CHECK-NEXT:    br i1 [[TMP1011]], label [[SX_ATOMIC_EXIT97:%.*]], label [[SX_ATOMIC_CONT98:%.*]]
+// CHECK:       sx.atomic.cont98:
+// CHECK-NEXT:    store i16 [[TMP1010]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT97]]
+// CHECK:       sx.atomic.exit97:
+// CHECK-NEXT:    [[TMP1012:%.*]] = extractvalue { i16, i1 } [[TMP1009]], 1
+// CHECK-NEXT:    [[TMP1013:%.*]] = zext i1 [[TMP1012]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1013]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1014:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1015:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1014]] acq_rel, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1015]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1016:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1017:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1016]] acq_rel, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1017]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1018:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1019:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1018]] acq_rel, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1019]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1020:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1021:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1020]] acq_rel, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1021]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1022:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1023:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1024:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1022]], i16 [[TMP1023]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1025:%.*]] = extractvalue { i16, i1 } [[TMP1024]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1025]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1026:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1027:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1028:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1026]], i16 [[TMP1027]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1029:%.*]] = extractvalue { i16, i1 } [[TMP1028]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1029]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1030:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1031:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1030]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1032:%.*]] = icmp ugt i16 [[TMP1031]], [[TMP1030]]
+// CHECK-NEXT:    [[TMP1033:%.*]] = select i1 [[TMP1032]], i16 [[TMP1030]], i16 [[TMP1031]]
+// CHECK-NEXT:    store volatile i16 [[TMP1033]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1034:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1035:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1034]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1036:%.*]] = icmp ult i16 [[TMP1035]], [[TMP1034]]
+// CHECK-NEXT:    [[TMP1037:%.*]] = select i1 [[TMP1036]], i16 [[TMP1034]], i16 [[TMP1035]]
+// CHECK-NEXT:    store volatile i16 [[TMP1037]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1038:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1039:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1038]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1040:%.*]] = icmp ult i16 [[TMP1039]], [[TMP1038]]
+// CHECK-NEXT:    [[TMP1041:%.*]] = select i1 [[TMP1040]], i16 [[TMP1038]], i16 [[TMP1039]]
+// CHECK-NEXT:    store volatile i16 [[TMP1041]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1042:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1043:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1042]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1044:%.*]] = icmp ugt i16 [[TMP1043]], [[TMP1042]]
+// CHECK-NEXT:    [[TMP1045:%.*]] = select i1 [[TMP1044]], i16 [[TMP1042]], i16 [[TMP1043]]
+// CHECK-NEXT:    store volatile i16 [[TMP1045]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1046:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1047:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1048:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1046]], i16 [[TMP1047]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1049:%.*]] = extractvalue { i16, i1 } [[TMP1048]], 0
+// CHECK-NEXT:    [[TMP1050:%.*]] = extractvalue { i16, i1 } [[TMP1048]], 1
+// CHECK-NEXT:    [[TMP1051:%.*]] = select i1 [[TMP1050]], i16 [[TMP1046]], i16 [[TMP1049]]
+// CHECK-NEXT:    store volatile i16 [[TMP1051]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1052:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1053:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1054:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1052]], i16 [[TMP1053]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1055:%.*]] = extractvalue { i16, i1 } [[TMP1054]], 0
+// CHECK-NEXT:    [[TMP1056:%.*]] = extractvalue { i16, i1 } [[TMP1054]], 1
+// CHECK-NEXT:    [[TMP1057:%.*]] = select i1 [[TMP1056]], i16 [[TMP1052]], i16 [[TMP1055]]
+// CHECK-NEXT:    store volatile i16 [[TMP1057]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1058:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1059:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1060:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1058]], i16 [[TMP1059]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1061:%.*]] = extractvalue { i16, i1 } [[TMP1060]], 0
+// CHECK-NEXT:    [[TMP1062:%.*]] = extractvalue { i16, i1 } [[TMP1060]], 1
+// CHECK-NEXT:    br i1 [[TMP1062]], label [[SX_ATOMIC_EXIT99:%.*]], label [[SX_ATOMIC_CONT100:%.*]]
+// CHECK:       sx.atomic.cont100:
+// CHECK-NEXT:    store i16 [[TMP1061]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT99]]
+// CHECK:       sx.atomic.exit99:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1063:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1064:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1065:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1063]], i16 [[TMP1064]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1066:%.*]] = extractvalue { i16, i1 } [[TMP1065]], 0
+// CHECK-NEXT:    [[TMP1067:%.*]] = extractvalue { i16, i1 } [[TMP1065]], 1
+// CHECK-NEXT:    br i1 [[TMP1067]], label [[SX_ATOMIC_EXIT101:%.*]], label [[SX_ATOMIC_CONT102:%.*]]
+// CHECK:       sx.atomic.cont102:
+// CHECK-NEXT:    store i16 [[TMP1066]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT101]]
+// CHECK:       sx.atomic.exit101:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1068:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1069:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1070:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1068]], i16 [[TMP1069]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1071:%.*]] = extractvalue { i16, i1 } [[TMP1070]], 1
+// CHECK-NEXT:    [[TMP1072:%.*]] = zext i1 [[TMP1071]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1072]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1073:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1074:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1075:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1073]], i16 [[TMP1074]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1076:%.*]] = extractvalue { i16, i1 } [[TMP1075]], 1
+// CHECK-NEXT:    [[TMP1077:%.*]] = zext i1 [[TMP1076]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1077]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1078:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1079:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1080:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1078]], i16 [[TMP1079]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1081:%.*]] = extractvalue { i16, i1 } [[TMP1080]], 0
+// CHECK-NEXT:    [[TMP1082:%.*]] = extractvalue { i16, i1 } [[TMP1080]], 1
+// CHECK-NEXT:    br i1 [[TMP1082]], label [[SX_ATOMIC_EXIT103:%.*]], label [[SX_ATOMIC_CONT104:%.*]]
+// CHECK:       sx.atomic.cont104:
+// CHECK-NEXT:    store i16 [[TMP1081]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT103]]
+// CHECK:       sx.atomic.exit103:
+// CHECK-NEXT:    [[TMP1083:%.*]] = extractvalue { i16, i1 } [[TMP1080]], 1
+// CHECK-NEXT:    [[TMP1084:%.*]] = zext i1 [[TMP1083]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1084]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1085:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1086:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1087:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1085]], i16 [[TMP1086]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1088:%.*]] = extractvalue { i16, i1 } [[TMP1087]], 0
+// CHECK-NEXT:    [[TMP1089:%.*]] = extractvalue { i16, i1 } [[TMP1087]], 1
+// CHECK-NEXT:    br i1 [[TMP1089]], label [[SX_ATOMIC_EXIT105:%.*]], label [[SX_ATOMIC_CONT106:%.*]]
+// CHECK:       sx.atomic.cont106:
+// CHECK-NEXT:    store i16 [[TMP1088]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT105]]
+// CHECK:       sx.atomic.exit105:
+// CHECK-NEXT:    [[TMP1090:%.*]] = extractvalue { i16, i1 } [[TMP1087]], 1
+// CHECK-NEXT:    [[TMP1091:%.*]] = zext i1 [[TMP1090]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1091]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1092:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1093:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1092]] acquire, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1093]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1094:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1095:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1094]] acquire, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1095]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1096:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1097:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1096]] acquire, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1097]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1098:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1099:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1098]] acquire, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1099]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1100:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1101:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1102:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1100]], i16 [[TMP1101]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1103:%.*]] = extractvalue { i16, i1 } [[TMP1102]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1103]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1104:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1105:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1106:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1104]], i16 [[TMP1105]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1107:%.*]] = extractvalue { i16, i1 } [[TMP1106]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1107]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1108:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1109:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1108]] acquire, align 2
+// CHECK-NEXT:    [[TMP1110:%.*]] = icmp ugt i16 [[TMP1109]], [[TMP1108]]
+// CHECK-NEXT:    [[TMP1111:%.*]] = select i1 [[TMP1110]], i16 [[TMP1108]], i16 [[TMP1109]]
+// CHECK-NEXT:    store volatile i16 [[TMP1111]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1112:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1113:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1112]] acquire, align 2
+// CHECK-NEXT:    [[TMP1114:%.*]] = icmp ult i16 [[TMP1113]], [[TMP1112]]
+// CHECK-NEXT:    [[TMP1115:%.*]] = select i1 [[TMP1114]], i16 [[TMP1112]], i16 [[TMP1113]]
+// CHECK-NEXT:    store volatile i16 [[TMP1115]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1116:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1117:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1116]] acquire, align 2
+// CHECK-NEXT:    [[TMP1118:%.*]] = icmp ult i16 [[TMP1117]], [[TMP1116]]
+// CHECK-NEXT:    [[TMP1119:%.*]] = select i1 [[TMP1118]], i16 [[TMP1116]], i16 [[TMP1117]]
+// CHECK-NEXT:    store volatile i16 [[TMP1119]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1120:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1121:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1120]] acquire, align 2
+// CHECK-NEXT:    [[TMP1122:%.*]] = icmp ugt i16 [[TMP1121]], [[TMP1120]]
+// CHECK-NEXT:    [[TMP1123:%.*]] = select i1 [[TMP1122]], i16 [[TMP1120]], i16 [[TMP1121]]
+// CHECK-NEXT:    store volatile i16 [[TMP1123]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1124:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1125:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1126:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1124]], i16 [[TMP1125]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1127:%.*]] = extractvalue { i16, i1 } [[TMP1126]], 0
+// CHECK-NEXT:    [[TMP1128:%.*]] = extractvalue { i16, i1 } [[TMP1126]], 1
+// CHECK-NEXT:    [[TMP1129:%.*]] = select i1 [[TMP1128]], i16 [[TMP1124]], i16 [[TMP1127]]
+// CHECK-NEXT:    store volatile i16 [[TMP1129]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1130:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1131:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1132:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1130]], i16 [[TMP1131]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1133:%.*]] = extractvalue { i16, i1 } [[TMP1132]], 0
+// CHECK-NEXT:    [[TMP1134:%.*]] = extractvalue { i16, i1 } [[TMP1132]], 1
+// CHECK-NEXT:    [[TMP1135:%.*]] = select i1 [[TMP1134]], i16 [[TMP1130]], i16 [[TMP1133]]
+// CHECK-NEXT:    store volatile i16 [[TMP1135]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1136:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1137:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1138:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1136]], i16 [[TMP1137]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1139:%.*]] = extractvalue { i16, i1 } [[TMP1138]], 0
+// CHECK-NEXT:    [[TMP1140:%.*]] = extractvalue { i16, i1 } [[TMP1138]], 1
+// CHECK-NEXT:    br i1 [[TMP1140]], label [[SX_ATOMIC_EXIT107:%.*]], label [[SX_ATOMIC_CONT108:%.*]]
+// CHECK:       sx.atomic.cont108:
+// CHECK-NEXT:    store i16 [[TMP1139]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT107]]
+// CHECK:       sx.atomic.exit107:
+// CHECK-NEXT:    [[TMP1141:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1142:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1143:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1141]], i16 [[TMP1142]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1144:%.*]] = extractvalue { i16, i1 } [[TMP1143]], 0
+// CHECK-NEXT:    [[TMP1145:%.*]] = extractvalue { i16, i1 } [[TMP1143]], 1
+// CHECK-NEXT:    br i1 [[TMP1145]], label [[SX_ATOMIC_EXIT109:%.*]], label [[SX_ATOMIC_CONT110:%.*]]
+// CHECK:       sx.atomic.cont110:
+// CHECK-NEXT:    store i16 [[TMP1144]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT109]]
+// CHECK:       sx.atomic.exit109:
+// CHECK-NEXT:    [[TMP1146:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1147:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1148:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1146]], i16 [[TMP1147]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1149:%.*]] = extractvalue { i16, i1 } [[TMP1148]], 1
+// CHECK-NEXT:    [[TMP1150:%.*]] = zext i1 [[TMP1149]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1150]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1151:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1152:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1153:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1151]], i16 [[TMP1152]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1154:%.*]] = extractvalue { i16, i1 } [[TMP1153]], 1
+// CHECK-NEXT:    [[TMP1155:%.*]] = zext i1 [[TMP1154]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1155]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1156:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1157:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1158:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1156]], i16 [[TMP1157]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1159:%.*]] = extractvalue { i16, i1 } [[TMP1158]], 0
+// CHECK-NEXT:    [[TMP1160:%.*]] = extractvalue { i16, i1 } [[TMP1158]], 1
+// CHECK-NEXT:    br i1 [[TMP1160]], label [[SX_ATOMIC_EXIT111:%.*]], label [[SX_ATOMIC_CONT112:%.*]]
+// CHECK:       sx.atomic.cont112:
+// CHECK-NEXT:    store i16 [[TMP1159]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT111]]
+// CHECK:       sx.atomic.exit111:
+// CHECK-NEXT:    [[TMP1161:%.*]] = extractvalue { i16, i1 } [[TMP1158]], 1
+// CHECK-NEXT:    [[TMP1162:%.*]] = zext i1 [[TMP1161]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1162]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1163:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1164:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1165:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1163]], i16 [[TMP1164]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1166:%.*]] = extractvalue { i16, i1 } [[TMP1165]], 0
+// CHECK-NEXT:    [[TMP1167:%.*]] = extractvalue { i16, i1 } [[TMP1165]], 1
+// CHECK-NEXT:    br i1 [[TMP1167]], label [[SX_ATOMIC_EXIT113:%.*]], label [[SX_ATOMIC_CONT114:%.*]]
+// CHECK:       sx.atomic.cont114:
+// CHECK-NEXT:    store i16 [[TMP1166]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT113]]
+// CHECK:       sx.atomic.exit113:
+// CHECK-NEXT:    [[TMP1168:%.*]] = extractvalue { i16, i1 } [[TMP1165]], 1
+// CHECK-NEXT:    [[TMP1169:%.*]] = zext i1 [[TMP1168]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1169]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1170:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1171:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1170]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1171]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1172:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1173:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1172]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1173]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1174:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1175:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1174]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1175]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1176:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1177:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1176]] monotonic, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1177]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1178:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1179:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1180:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1178]], i16 [[TMP1179]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1181:%.*]] = extractvalue { i16, i1 } [[TMP1180]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1181]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1182:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1183:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1184:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1182]], i16 [[TMP1183]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1185:%.*]] = extractvalue { i16, i1 } [[TMP1184]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1185]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1186:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1187:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1186]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1188:%.*]] = icmp ugt i16 [[TMP1187]], [[TMP1186]]
+// CHECK-NEXT:    [[TMP1189:%.*]] = select i1 [[TMP1188]], i16 [[TMP1186]], i16 [[TMP1187]]
+// CHECK-NEXT:    store volatile i16 [[TMP1189]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1190:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1191:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1190]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1192:%.*]] = icmp ult i16 [[TMP1191]], [[TMP1190]]
+// CHECK-NEXT:    [[TMP1193:%.*]] = select i1 [[TMP1192]], i16 [[TMP1190]], i16 [[TMP1191]]
+// CHECK-NEXT:    store volatile i16 [[TMP1193]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1194:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1195:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1194]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1196:%.*]] = icmp ult i16 [[TMP1195]], [[TMP1194]]
+// CHECK-NEXT:    [[TMP1197:%.*]] = select i1 [[TMP1196]], i16 [[TMP1194]], i16 [[TMP1195]]
+// CHECK-NEXT:    store volatile i16 [[TMP1197]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1198:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1199:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1198]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1200:%.*]] = icmp ugt i16 [[TMP1199]], [[TMP1198]]
+// CHECK-NEXT:    [[TMP1201:%.*]] = select i1 [[TMP1200]], i16 [[TMP1198]], i16 [[TMP1199]]
+// CHECK-NEXT:    store volatile i16 [[TMP1201]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1202:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1203:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1204:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1202]], i16 [[TMP1203]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1205:%.*]] = extractvalue { i16, i1 } [[TMP1204]], 0
+// CHECK-NEXT:    [[TMP1206:%.*]] = extractvalue { i16, i1 } [[TMP1204]], 1
+// CHECK-NEXT:    [[TMP1207:%.*]] = select i1 [[TMP1206]], i16 [[TMP1202]], i16 [[TMP1205]]
+// CHECK-NEXT:    store volatile i16 [[TMP1207]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1208:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1209:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1210:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1208]], i16 [[TMP1209]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1211:%.*]] = extractvalue { i16, i1 } [[TMP1210]], 0
+// CHECK-NEXT:    [[TMP1212:%.*]] = extractvalue { i16, i1 } [[TMP1210]], 1
+// CHECK-NEXT:    [[TMP1213:%.*]] = select i1 [[TMP1212]], i16 [[TMP1208]], i16 [[TMP1211]]
+// CHECK-NEXT:    store volatile i16 [[TMP1213]], i16* [[SV]], align 2
+// CHECK-NEXT:    [[TMP1214:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1215:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1216:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1214]], i16 [[TMP1215]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1217:%.*]] = extractvalue { i16, i1 } [[TMP1216]], 0
+// CHECK-NEXT:    [[TMP1218:%.*]] = extractvalue { i16, i1 } [[TMP1216]], 1
+// CHECK-NEXT:    br i1 [[TMP1218]], label [[SX_ATOMIC_EXIT115:%.*]], label [[SX_ATOMIC_CONT116:%.*]]
+// CHECK:       sx.atomic.cont116:
+// CHECK-NEXT:    store i16 [[TMP1217]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT115]]
+// CHECK:       sx.atomic.exit115:
+// CHECK-NEXT:    [[TMP1219:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1220:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1221:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1219]], i16 [[TMP1220]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1222:%.*]] = extractvalue { i16, i1 } [[TMP1221]], 0
+// CHECK-NEXT:    [[TMP1223:%.*]] = extractvalue { i16, i1 } [[TMP1221]], 1
+// CHECK-NEXT:    br i1 [[TMP1223]], label [[SX_ATOMIC_EXIT117:%.*]], label [[SX_ATOMIC_CONT118:%.*]]
+// CHECK:       sx.atomic.cont118:
+// CHECK-NEXT:    store i16 [[TMP1222]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT117]]
+// CHECK:       sx.atomic.exit117:
+// CHECK-NEXT:    [[TMP1224:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1225:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1226:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1224]], i16 [[TMP1225]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1227:%.*]] = extractvalue { i16, i1 } [[TMP1226]], 1
+// CHECK-NEXT:    [[TMP1228:%.*]] = zext i1 [[TMP1227]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1228]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1229:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1230:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1231:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1229]], i16 [[TMP1230]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1232:%.*]] = extractvalue { i16, i1 } [[TMP1231]], 1
+// CHECK-NEXT:    [[TMP1233:%.*]] = zext i1 [[TMP1232]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1233]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1234:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1235:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1236:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1234]], i16 [[TMP1235]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1237:%.*]] = extractvalue { i16, i1 } [[TMP1236]], 0
+// CHECK-NEXT:    [[TMP1238:%.*]] = extractvalue { i16, i1 } [[TMP1236]], 1
+// CHECK-NEXT:    br i1 [[TMP1238]], label [[SX_ATOMIC_EXIT119:%.*]], label [[SX_ATOMIC_CONT120:%.*]]
+// CHECK:       sx.atomic.cont120:
+// CHECK-NEXT:    store i16 [[TMP1237]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT119]]
+// CHECK:       sx.atomic.exit119:
+// CHECK-NEXT:    [[TMP1239:%.*]] = extractvalue { i16, i1 } [[TMP1236]], 1
+// CHECK-NEXT:    [[TMP1240:%.*]] = zext i1 [[TMP1239]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1240]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1241:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1242:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1243:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1241]], i16 [[TMP1242]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1244:%.*]] = extractvalue { i16, i1 } [[TMP1243]], 0
+// CHECK-NEXT:    [[TMP1245:%.*]] = extractvalue { i16, i1 } [[TMP1243]], 1
+// CHECK-NEXT:    br i1 [[TMP1245]], label [[SX_ATOMIC_EXIT121:%.*]], label [[SX_ATOMIC_CONT122:%.*]]
+// CHECK:       sx.atomic.cont122:
+// CHECK-NEXT:    store i16 [[TMP1244]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT121]]
+// CHECK:       sx.atomic.exit121:
+// CHECK-NEXT:    [[TMP1246:%.*]] = extractvalue { i16, i1 } [[TMP1243]], 1
+// CHECK-NEXT:    [[TMP1247:%.*]] = zext i1 [[TMP1246]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1247]], i16* [[SR]], align 2
+// CHECK-NEXT:    [[TMP1248:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1249:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1248]] release, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1249]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1250:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1251:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1250]] release, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1251]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1252:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1253:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1252]] release, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1253]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1254:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1255:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1254]] release, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1255]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1256:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1257:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1258:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1256]], i16 [[TMP1257]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1259:%.*]] = extractvalue { i16, i1 } [[TMP1258]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1259]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1260:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1261:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1262:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1260]], i16 [[TMP1261]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1263:%.*]] = extractvalue { i16, i1 } [[TMP1262]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1263]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1264:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1265:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1264]] release, align 2
+// CHECK-NEXT:    [[TMP1266:%.*]] = icmp ugt i16 [[TMP1265]], [[TMP1264]]
+// CHECK-NEXT:    [[TMP1267:%.*]] = select i1 [[TMP1266]], i16 [[TMP1264]], i16 [[TMP1265]]
+// CHECK-NEXT:    store volatile i16 [[TMP1267]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1268:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1269:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1268]] release, align 2
+// CHECK-NEXT:    [[TMP1270:%.*]] = icmp ult i16 [[TMP1269]], [[TMP1268]]
+// CHECK-NEXT:    [[TMP1271:%.*]] = select i1 [[TMP1270]], i16 [[TMP1268]], i16 [[TMP1269]]
+// CHECK-NEXT:    store volatile i16 [[TMP1271]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1272:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1273:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1272]] release, align 2
+// CHECK-NEXT:    [[TMP1274:%.*]] = icmp ult i16 [[TMP1273]], [[TMP1272]]
+// CHECK-NEXT:    [[TMP1275:%.*]] = select i1 [[TMP1274]], i16 [[TMP1272]], i16 [[TMP1273]]
+// CHECK-NEXT:    store volatile i16 [[TMP1275]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1276:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1277:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1276]] release, align 2
+// CHECK-NEXT:    [[TMP1278:%.*]] = icmp ugt i16 [[TMP1277]], [[TMP1276]]
+// CHECK-NEXT:    [[TMP1279:%.*]] = select i1 [[TMP1278]], i16 [[TMP1276]], i16 [[TMP1277]]
+// CHECK-NEXT:    store volatile i16 [[TMP1279]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1280:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1281:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1282:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1280]], i16 [[TMP1281]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1283:%.*]] = extractvalue { i16, i1 } [[TMP1282]], 0
+// CHECK-NEXT:    [[TMP1284:%.*]] = extractvalue { i16, i1 } [[TMP1282]], 1
+// CHECK-NEXT:    [[TMP1285:%.*]] = select i1 [[TMP1284]], i16 [[TMP1280]], i16 [[TMP1283]]
+// CHECK-NEXT:    store volatile i16 [[TMP1285]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1286:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1287:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1288:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1286]], i16 [[TMP1287]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1289:%.*]] = extractvalue { i16, i1 } [[TMP1288]], 0
+// CHECK-NEXT:    [[TMP1290:%.*]] = extractvalue { i16, i1 } [[TMP1288]], 1
+// CHECK-NEXT:    [[TMP1291:%.*]] = select i1 [[TMP1290]], i16 [[TMP1286]], i16 [[TMP1289]]
+// CHECK-NEXT:    store volatile i16 [[TMP1291]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1292:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1293:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1294:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1292]], i16 [[TMP1293]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1295:%.*]] = extractvalue { i16, i1 } [[TMP1294]], 0
+// CHECK-NEXT:    [[TMP1296:%.*]] = extractvalue { i16, i1 } [[TMP1294]], 1
+// CHECK-NEXT:    br i1 [[TMP1296]], label [[SX_ATOMIC_EXIT123:%.*]], label [[SX_ATOMIC_CONT124:%.*]]
+// CHECK:       sx.atomic.cont124:
+// CHECK-NEXT:    store i16 [[TMP1295]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT123]]
+// CHECK:       sx.atomic.exit123:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1297:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1298:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1299:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1297]], i16 [[TMP1298]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1300:%.*]] = extractvalue { i16, i1 } [[TMP1299]], 0
+// CHECK-NEXT:    [[TMP1301:%.*]] = extractvalue { i16, i1 } [[TMP1299]], 1
+// CHECK-NEXT:    br i1 [[TMP1301]], label [[SX_ATOMIC_EXIT125:%.*]], label [[SX_ATOMIC_CONT126:%.*]]
+// CHECK:       sx.atomic.cont126:
+// CHECK-NEXT:    store i16 [[TMP1300]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT125]]
+// CHECK:       sx.atomic.exit125:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1302:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1303:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1304:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1302]], i16 [[TMP1303]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1305:%.*]] = extractvalue { i16, i1 } [[TMP1304]], 1
+// CHECK-NEXT:    [[TMP1306:%.*]] = zext i1 [[TMP1305]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1306]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1307:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1308:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1309:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1307]], i16 [[TMP1308]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1310:%.*]] = extractvalue { i16, i1 } [[TMP1309]], 1
+// CHECK-NEXT:    [[TMP1311:%.*]] = zext i1 [[TMP1310]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1311]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1312:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1313:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1314:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1312]], i16 [[TMP1313]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1315:%.*]] = extractvalue { i16, i1 } [[TMP1314]], 0
+// CHECK-NEXT:    [[TMP1316:%.*]] = extractvalue { i16, i1 } [[TMP1314]], 1
+// CHECK-NEXT:    br i1 [[TMP1316]], label [[SX_ATOMIC_EXIT127:%.*]], label [[SX_ATOMIC_CONT128:%.*]]
+// CHECK:       sx.atomic.cont128:
+// CHECK-NEXT:    store i16 [[TMP1315]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT127]]
+// CHECK:       sx.atomic.exit127:
+// CHECK-NEXT:    [[TMP1317:%.*]] = extractvalue { i16, i1 } [[TMP1314]], 1
+// CHECK-NEXT:    [[TMP1318:%.*]] = zext i1 [[TMP1317]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1318]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1319:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1320:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1321:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1319]], i16 [[TMP1320]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1322:%.*]] = extractvalue { i16, i1 } [[TMP1321]], 0
+// CHECK-NEXT:    [[TMP1323:%.*]] = extractvalue { i16, i1 } [[TMP1321]], 1
+// CHECK-NEXT:    br i1 [[TMP1323]], label [[SX_ATOMIC_EXIT129:%.*]], label [[SX_ATOMIC_CONT130:%.*]]
+// CHECK:       sx.atomic.cont130:
+// CHECK-NEXT:    store i16 [[TMP1322]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT129]]
+// CHECK:       sx.atomic.exit129:
+// CHECK-NEXT:    [[TMP1324:%.*]] = extractvalue { i16, i1 } [[TMP1321]], 1
+// CHECK-NEXT:    [[TMP1325:%.*]] = zext i1 [[TMP1324]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1325]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1326:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1327:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1326]] seq_cst, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1327]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1328:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1329:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1328]] seq_cst, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1329]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1330:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1331:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1330]] seq_cst, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1331]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1332:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1333:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1332]] seq_cst, align 2
+// CHECK-NEXT:    store volatile i16 [[TMP1333]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1334:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1335:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1336:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1334]], i16 [[TMP1335]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1337:%.*]] = extractvalue { i16, i1 } [[TMP1336]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1337]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1338:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1339:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1340:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1338]], i16 [[TMP1339]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1341:%.*]] = extractvalue { i16, i1 } [[TMP1340]], 0
+// CHECK-NEXT:    store volatile i16 [[TMP1341]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1342:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1343:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1342]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1344:%.*]] = icmp ugt i16 [[TMP1343]], [[TMP1342]]
+// CHECK-NEXT:    [[TMP1345:%.*]] = select i1 [[TMP1344]], i16 [[TMP1342]], i16 [[TMP1343]]
+// CHECK-NEXT:    store volatile i16 [[TMP1345]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1346:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1347:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1346]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1348:%.*]] = icmp ult i16 [[TMP1347]], [[TMP1346]]
+// CHECK-NEXT:    [[TMP1349:%.*]] = select i1 [[TMP1348]], i16 [[TMP1346]], i16 [[TMP1347]]
+// CHECK-NEXT:    store volatile i16 [[TMP1349]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1350:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1351:%.*]] = atomicrmw umin i16* [[SX]], i16 [[TMP1350]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1352:%.*]] = icmp ult i16 [[TMP1351]], [[TMP1350]]
+// CHECK-NEXT:    [[TMP1353:%.*]] = select i1 [[TMP1352]], i16 [[TMP1350]], i16 [[TMP1351]]
+// CHECK-NEXT:    store volatile i16 [[TMP1353]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1354:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1355:%.*]] = atomicrmw umax i16* [[SX]], i16 [[TMP1354]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1356:%.*]] = icmp ugt i16 [[TMP1355]], [[TMP1354]]
+// CHECK-NEXT:    [[TMP1357:%.*]] = select i1 [[TMP1356]], i16 [[TMP1354]], i16 [[TMP1355]]
+// CHECK-NEXT:    store volatile i16 [[TMP1357]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1358:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1359:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1360:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1358]], i16 [[TMP1359]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1361:%.*]] = extractvalue { i16, i1 } [[TMP1360]], 0
+// CHECK-NEXT:    [[TMP1362:%.*]] = extractvalue { i16, i1 } [[TMP1360]], 1
+// CHECK-NEXT:    [[TMP1363:%.*]] = select i1 [[TMP1362]], i16 [[TMP1358]], i16 [[TMP1361]]
+// CHECK-NEXT:    store volatile i16 [[TMP1363]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1364:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1365:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1366:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1364]], i16 [[TMP1365]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1367:%.*]] = extractvalue { i16, i1 } [[TMP1366]], 0
+// CHECK-NEXT:    [[TMP1368:%.*]] = extractvalue { i16, i1 } [[TMP1366]], 1
+// CHECK-NEXT:    [[TMP1369:%.*]] = select i1 [[TMP1368]], i16 [[TMP1364]], i16 [[TMP1367]]
+// CHECK-NEXT:    store volatile i16 [[TMP1369]], i16* [[SV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1370:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1371:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1372:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1370]], i16 [[TMP1371]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1373:%.*]] = extractvalue { i16, i1 } [[TMP1372]], 0
+// CHECK-NEXT:    [[TMP1374:%.*]] = extractvalue { i16, i1 } [[TMP1372]], 1
+// CHECK-NEXT:    br i1 [[TMP1374]], label [[SX_ATOMIC_EXIT131:%.*]], label [[SX_ATOMIC_CONT132:%.*]]
+// CHECK:       sx.atomic.cont132:
+// CHECK-NEXT:    store i16 [[TMP1373]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT131]]
+// CHECK:       sx.atomic.exit131:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1375:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1376:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1377:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1375]], i16 [[TMP1376]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1378:%.*]] = extractvalue { i16, i1 } [[TMP1377]], 0
+// CHECK-NEXT:    [[TMP1379:%.*]] = extractvalue { i16, i1 } [[TMP1377]], 1
+// CHECK-NEXT:    br i1 [[TMP1379]], label [[SX_ATOMIC_EXIT133:%.*]], label [[SX_ATOMIC_CONT134:%.*]]
+// CHECK:       sx.atomic.cont134:
+// CHECK-NEXT:    store i16 [[TMP1378]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT133]]
+// CHECK:       sx.atomic.exit133:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1380:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1381:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1382:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1380]], i16 [[TMP1381]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1383:%.*]] = extractvalue { i16, i1 } [[TMP1382]], 1
+// CHECK-NEXT:    [[TMP1384:%.*]] = zext i1 [[TMP1383]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1384]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1385:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1386:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1387:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1385]], i16 [[TMP1386]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1388:%.*]] = extractvalue { i16, i1 } [[TMP1387]], 1
+// CHECK-NEXT:    [[TMP1389:%.*]] = zext i1 [[TMP1388]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1389]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1390:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1391:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1392:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1390]], i16 [[TMP1391]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1393:%.*]] = extractvalue { i16, i1 } [[TMP1392]], 0
+// CHECK-NEXT:    [[TMP1394:%.*]] = extractvalue { i16, i1 } [[TMP1392]], 1
+// CHECK-NEXT:    br i1 [[TMP1394]], label [[SX_ATOMIC_EXIT135:%.*]], label [[SX_ATOMIC_CONT136:%.*]]
+// CHECK:       sx.atomic.cont136:
+// CHECK-NEXT:    store i16 [[TMP1393]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT135]]
+// CHECK:       sx.atomic.exit135:
+// CHECK-NEXT:    [[TMP1395:%.*]] = extractvalue { i16, i1 } [[TMP1392]], 1
+// CHECK-NEXT:    [[TMP1396:%.*]] = zext i1 [[TMP1395]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1396]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1397:%.*]] = load i16, i16* [[SE]], align 2
+// CHECK-NEXT:    [[TMP1398:%.*]] = load i16, i16* [[SD]], align 2
+// CHECK-NEXT:    [[TMP1399:%.*]] = cmpxchg i16* [[SX]], i16 [[TMP1397]], i16 [[TMP1398]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1400:%.*]] = extractvalue { i16, i1 } [[TMP1399]], 0
+// CHECK-NEXT:    [[TMP1401:%.*]] = extractvalue { i16, i1 } [[TMP1399]], 1
+// CHECK-NEXT:    br i1 [[TMP1401]], label [[SX_ATOMIC_EXIT137:%.*]], label [[SX_ATOMIC_CONT138:%.*]]
+// CHECK:       sx.atomic.cont138:
+// CHECK-NEXT:    store i16 [[TMP1400]], i16* [[SV]], align 2
+// CHECK-NEXT:    br label [[SX_ATOMIC_EXIT137]]
+// CHECK:       sx.atomic.exit137:
+// CHECK-NEXT:    [[TMP1402:%.*]] = extractvalue { i16, i1 } [[TMP1399]], 1
+// CHECK-NEXT:    [[TMP1403:%.*]] = zext i1 [[TMP1402]] to i16
+// CHECK-NEXT:    store volatile i16 [[TMP1403]], i16* [[SR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1404:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1405:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1404]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1405]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1406:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1407:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1406]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1407]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1408:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1409:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1408]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1409]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1410:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1411:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1410]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1411]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1412:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1413:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1414:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1412]], i16 [[TMP1413]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1415:%.*]] = extractvalue { i16, i1 } [[TMP1414]], 0
+// CHECK-NEXT:    store i16 [[TMP1415]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1416:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1417:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1418:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1416]], i16 [[TMP1417]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1419:%.*]] = extractvalue { i16, i1 } [[TMP1418]], 0
+// CHECK-NEXT:    store i16 [[TMP1419]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1420:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1421:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1420]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1422:%.*]] = icmp ugt i16 [[TMP1421]], [[TMP1420]]
+// CHECK-NEXT:    [[TMP1423:%.*]] = select i1 [[TMP1422]], i16 [[TMP1420]], i16 [[TMP1421]]
+// CHECK-NEXT:    store i16 [[TMP1423]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1424:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1425:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1424]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1426:%.*]] = icmp ult i16 [[TMP1425]], [[TMP1424]]
+// CHECK-NEXT:    [[TMP1427:%.*]] = select i1 [[TMP1426]], i16 [[TMP1424]], i16 [[TMP1425]]
+// CHECK-NEXT:    store i16 [[TMP1427]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1428:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1429:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1428]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1430:%.*]] = icmp ult i16 [[TMP1429]], [[TMP1428]]
+// CHECK-NEXT:    [[TMP1431:%.*]] = select i1 [[TMP1430]], i16 [[TMP1428]], i16 [[TMP1429]]
+// CHECK-NEXT:    store i16 [[TMP1431]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1432:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1433:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1432]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1434:%.*]] = icmp ugt i16 [[TMP1433]], [[TMP1432]]
+// CHECK-NEXT:    [[TMP1435:%.*]] = select i1 [[TMP1434]], i16 [[TMP1432]], i16 [[TMP1433]]
+// CHECK-NEXT:    store i16 [[TMP1435]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1436:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1437:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1438:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1436]], i16 [[TMP1437]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1439:%.*]] = extractvalue { i16, i1 } [[TMP1438]], 0
+// CHECK-NEXT:    [[TMP1440:%.*]] = extractvalue { i16, i1 } [[TMP1438]], 1
+// CHECK-NEXT:    [[TMP1441:%.*]] = select i1 [[TMP1440]], i16 [[TMP1436]], i16 [[TMP1439]]
+// CHECK-NEXT:    store i16 [[TMP1441]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1442:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1443:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1444:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1442]], i16 [[TMP1443]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1445:%.*]] = extractvalue { i16, i1 } [[TMP1444]], 0
+// CHECK-NEXT:    [[TMP1446:%.*]] = extractvalue { i16, i1 } [[TMP1444]], 1
+// CHECK-NEXT:    [[TMP1447:%.*]] = select i1 [[TMP1446]], i16 [[TMP1442]], i16 [[TMP1445]]
+// CHECK-NEXT:    store i16 [[TMP1447]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1448:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1449:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1450:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1448]], i16 [[TMP1449]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1451:%.*]] = extractvalue { i16, i1 } [[TMP1450]], 0
+// CHECK-NEXT:    [[TMP1452:%.*]] = extractvalue { i16, i1 } [[TMP1450]], 1
+// CHECK-NEXT:    br i1 [[TMP1452]], label [[USX_ATOMIC_EXIT:%.*]], label [[USX_ATOMIC_CONT:%.*]]
+// CHECK:       usx.atomic.cont:
+// CHECK-NEXT:    store i16 [[TMP1451]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT]]
+// CHECK:       usx.atomic.exit:
+// CHECK-NEXT:    [[TMP1453:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1454:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1455:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1453]], i16 [[TMP1454]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1456:%.*]] = extractvalue { i16, i1 } [[TMP1455]], 0
+// CHECK-NEXT:    [[TMP1457:%.*]] = extractvalue { i16, i1 } [[TMP1455]], 1
+// CHECK-NEXT:    br i1 [[TMP1457]], label [[USX_ATOMIC_EXIT139:%.*]], label [[USX_ATOMIC_CONT140:%.*]]
+// CHECK:       usx.atomic.cont140:
+// CHECK-NEXT:    store i16 [[TMP1456]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT139]]
+// CHECK:       usx.atomic.exit139:
+// CHECK-NEXT:    [[TMP1458:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1459:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1460:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1458]], i16 [[TMP1459]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1461:%.*]] = extractvalue { i16, i1 } [[TMP1460]], 1
+// CHECK-NEXT:    [[TMP1462:%.*]] = zext i1 [[TMP1461]] to i16
+// CHECK-NEXT:    store i16 [[TMP1462]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1463:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1464:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1465:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1463]], i16 [[TMP1464]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1466:%.*]] = extractvalue { i16, i1 } [[TMP1465]], 1
+// CHECK-NEXT:    [[TMP1467:%.*]] = zext i1 [[TMP1466]] to i16
+// CHECK-NEXT:    store i16 [[TMP1467]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1468:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1469:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1470:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1468]], i16 [[TMP1469]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1471:%.*]] = extractvalue { i16, i1 } [[TMP1470]], 0
+// CHECK-NEXT:    [[TMP1472:%.*]] = extractvalue { i16, i1 } [[TMP1470]], 1
+// CHECK-NEXT:    br i1 [[TMP1472]], label [[USX_ATOMIC_EXIT141:%.*]], label [[USX_ATOMIC_CONT142:%.*]]
+// CHECK:       usx.atomic.cont142:
+// CHECK-NEXT:    store i16 [[TMP1471]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT141]]
+// CHECK:       usx.atomic.exit141:
+// CHECK-NEXT:    [[TMP1473:%.*]] = extractvalue { i16, i1 } [[TMP1470]], 1
+// CHECK-NEXT:    [[TMP1474:%.*]] = zext i1 [[TMP1473]] to i16
+// CHECK-NEXT:    store i16 [[TMP1474]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1475:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1476:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1477:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1475]], i16 [[TMP1476]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1478:%.*]] = extractvalue { i16, i1 } [[TMP1477]], 0
+// CHECK-NEXT:    [[TMP1479:%.*]] = extractvalue { i16, i1 } [[TMP1477]], 1
+// CHECK-NEXT:    br i1 [[TMP1479]], label [[USX_ATOMIC_EXIT143:%.*]], label [[USX_ATOMIC_CONT144:%.*]]
+// CHECK:       usx.atomic.cont144:
+// CHECK-NEXT:    store i16 [[TMP1478]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT143]]
+// CHECK:       usx.atomic.exit143:
+// CHECK-NEXT:    [[TMP1480:%.*]] = extractvalue { i16, i1 } [[TMP1477]], 1
+// CHECK-NEXT:    [[TMP1481:%.*]] = zext i1 [[TMP1480]] to i16
+// CHECK-NEXT:    store i16 [[TMP1481]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1482:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1483:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1482]] acq_rel, align 2
+// CHECK-NEXT:    store i16 [[TMP1483]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1484:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1485:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1484]] acq_rel, align 2
+// CHECK-NEXT:    store i16 [[TMP1485]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1486:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1487:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1486]] acq_rel, align 2
+// CHECK-NEXT:    store i16 [[TMP1487]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1488:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1489:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1488]] acq_rel, align 2
+// CHECK-NEXT:    store i16 [[TMP1489]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1490:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1491:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1492:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1490]], i16 [[TMP1491]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1493:%.*]] = extractvalue { i16, i1 } [[TMP1492]], 0
+// CHECK-NEXT:    store i16 [[TMP1493]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1494:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1495:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1496:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1494]], i16 [[TMP1495]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1497:%.*]] = extractvalue { i16, i1 } [[TMP1496]], 0
+// CHECK-NEXT:    store i16 [[TMP1497]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1498:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1499:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1498]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1500:%.*]] = icmp ugt i16 [[TMP1499]], [[TMP1498]]
+// CHECK-NEXT:    [[TMP1501:%.*]] = select i1 [[TMP1500]], i16 [[TMP1498]], i16 [[TMP1499]]
+// CHECK-NEXT:    store i16 [[TMP1501]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1502:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1503:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1502]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1504:%.*]] = icmp ult i16 [[TMP1503]], [[TMP1502]]
+// CHECK-NEXT:    [[TMP1505:%.*]] = select i1 [[TMP1504]], i16 [[TMP1502]], i16 [[TMP1503]]
+// CHECK-NEXT:    store i16 [[TMP1505]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1506:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1507:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1506]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1508:%.*]] = icmp ult i16 [[TMP1507]], [[TMP1506]]
+// CHECK-NEXT:    [[TMP1509:%.*]] = select i1 [[TMP1508]], i16 [[TMP1506]], i16 [[TMP1507]]
+// CHECK-NEXT:    store i16 [[TMP1509]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1510:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1511:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1510]] acq_rel, align 2
+// CHECK-NEXT:    [[TMP1512:%.*]] = icmp ugt i16 [[TMP1511]], [[TMP1510]]
+// CHECK-NEXT:    [[TMP1513:%.*]] = select i1 [[TMP1512]], i16 [[TMP1510]], i16 [[TMP1511]]
+// CHECK-NEXT:    store i16 [[TMP1513]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1514:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1515:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1516:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1514]], i16 [[TMP1515]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1517:%.*]] = extractvalue { i16, i1 } [[TMP1516]], 0
+// CHECK-NEXT:    [[TMP1518:%.*]] = extractvalue { i16, i1 } [[TMP1516]], 1
+// CHECK-NEXT:    [[TMP1519:%.*]] = select i1 [[TMP1518]], i16 [[TMP1514]], i16 [[TMP1517]]
+// CHECK-NEXT:    store i16 [[TMP1519]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1520:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1521:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1522:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1520]], i16 [[TMP1521]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1523:%.*]] = extractvalue { i16, i1 } [[TMP1522]], 0
+// CHECK-NEXT:    [[TMP1524:%.*]] = extractvalue { i16, i1 } [[TMP1522]], 1
+// CHECK-NEXT:    [[TMP1525:%.*]] = select i1 [[TMP1524]], i16 [[TMP1520]], i16 [[TMP1523]]
+// CHECK-NEXT:    store i16 [[TMP1525]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1526:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1527:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1528:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1526]], i16 [[TMP1527]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1529:%.*]] = extractvalue { i16, i1 } [[TMP1528]], 0
+// CHECK-NEXT:    [[TMP1530:%.*]] = extractvalue { i16, i1 } [[TMP1528]], 1
+// CHECK-NEXT:    br i1 [[TMP1530]], label [[USX_ATOMIC_EXIT145:%.*]], label [[USX_ATOMIC_CONT146:%.*]]
+// CHECK:       usx.atomic.cont146:
+// CHECK-NEXT:    store i16 [[TMP1529]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT145]]
+// CHECK:       usx.atomic.exit145:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1531:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1532:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1533:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1531]], i16 [[TMP1532]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1534:%.*]] = extractvalue { i16, i1 } [[TMP1533]], 0
+// CHECK-NEXT:    [[TMP1535:%.*]] = extractvalue { i16, i1 } [[TMP1533]], 1
+// CHECK-NEXT:    br i1 [[TMP1535]], label [[USX_ATOMIC_EXIT147:%.*]], label [[USX_ATOMIC_CONT148:%.*]]
+// CHECK:       usx.atomic.cont148:
+// CHECK-NEXT:    store i16 [[TMP1534]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT147]]
+// CHECK:       usx.atomic.exit147:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1536:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1537:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1538:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1536]], i16 [[TMP1537]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1539:%.*]] = extractvalue { i16, i1 } [[TMP1538]], 1
+// CHECK-NEXT:    [[TMP1540:%.*]] = zext i1 [[TMP1539]] to i16
+// CHECK-NEXT:    store i16 [[TMP1540]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1541:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1542:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1543:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1541]], i16 [[TMP1542]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1544:%.*]] = extractvalue { i16, i1 } [[TMP1543]], 1
+// CHECK-NEXT:    [[TMP1545:%.*]] = zext i1 [[TMP1544]] to i16
+// CHECK-NEXT:    store i16 [[TMP1545]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1546:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1547:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1548:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1546]], i16 [[TMP1547]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1549:%.*]] = extractvalue { i16, i1 } [[TMP1548]], 0
+// CHECK-NEXT:    [[TMP1550:%.*]] = extractvalue { i16, i1 } [[TMP1548]], 1
+// CHECK-NEXT:    br i1 [[TMP1550]], label [[USX_ATOMIC_EXIT149:%.*]], label [[USX_ATOMIC_CONT150:%.*]]
+// CHECK:       usx.atomic.cont150:
+// CHECK-NEXT:    store i16 [[TMP1549]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT149]]
+// CHECK:       usx.atomic.exit149:
+// CHECK-NEXT:    [[TMP1551:%.*]] = extractvalue { i16, i1 } [[TMP1548]], 1
+// CHECK-NEXT:    [[TMP1552:%.*]] = zext i1 [[TMP1551]] to i16
+// CHECK-NEXT:    store i16 [[TMP1552]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1553:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1554:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1555:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1553]], i16 [[TMP1554]] acq_rel acquire, align 2
+// CHECK-NEXT:    [[TMP1556:%.*]] = extractvalue { i16, i1 } [[TMP1555]], 0
+// CHECK-NEXT:    [[TMP1557:%.*]] = extractvalue { i16, i1 } [[TMP1555]], 1
+// CHECK-NEXT:    br i1 [[TMP1557]], label [[USX_ATOMIC_EXIT151:%.*]], label [[USX_ATOMIC_CONT152:%.*]]
+// CHECK:       usx.atomic.cont152:
+// CHECK-NEXT:    store i16 [[TMP1556]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT151]]
+// CHECK:       usx.atomic.exit151:
+// CHECK-NEXT:    [[TMP1558:%.*]] = extractvalue { i16, i1 } [[TMP1555]], 1
+// CHECK-NEXT:    [[TMP1559:%.*]] = zext i1 [[TMP1558]] to i16
+// CHECK-NEXT:    store i16 [[TMP1559]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1560:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1561:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1560]] acquire, align 2
+// CHECK-NEXT:    store i16 [[TMP1561]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1562:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1563:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1562]] acquire, align 2
+// CHECK-NEXT:    store i16 [[TMP1563]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1564:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1565:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1564]] acquire, align 2
+// CHECK-NEXT:    store i16 [[TMP1565]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1566:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1567:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1566]] acquire, align 2
+// CHECK-NEXT:    store i16 [[TMP1567]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1568:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1569:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1570:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1568]], i16 [[TMP1569]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1571:%.*]] = extractvalue { i16, i1 } [[TMP1570]], 0
+// CHECK-NEXT:    store i16 [[TMP1571]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1572:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1573:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1574:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1572]], i16 [[TMP1573]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1575:%.*]] = extractvalue { i16, i1 } [[TMP1574]], 0
+// CHECK-NEXT:    store i16 [[TMP1575]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1576:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1577:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1576]] acquire, align 2
+// CHECK-NEXT:    [[TMP1578:%.*]] = icmp ugt i16 [[TMP1577]], [[TMP1576]]
+// CHECK-NEXT:    [[TMP1579:%.*]] = select i1 [[TMP1578]], i16 [[TMP1576]], i16 [[TMP1577]]
+// CHECK-NEXT:    store i16 [[TMP1579]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1580:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1581:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1580]] acquire, align 2
+// CHECK-NEXT:    [[TMP1582:%.*]] = icmp ult i16 [[TMP1581]], [[TMP1580]]
+// CHECK-NEXT:    [[TMP1583:%.*]] = select i1 [[TMP1582]], i16 [[TMP1580]], i16 [[TMP1581]]
+// CHECK-NEXT:    store i16 [[TMP1583]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1584:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1585:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1584]] acquire, align 2
+// CHECK-NEXT:    [[TMP1586:%.*]] = icmp ult i16 [[TMP1585]], [[TMP1584]]
+// CHECK-NEXT:    [[TMP1587:%.*]] = select i1 [[TMP1586]], i16 [[TMP1584]], i16 [[TMP1585]]
+// CHECK-NEXT:    store i16 [[TMP1587]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1588:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1589:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1588]] acquire, align 2
+// CHECK-NEXT:    [[TMP1590:%.*]] = icmp ugt i16 [[TMP1589]], [[TMP1588]]
+// CHECK-NEXT:    [[TMP1591:%.*]] = select i1 [[TMP1590]], i16 [[TMP1588]], i16 [[TMP1589]]
+// CHECK-NEXT:    store i16 [[TMP1591]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1592:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1593:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1594:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1592]], i16 [[TMP1593]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1595:%.*]] = extractvalue { i16, i1 } [[TMP1594]], 0
+// CHECK-NEXT:    [[TMP1596:%.*]] = extractvalue { i16, i1 } [[TMP1594]], 1
+// CHECK-NEXT:    [[TMP1597:%.*]] = select i1 [[TMP1596]], i16 [[TMP1592]], i16 [[TMP1595]]
+// CHECK-NEXT:    store i16 [[TMP1597]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1598:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1599:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1600:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1598]], i16 [[TMP1599]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1601:%.*]] = extractvalue { i16, i1 } [[TMP1600]], 0
+// CHECK-NEXT:    [[TMP1602:%.*]] = extractvalue { i16, i1 } [[TMP1600]], 1
+// CHECK-NEXT:    [[TMP1603:%.*]] = select i1 [[TMP1602]], i16 [[TMP1598]], i16 [[TMP1601]]
+// CHECK-NEXT:    store i16 [[TMP1603]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1604:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1605:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1606:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1604]], i16 [[TMP1605]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1607:%.*]] = extractvalue { i16, i1 } [[TMP1606]], 0
+// CHECK-NEXT:    [[TMP1608:%.*]] = extractvalue { i16, i1 } [[TMP1606]], 1
+// CHECK-NEXT:    br i1 [[TMP1608]], label [[USX_ATOMIC_EXIT153:%.*]], label [[USX_ATOMIC_CONT154:%.*]]
+// CHECK:       usx.atomic.cont154:
+// CHECK-NEXT:    store i16 [[TMP1607]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT153]]
+// CHECK:       usx.atomic.exit153:
+// CHECK-NEXT:    [[TMP1609:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1610:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1611:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1609]], i16 [[TMP1610]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1612:%.*]] = extractvalue { i16, i1 } [[TMP1611]], 0
+// CHECK-NEXT:    [[TMP1613:%.*]] = extractvalue { i16, i1 } [[TMP1611]], 1
+// CHECK-NEXT:    br i1 [[TMP1613]], label [[USX_ATOMIC_EXIT155:%.*]], label [[USX_ATOMIC_CONT156:%.*]]
+// CHECK:       usx.atomic.cont156:
+// CHECK-NEXT:    store i16 [[TMP1612]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT155]]
+// CHECK:       usx.atomic.exit155:
+// CHECK-NEXT:    [[TMP1614:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1615:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1616:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1614]], i16 [[TMP1615]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1617:%.*]] = extractvalue { i16, i1 } [[TMP1616]], 1
+// CHECK-NEXT:    [[TMP1618:%.*]] = zext i1 [[TMP1617]] to i16
+// CHECK-NEXT:    store i16 [[TMP1618]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1619:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1620:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1621:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1619]], i16 [[TMP1620]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1622:%.*]] = extractvalue { i16, i1 } [[TMP1621]], 1
+// CHECK-NEXT:    [[TMP1623:%.*]] = zext i1 [[TMP1622]] to i16
+// CHECK-NEXT:    store i16 [[TMP1623]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1624:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1625:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1626:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1624]], i16 [[TMP1625]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1627:%.*]] = extractvalue { i16, i1 } [[TMP1626]], 0
+// CHECK-NEXT:    [[TMP1628:%.*]] = extractvalue { i16, i1 } [[TMP1626]], 1
+// CHECK-NEXT:    br i1 [[TMP1628]], label [[USX_ATOMIC_EXIT157:%.*]], label [[USX_ATOMIC_CONT158:%.*]]
+// CHECK:       usx.atomic.cont158:
+// CHECK-NEXT:    store i16 [[TMP1627]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT157]]
+// CHECK:       usx.atomic.exit157:
+// CHECK-NEXT:    [[TMP1629:%.*]] = extractvalue { i16, i1 } [[TMP1626]], 1
+// CHECK-NEXT:    [[TMP1630:%.*]] = zext i1 [[TMP1629]] to i16
+// CHECK-NEXT:    store i16 [[TMP1630]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1631:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1632:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1633:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1631]], i16 [[TMP1632]] acquire acquire, align 2
+// CHECK-NEXT:    [[TMP1634:%.*]] = extractvalue { i16, i1 } [[TMP1633]], 0
+// CHECK-NEXT:    [[TMP1635:%.*]] = extractvalue { i16, i1 } [[TMP1633]], 1
+// CHECK-NEXT:    br i1 [[TMP1635]], label [[USX_ATOMIC_EXIT159:%.*]], label [[USX_ATOMIC_CONT160:%.*]]
+// CHECK:       usx.atomic.cont160:
+// CHECK-NEXT:    store i16 [[TMP1634]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT159]]
+// CHECK:       usx.atomic.exit159:
+// CHECK-NEXT:    [[TMP1636:%.*]] = extractvalue { i16, i1 } [[TMP1633]], 1
+// CHECK-NEXT:    [[TMP1637:%.*]] = zext i1 [[TMP1636]] to i16
+// CHECK-NEXT:    store i16 [[TMP1637]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1638:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1639:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1638]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1639]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1640:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1641:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1640]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1641]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1642:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1643:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1642]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1643]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1644:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1645:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1644]] monotonic, align 2
+// CHECK-NEXT:    store i16 [[TMP1645]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1646:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1647:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1648:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1646]], i16 [[TMP1647]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1649:%.*]] = extractvalue { i16, i1 } [[TMP1648]], 0
+// CHECK-NEXT:    store i16 [[TMP1649]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1650:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1651:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1652:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1650]], i16 [[TMP1651]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1653:%.*]] = extractvalue { i16, i1 } [[TMP1652]], 0
+// CHECK-NEXT:    store i16 [[TMP1653]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1654:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1655:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1654]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1656:%.*]] = icmp ugt i16 [[TMP1655]], [[TMP1654]]
+// CHECK-NEXT:    [[TMP1657:%.*]] = select i1 [[TMP1656]], i16 [[TMP1654]], i16 [[TMP1655]]
+// CHECK-NEXT:    store i16 [[TMP1657]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1658:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1659:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1658]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1660:%.*]] = icmp ult i16 [[TMP1659]], [[TMP1658]]
+// CHECK-NEXT:    [[TMP1661:%.*]] = select i1 [[TMP1660]], i16 [[TMP1658]], i16 [[TMP1659]]
+// CHECK-NEXT:    store i16 [[TMP1661]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1662:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1663:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1662]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1664:%.*]] = icmp ult i16 [[TMP1663]], [[TMP1662]]
+// CHECK-NEXT:    [[TMP1665:%.*]] = select i1 [[TMP1664]], i16 [[TMP1662]], i16 [[TMP1663]]
+// CHECK-NEXT:    store i16 [[TMP1665]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1666:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1667:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1666]] monotonic, align 2
+// CHECK-NEXT:    [[TMP1668:%.*]] = icmp ugt i16 [[TMP1667]], [[TMP1666]]
+// CHECK-NEXT:    [[TMP1669:%.*]] = select i1 [[TMP1668]], i16 [[TMP1666]], i16 [[TMP1667]]
+// CHECK-NEXT:    store i16 [[TMP1669]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1670:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1671:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1672:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1670]], i16 [[TMP1671]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1673:%.*]] = extractvalue { i16, i1 } [[TMP1672]], 0
+// CHECK-NEXT:    [[TMP1674:%.*]] = extractvalue { i16, i1 } [[TMP1672]], 1
+// CHECK-NEXT:    [[TMP1675:%.*]] = select i1 [[TMP1674]], i16 [[TMP1670]], i16 [[TMP1673]]
+// CHECK-NEXT:    store i16 [[TMP1675]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1676:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1677:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1678:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1676]], i16 [[TMP1677]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1679:%.*]] = extractvalue { i16, i1 } [[TMP1678]], 0
+// CHECK-NEXT:    [[TMP1680:%.*]] = extractvalue { i16, i1 } [[TMP1678]], 1
+// CHECK-NEXT:    [[TMP1681:%.*]] = select i1 [[TMP1680]], i16 [[TMP1676]], i16 [[TMP1679]]
+// CHECK-NEXT:    store i16 [[TMP1681]], i16* [[USV]], align 2
+// CHECK-NEXT:    [[TMP1682:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1683:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1684:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1682]], i16 [[TMP1683]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1685:%.*]] = extractvalue { i16, i1 } [[TMP1684]], 0
+// CHECK-NEXT:    [[TMP1686:%.*]] = extractvalue { i16, i1 } [[TMP1684]], 1
+// CHECK-NEXT:    br i1 [[TMP1686]], label [[USX_ATOMIC_EXIT161:%.*]], label [[USX_ATOMIC_CONT162:%.*]]
+// CHECK:       usx.atomic.cont162:
+// CHECK-NEXT:    store i16 [[TMP1685]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT161]]
+// CHECK:       usx.atomic.exit161:
+// CHECK-NEXT:    [[TMP1687:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1688:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1689:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1687]], i16 [[TMP1688]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1690:%.*]] = extractvalue { i16, i1 } [[TMP1689]], 0
+// CHECK-NEXT:    [[TMP1691:%.*]] = extractvalue { i16, i1 } [[TMP1689]], 1
+// CHECK-NEXT:    br i1 [[TMP1691]], label [[USX_ATOMIC_EXIT163:%.*]], label [[USX_ATOMIC_CONT164:%.*]]
+// CHECK:       usx.atomic.cont164:
+// CHECK-NEXT:    store i16 [[TMP1690]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT163]]
+// CHECK:       usx.atomic.exit163:
+// CHECK-NEXT:    [[TMP1692:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1693:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1694:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1692]], i16 [[TMP1693]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1695:%.*]] = extractvalue { i16, i1 } [[TMP1694]], 1
+// CHECK-NEXT:    [[TMP1696:%.*]] = zext i1 [[TMP1695]] to i16
+// CHECK-NEXT:    store i16 [[TMP1696]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1697:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1698:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1699:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1697]], i16 [[TMP1698]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1700:%.*]] = extractvalue { i16, i1 } [[TMP1699]], 1
+// CHECK-NEXT:    [[TMP1701:%.*]] = zext i1 [[TMP1700]] to i16
+// CHECK-NEXT:    store i16 [[TMP1701]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1702:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1703:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1704:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1702]], i16 [[TMP1703]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1705:%.*]] = extractvalue { i16, i1 } [[TMP1704]], 0
+// CHECK-NEXT:    [[TMP1706:%.*]] = extractvalue { i16, i1 } [[TMP1704]], 1
+// CHECK-NEXT:    br i1 [[TMP1706]], label [[USX_ATOMIC_EXIT165:%.*]], label [[USX_ATOMIC_CONT166:%.*]]
+// CHECK:       usx.atomic.cont166:
+// CHECK-NEXT:    store i16 [[TMP1705]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT165]]
+// CHECK:       usx.atomic.exit165:
+// CHECK-NEXT:    [[TMP1707:%.*]] = extractvalue { i16, i1 } [[TMP1704]], 1
+// CHECK-NEXT:    [[TMP1708:%.*]] = zext i1 [[TMP1707]] to i16
+// CHECK-NEXT:    store i16 [[TMP1708]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1709:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1710:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1711:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1709]], i16 [[TMP1710]] monotonic monotonic, align 2
+// CHECK-NEXT:    [[TMP1712:%.*]] = extractvalue { i16, i1 } [[TMP1711]], 0
+// CHECK-NEXT:    [[TMP1713:%.*]] = extractvalue { i16, i1 } [[TMP1711]], 1
+// CHECK-NEXT:    br i1 [[TMP1713]], label [[USX_ATOMIC_EXIT167:%.*]], label [[USX_ATOMIC_CONT168:%.*]]
+// CHECK:       usx.atomic.cont168:
+// CHECK-NEXT:    store i16 [[TMP1712]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT167]]
+// CHECK:       usx.atomic.exit167:
+// CHECK-NEXT:    [[TMP1714:%.*]] = extractvalue { i16, i1 } [[TMP1711]], 1
+// CHECK-NEXT:    [[TMP1715:%.*]] = zext i1 [[TMP1714]] to i16
+// CHECK-NEXT:    store i16 [[TMP1715]], i16* [[USR]], align 2
+// CHECK-NEXT:    [[TMP1716:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1717:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1716]] release, align 2
+// CHECK-NEXT:    store i16 [[TMP1717]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1718:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1719:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1718]] release, align 2
+// CHECK-NEXT:    store i16 [[TMP1719]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1720:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1721:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1720]] release, align 2
+// CHECK-NEXT:    store i16 [[TMP1721]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1722:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1723:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1722]] release, align 2
+// CHECK-NEXT:    store i16 [[TMP1723]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1724:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1725:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1726:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1724]], i16 [[TMP1725]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1727:%.*]] = extractvalue { i16, i1 } [[TMP1726]], 0
+// CHECK-NEXT:    store i16 [[TMP1727]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1728:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1729:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1730:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1728]], i16 [[TMP1729]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1731:%.*]] = extractvalue { i16, i1 } [[TMP1730]], 0
+// CHECK-NEXT:    store i16 [[TMP1731]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1732:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1733:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1732]] release, align 2
+// CHECK-NEXT:    [[TMP1734:%.*]] = icmp ugt i16 [[TMP1733]], [[TMP1732]]
+// CHECK-NEXT:    [[TMP1735:%.*]] = select i1 [[TMP1734]], i16 [[TMP1732]], i16 [[TMP1733]]
+// CHECK-NEXT:    store i16 [[TMP1735]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1736:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1737:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1736]] release, align 2
+// CHECK-NEXT:    [[TMP1738:%.*]] = icmp ult i16 [[TMP1737]], [[TMP1736]]
+// CHECK-NEXT:    [[TMP1739:%.*]] = select i1 [[TMP1738]], i16 [[TMP1736]], i16 [[TMP1737]]
+// CHECK-NEXT:    store i16 [[TMP1739]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1740:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1741:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1740]] release, align 2
+// CHECK-NEXT:    [[TMP1742:%.*]] = icmp ult i16 [[TMP1741]], [[TMP1740]]
+// CHECK-NEXT:    [[TMP1743:%.*]] = select i1 [[TMP1742]], i16 [[TMP1740]], i16 [[TMP1741]]
+// CHECK-NEXT:    store i16 [[TMP1743]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1744:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1745:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1744]] release, align 2
+// CHECK-NEXT:    [[TMP1746:%.*]] = icmp ugt i16 [[TMP1745]], [[TMP1744]]
+// CHECK-NEXT:    [[TMP1747:%.*]] = select i1 [[TMP1746]], i16 [[TMP1744]], i16 [[TMP1745]]
+// CHECK-NEXT:    store i16 [[TMP1747]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1748:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1749:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1750:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1748]], i16 [[TMP1749]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1751:%.*]] = extractvalue { i16, i1 } [[TMP1750]], 0
+// CHECK-NEXT:    [[TMP1752:%.*]] = extractvalue { i16, i1 } [[TMP1750]], 1
+// CHECK-NEXT:    [[TMP1753:%.*]] = select i1 [[TMP1752]], i16 [[TMP1748]], i16 [[TMP1751]]
+// CHECK-NEXT:    store i16 [[TMP1753]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1754:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1755:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1756:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1754]], i16 [[TMP1755]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1757:%.*]] = extractvalue { i16, i1 } [[TMP1756]], 0
+// CHECK-NEXT:    [[TMP1758:%.*]] = extractvalue { i16, i1 } [[TMP1756]], 1
+// CHECK-NEXT:    [[TMP1759:%.*]] = select i1 [[TMP1758]], i16 [[TMP1754]], i16 [[TMP1757]]
+// CHECK-NEXT:    store i16 [[TMP1759]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1760:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1761:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1762:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1760]], i16 [[TMP1761]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1763:%.*]] = extractvalue { i16, i1 } [[TMP1762]], 0
+// CHECK-NEXT:    [[TMP1764:%.*]] = extractvalue { i16, i1 } [[TMP1762]], 1
+// CHECK-NEXT:    br i1 [[TMP1764]], label [[USX_ATOMIC_EXIT169:%.*]], label [[USX_ATOMIC_CONT170:%.*]]
+// CHECK:       usx.atomic.cont170:
+// CHECK-NEXT:    store i16 [[TMP1763]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT169]]
+// CHECK:       usx.atomic.exit169:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1765:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1766:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1767:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1765]], i16 [[TMP1766]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1768:%.*]] = extractvalue { i16, i1 } [[TMP1767]], 0
+// CHECK-NEXT:    [[TMP1769:%.*]] = extractvalue { i16, i1 } [[TMP1767]], 1
+// CHECK-NEXT:    br i1 [[TMP1769]], label [[USX_ATOMIC_EXIT171:%.*]], label [[USX_ATOMIC_CONT172:%.*]]
+// CHECK:       usx.atomic.cont172:
+// CHECK-NEXT:    store i16 [[TMP1768]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT171]]
+// CHECK:       usx.atomic.exit171:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1770:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1771:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1772:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1770]], i16 [[TMP1771]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1773:%.*]] = extractvalue { i16, i1 } [[TMP1772]], 1
+// CHECK-NEXT:    [[TMP1774:%.*]] = zext i1 [[TMP1773]] to i16
+// CHECK-NEXT:    store i16 [[TMP1774]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1775:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1776:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1777:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1775]], i16 [[TMP1776]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1778:%.*]] = extractvalue { i16, i1 } [[TMP1777]], 1
+// CHECK-NEXT:    [[TMP1779:%.*]] = zext i1 [[TMP1778]] to i16
+// CHECK-NEXT:    store i16 [[TMP1779]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1780:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1781:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1782:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1780]], i16 [[TMP1781]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1783:%.*]] = extractvalue { i16, i1 } [[TMP1782]], 0
+// CHECK-NEXT:    [[TMP1784:%.*]] = extractvalue { i16, i1 } [[TMP1782]], 1
+// CHECK-NEXT:    br i1 [[TMP1784]], label [[USX_ATOMIC_EXIT173:%.*]], label [[USX_ATOMIC_CONT174:%.*]]
+// CHECK:       usx.atomic.cont174:
+// CHECK-NEXT:    store i16 [[TMP1783]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT173]]
+// CHECK:       usx.atomic.exit173:
+// CHECK-NEXT:    [[TMP1785:%.*]] = extractvalue { i16, i1 } [[TMP1782]], 1
+// CHECK-NEXT:    [[TMP1786:%.*]] = zext i1 [[TMP1785]] to i16
+// CHECK-NEXT:    store i16 [[TMP1786]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1787:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1788:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1789:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1787]], i16 [[TMP1788]] release monotonic, align 2
+// CHECK-NEXT:    [[TMP1790:%.*]] = extractvalue { i16, i1 } [[TMP1789]], 0
+// CHECK-NEXT:    [[TMP1791:%.*]] = extractvalue { i16, i1 } [[TMP1789]], 1
+// CHECK-NEXT:    br i1 [[TMP1791]], label [[USX_ATOMIC_EXIT175:%.*]], label [[USX_ATOMIC_CONT176:%.*]]
+// CHECK:       usx.atomic.cont176:
+// CHECK-NEXT:    store i16 [[TMP1790]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT175]]
+// CHECK:       usx.atomic.exit175:
+// CHECK-NEXT:    [[TMP1792:%.*]] = extractvalue { i16, i1 } [[TMP1789]], 1
+// CHECK-NEXT:    [[TMP1793:%.*]] = zext i1 [[TMP1792]] to i16
+// CHECK-NEXT:    store i16 [[TMP1793]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1794:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1795:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1794]] seq_cst, align 2
+// CHECK-NEXT:    store i16 [[TMP1795]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1796:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1797:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1796]] seq_cst, align 2
+// CHECK-NEXT:    store i16 [[TMP1797]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1798:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1799:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1798]] seq_cst, align 2
+// CHECK-NEXT:    store i16 [[TMP1799]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1800:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1801:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1800]] seq_cst, align 2
+// CHECK-NEXT:    store i16 [[TMP1801]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1802:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1803:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1804:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1802]], i16 [[TMP1803]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1805:%.*]] = extractvalue { i16, i1 } [[TMP1804]], 0
+// CHECK-NEXT:    store i16 [[TMP1805]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1806:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1807:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1808:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1806]], i16 [[TMP1807]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1809:%.*]] = extractvalue { i16, i1 } [[TMP1808]], 0
+// CHECK-NEXT:    store i16 [[TMP1809]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1810:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1811:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1810]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1812:%.*]] = icmp ugt i16 [[TMP1811]], [[TMP1810]]
+// CHECK-NEXT:    [[TMP1813:%.*]] = select i1 [[TMP1812]], i16 [[TMP1810]], i16 [[TMP1811]]
+// CHECK-NEXT:    store i16 [[TMP1813]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1814:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1815:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1814]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1816:%.*]] = icmp ult i16 [[TMP1815]], [[TMP1814]]
+// CHECK-NEXT:    [[TMP1817:%.*]] = select i1 [[TMP1816]], i16 [[TMP1814]], i16 [[TMP1815]]
+// CHECK-NEXT:    store i16 [[TMP1817]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1818:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1819:%.*]] = atomicrmw umin i16* [[USX]], i16 [[TMP1818]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1820:%.*]] = icmp ult i16 [[TMP1819]], [[TMP1818]]
+// CHECK-NEXT:    [[TMP1821:%.*]] = select i1 [[TMP1820]], i16 [[TMP1818]], i16 [[TMP1819]]
+// CHECK-NEXT:    store i16 [[TMP1821]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1822:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1823:%.*]] = atomicrmw umax i16* [[USX]], i16 [[TMP1822]] seq_cst, align 2
+// CHECK-NEXT:    [[TMP1824:%.*]] = icmp ugt i16 [[TMP1823]], [[TMP1822]]
+// CHECK-NEXT:    [[TMP1825:%.*]] = select i1 [[TMP1824]], i16 [[TMP1822]], i16 [[TMP1823]]
+// CHECK-NEXT:    store i16 [[TMP1825]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1826:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1827:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1828:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1826]], i16 [[TMP1827]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1829:%.*]] = extractvalue { i16, i1 } [[TMP1828]], 0
+// CHECK-NEXT:    [[TMP1830:%.*]] = extractvalue { i16, i1 } [[TMP1828]], 1
+// CHECK-NEXT:    [[TMP1831:%.*]] = select i1 [[TMP1830]], i16 [[TMP1826]], i16 [[TMP1829]]
+// CHECK-NEXT:    store i16 [[TMP1831]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1832:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1833:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1834:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1832]], i16 [[TMP1833]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1835:%.*]] = extractvalue { i16, i1 } [[TMP1834]], 0
+// CHECK-NEXT:    [[TMP1836:%.*]] = extractvalue { i16, i1 } [[TMP1834]], 1
+// CHECK-NEXT:    [[TMP1837:%.*]] = select i1 [[TMP1836]], i16 [[TMP1832]], i16 [[TMP1835]]
+// CHECK-NEXT:    store i16 [[TMP1837]], i16* [[USV]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1838:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1839:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1840:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1838]], i16 [[TMP1839]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1841:%.*]] = extractvalue { i16, i1 } [[TMP1840]], 0
+// CHECK-NEXT:    [[TMP1842:%.*]] = extractvalue { i16, i1 } [[TMP1840]], 1
+// CHECK-NEXT:    br i1 [[TMP1842]], label [[USX_ATOMIC_EXIT177:%.*]], label [[USX_ATOMIC_CONT178:%.*]]
+// CHECK:       usx.atomic.cont178:
+// CHECK-NEXT:    store i16 [[TMP1841]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT177]]
+// CHECK:       usx.atomic.exit177:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1843:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1844:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1845:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1843]], i16 [[TMP1844]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1846:%.*]] = extractvalue { i16, i1 } [[TMP1845]], 0
+// CHECK-NEXT:    [[TMP1847:%.*]] = extractvalue { i16, i1 } [[TMP1845]], 1
+// CHECK-NEXT:    br i1 [[TMP1847]], label [[USX_ATOMIC_EXIT179:%.*]], label [[USX_ATOMIC_CONT180:%.*]]
+// CHECK:       usx.atomic.cont180:
+// CHECK-NEXT:    store i16 [[TMP1846]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT179]]
+// CHECK:       usx.atomic.exit179:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1848:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1849:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1850:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1848]], i16 [[TMP1849]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1851:%.*]] = extractvalue { i16, i1 } [[TMP1850]], 1
+// CHECK-NEXT:    [[TMP1852:%.*]] = zext i1 [[TMP1851]] to i16
+// CHECK-NEXT:    store i16 [[TMP1852]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1853:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1854:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1855:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1853]], i16 [[TMP1854]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1856:%.*]] = extractvalue { i16, i1 } [[TMP1855]], 1
+// CHECK-NEXT:    [[TMP1857:%.*]] = zext i1 [[TMP1856]] to i16
+// CHECK-NEXT:    store i16 [[TMP1857]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1858:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1859:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1860:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1858]], i16 [[TMP1859]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1861:%.*]] = extractvalue { i16, i1 } [[TMP1860]], 0
+// CHECK-NEXT:    [[TMP1862:%.*]] = extractvalue { i16, i1 } [[TMP1860]], 1
+// CHECK-NEXT:    br i1 [[TMP1862]], label [[USX_ATOMIC_EXIT181:%.*]], label [[USX_ATOMIC_CONT182:%.*]]
+// CHECK:       usx.atomic.cont182:
+// CHECK-NEXT:    store i16 [[TMP1861]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT181]]
+// CHECK:       usx.atomic.exit181:
+// CHECK-NEXT:    [[TMP1863:%.*]] = extractvalue { i16, i1 } [[TMP1860]], 1
+// CHECK-NEXT:    [[TMP1864:%.*]] = zext i1 [[TMP1863]] to i16
+// CHECK-NEXT:    store i16 [[TMP1864]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1865:%.*]] = load i16, i16* [[USE]], align 2
+// CHECK-NEXT:    [[TMP1866:%.*]] = load i16, i16* [[USD]], align 2
+// CHECK-NEXT:    [[TMP1867:%.*]] = cmpxchg i16* [[USX]], i16 [[TMP1865]], i16 [[TMP1866]] seq_cst seq_cst, align 2
+// CHECK-NEXT:    [[TMP1868:%.*]] = extractvalue { i16, i1 } [[TMP1867]], 0
+// CHECK-NEXT:    [[TMP1869:%.*]] = extractvalue { i16, i1 } [[TMP1867]], 1
+// CHECK-NEXT:    br i1 [[TMP1869]], label [[USX_ATOMIC_EXIT183:%.*]], label [[USX_ATOMIC_CONT184:%.*]]
+// CHECK:       usx.atomic.cont184:
+// CHECK-NEXT:    store i16 [[TMP1868]], i16* [[USV]], align 2
+// CHECK-NEXT:    br label [[USX_ATOMIC_EXIT183]]
+// CHECK:       usx.atomic.exit183:
+// CHECK-NEXT:    [[TMP1870:%.*]] = extractvalue { i16, i1 } [[TMP1867]], 1
+// CHECK-NEXT:    [[TMP1871:%.*]] = zext i1 [[TMP1870]] to i16
+// CHECK-NEXT:    store i16 [[TMP1871]], i16* [[USR]], align 2
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1872:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1873:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1872]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1873]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1874:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1875:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1874]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1875]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1876:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1877:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1876]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1877]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1878:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1879:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1878]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1879]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1880:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1881:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1882:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1880]], i32 [[TMP1881]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1883:%.*]] = extractvalue { i32, i1 } [[TMP1882]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP1883]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1884:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1885:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1886:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1884]], i32 [[TMP1885]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1887:%.*]] = extractvalue { i32, i1 } [[TMP1886]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP1887]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1888:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1889:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1888]] monotonic, align 4
+// CHECK-NEXT:    [[TMP1890:%.*]] = icmp ugt i32 [[TMP1889]], [[TMP1888]]
+// CHECK-NEXT:    [[TMP1891:%.*]] = select i1 [[TMP1890]], i32 [[TMP1888]], i32 [[TMP1889]]
+// CHECK-NEXT:    store volatile i32 [[TMP1891]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1892:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1893:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1892]] monotonic, align 4
+// CHECK-NEXT:    [[TMP1894:%.*]] = icmp ult i32 [[TMP1893]], [[TMP1892]]
+// CHECK-NEXT:    [[TMP1895:%.*]] = select i1 [[TMP1894]], i32 [[TMP1892]], i32 [[TMP1893]]
+// CHECK-NEXT:    store volatile i32 [[TMP1895]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1896:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1897:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1896]] monotonic, align 4
+// CHECK-NEXT:    [[TMP1898:%.*]] = icmp ult i32 [[TMP1897]], [[TMP1896]]
+// CHECK-NEXT:    [[TMP1899:%.*]] = select i1 [[TMP1898]], i32 [[TMP1896]], i32 [[TMP1897]]
+// CHECK-NEXT:    store volatile i32 [[TMP1899]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1900:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1901:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1900]] monotonic, align 4
+// CHECK-NEXT:    [[TMP1902:%.*]] = icmp ugt i32 [[TMP1901]], [[TMP1900]]
+// CHECK-NEXT:    [[TMP1903:%.*]] = select i1 [[TMP1902]], i32 [[TMP1900]], i32 [[TMP1901]]
+// CHECK-NEXT:    store volatile i32 [[TMP1903]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1904:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1905:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1906:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1904]], i32 [[TMP1905]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1907:%.*]] = extractvalue { i32, i1 } [[TMP1906]], 0
+// CHECK-NEXT:    [[TMP1908:%.*]] = extractvalue { i32, i1 } [[TMP1906]], 1
+// CHECK-NEXT:    [[TMP1909:%.*]] = select i1 [[TMP1908]], i32 [[TMP1904]], i32 [[TMP1907]]
+// CHECK-NEXT:    store volatile i32 [[TMP1909]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1910:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1911:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1912:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1910]], i32 [[TMP1911]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1913:%.*]] = extractvalue { i32, i1 } [[TMP1912]], 0
+// CHECK-NEXT:    [[TMP1914:%.*]] = extractvalue { i32, i1 } [[TMP1912]], 1
+// CHECK-NEXT:    [[TMP1915:%.*]] = select i1 [[TMP1914]], i32 [[TMP1910]], i32 [[TMP1913]]
+// CHECK-NEXT:    store volatile i32 [[TMP1915]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP1916:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1917:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1918:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1916]], i32 [[TMP1917]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1919:%.*]] = extractvalue { i32, i1 } [[TMP1918]], 0
+// CHECK-NEXT:    [[TMP1920:%.*]] = extractvalue { i32, i1 } [[TMP1918]], 1
+// CHECK-NEXT:    br i1 [[TMP1920]], label [[IX_ATOMIC_EXIT:%.*]], label [[IX_ATOMIC_CONT:%.*]]
+// CHECK:       ix.atomic.cont:
+// CHECK-NEXT:    store i32 [[TMP1919]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT]]
+// CHECK:       ix.atomic.exit:
+// CHECK-NEXT:    [[TMP1921:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1922:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1923:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1921]], i32 [[TMP1922]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1924:%.*]] = extractvalue { i32, i1 } [[TMP1923]], 0
+// CHECK-NEXT:    [[TMP1925:%.*]] = extractvalue { i32, i1 } [[TMP1923]], 1
+// CHECK-NEXT:    br i1 [[TMP1925]], label [[IX_ATOMIC_EXIT185:%.*]], label [[IX_ATOMIC_CONT186:%.*]]
+// CHECK:       ix.atomic.cont186:
+// CHECK-NEXT:    store i32 [[TMP1924]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT185]]
+// CHECK:       ix.atomic.exit185:
+// CHECK-NEXT:    [[TMP1926:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1927:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1928:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1926]], i32 [[TMP1927]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1929:%.*]] = extractvalue { i32, i1 } [[TMP1928]], 1
+// CHECK-NEXT:    [[TMP1930:%.*]] = zext i1 [[TMP1929]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP1930]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP1931:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1932:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1933:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1931]], i32 [[TMP1932]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1934:%.*]] = extractvalue { i32, i1 } [[TMP1933]], 1
+// CHECK-NEXT:    [[TMP1935:%.*]] = zext i1 [[TMP1934]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP1935]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP1936:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1937:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1938:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1936]], i32 [[TMP1937]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1939:%.*]] = extractvalue { i32, i1 } [[TMP1938]], 0
+// CHECK-NEXT:    [[TMP1940:%.*]] = extractvalue { i32, i1 } [[TMP1938]], 1
+// CHECK-NEXT:    br i1 [[TMP1940]], label [[IX_ATOMIC_EXIT187:%.*]], label [[IX_ATOMIC_CONT188:%.*]]
+// CHECK:       ix.atomic.cont188:
+// CHECK-NEXT:    store i32 [[TMP1939]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT187]]
+// CHECK:       ix.atomic.exit187:
+// CHECK-NEXT:    [[TMP1941:%.*]] = extractvalue { i32, i1 } [[TMP1938]], 1
+// CHECK-NEXT:    [[TMP1942:%.*]] = zext i1 [[TMP1941]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP1942]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP1943:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1944:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1945:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1943]], i32 [[TMP1944]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP1946:%.*]] = extractvalue { i32, i1 } [[TMP1945]], 0
+// CHECK-NEXT:    [[TMP1947:%.*]] = extractvalue { i32, i1 } [[TMP1945]], 1
+// CHECK-NEXT:    br i1 [[TMP1947]], label [[IX_ATOMIC_EXIT189:%.*]], label [[IX_ATOMIC_CONT190:%.*]]
+// CHECK:       ix.atomic.cont190:
+// CHECK-NEXT:    store i32 [[TMP1946]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT189]]
+// CHECK:       ix.atomic.exit189:
+// CHECK-NEXT:    [[TMP1948:%.*]] = extractvalue { i32, i1 } [[TMP1945]], 1
+// CHECK-NEXT:    [[TMP1949:%.*]] = zext i1 [[TMP1948]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP1949]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP1950:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1951:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1950]] acq_rel, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1951]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1952:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1953:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1952]] acq_rel, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1953]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1954:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1955:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1954]] acq_rel, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1955]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1956:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1957:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1956]] acq_rel, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP1957]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1958:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1959:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1960:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1958]], i32 [[TMP1959]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP1961:%.*]] = extractvalue { i32, i1 } [[TMP1960]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP1961]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1962:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1963:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1964:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1962]], i32 [[TMP1963]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP1965:%.*]] = extractvalue { i32, i1 } [[TMP1964]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP1965]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1966:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1967:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1966]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP1968:%.*]] = icmp ugt i32 [[TMP1967]], [[TMP1966]]
+// CHECK-NEXT:    [[TMP1969:%.*]] = select i1 [[TMP1968]], i32 [[TMP1966]], i32 [[TMP1967]]
+// CHECK-NEXT:    store volatile i32 [[TMP1969]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1970:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1971:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1970]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP1972:%.*]] = icmp ult i32 [[TMP1971]], [[TMP1970]]
+// CHECK-NEXT:    [[TMP1973:%.*]] = select i1 [[TMP1972]], i32 [[TMP1970]], i32 [[TMP1971]]
+// CHECK-NEXT:    store volatile i32 [[TMP1973]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1974:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1975:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP1974]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP1976:%.*]] = icmp ult i32 [[TMP1975]], [[TMP1974]]
+// CHECK-NEXT:    [[TMP1977:%.*]] = select i1 [[TMP1976]], i32 [[TMP1974]], i32 [[TMP1975]]
+// CHECK-NEXT:    store volatile i32 [[TMP1977]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1978:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1979:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP1978]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP1980:%.*]] = icmp ugt i32 [[TMP1979]], [[TMP1978]]
+// CHECK-NEXT:    [[TMP1981:%.*]] = select i1 [[TMP1980]], i32 [[TMP1978]], i32 [[TMP1979]]
+// CHECK-NEXT:    store volatile i32 [[TMP1981]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1982:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1983:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1984:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1982]], i32 [[TMP1983]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP1985:%.*]] = extractvalue { i32, i1 } [[TMP1984]], 0
+// CHECK-NEXT:    [[TMP1986:%.*]] = extractvalue { i32, i1 } [[TMP1984]], 1
+// CHECK-NEXT:    [[TMP1987:%.*]] = select i1 [[TMP1986]], i32 [[TMP1982]], i32 [[TMP1985]]
+// CHECK-NEXT:    store volatile i32 [[TMP1987]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1988:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1989:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1990:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1988]], i32 [[TMP1989]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP1991:%.*]] = extractvalue { i32, i1 } [[TMP1990]], 0
+// CHECK-NEXT:    [[TMP1992:%.*]] = extractvalue { i32, i1 } [[TMP1990]], 1
+// CHECK-NEXT:    [[TMP1993:%.*]] = select i1 [[TMP1992]], i32 [[TMP1988]], i32 [[TMP1991]]
+// CHECK-NEXT:    store volatile i32 [[TMP1993]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1994:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP1995:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP1996:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1994]], i32 [[TMP1995]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP1997:%.*]] = extractvalue { i32, i1 } [[TMP1996]], 0
+// CHECK-NEXT:    [[TMP1998:%.*]] = extractvalue { i32, i1 } [[TMP1996]], 1
+// CHECK-NEXT:    br i1 [[TMP1998]], label [[IX_ATOMIC_EXIT191:%.*]], label [[IX_ATOMIC_CONT192:%.*]]
+// CHECK:       ix.atomic.cont192:
+// CHECK-NEXT:    store i32 [[TMP1997]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT191]]
+// CHECK:       ix.atomic.exit191:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP1999:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2000:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2001:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP1999]], i32 [[TMP2000]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2002:%.*]] = extractvalue { i32, i1 } [[TMP2001]], 0
+// CHECK-NEXT:    [[TMP2003:%.*]] = extractvalue { i32, i1 } [[TMP2001]], 1
+// CHECK-NEXT:    br i1 [[TMP2003]], label [[IX_ATOMIC_EXIT193:%.*]], label [[IX_ATOMIC_CONT194:%.*]]
+// CHECK:       ix.atomic.cont194:
+// CHECK-NEXT:    store i32 [[TMP2002]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT193]]
+// CHECK:       ix.atomic.exit193:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2004:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2005:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2006:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2004]], i32 [[TMP2005]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2007:%.*]] = extractvalue { i32, i1 } [[TMP2006]], 1
+// CHECK-NEXT:    [[TMP2008:%.*]] = zext i1 [[TMP2007]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2008]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2009:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2010:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2011:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2009]], i32 [[TMP2010]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2012:%.*]] = extractvalue { i32, i1 } [[TMP2011]], 1
+// CHECK-NEXT:    [[TMP2013:%.*]] = zext i1 [[TMP2012]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2013]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2014:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2015:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2016:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2014]], i32 [[TMP2015]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2017:%.*]] = extractvalue { i32, i1 } [[TMP2016]], 0
+// CHECK-NEXT:    [[TMP2018:%.*]] = extractvalue { i32, i1 } [[TMP2016]], 1
+// CHECK-NEXT:    br i1 [[TMP2018]], label [[IX_ATOMIC_EXIT195:%.*]], label [[IX_ATOMIC_CONT196:%.*]]
+// CHECK:       ix.atomic.cont196:
+// CHECK-NEXT:    store i32 [[TMP2017]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT195]]
+// CHECK:       ix.atomic.exit195:
+// CHECK-NEXT:    [[TMP2019:%.*]] = extractvalue { i32, i1 } [[TMP2016]], 1
+// CHECK-NEXT:    [[TMP2020:%.*]] = zext i1 [[TMP2019]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2020]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2021:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2022:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2023:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2021]], i32 [[TMP2022]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2024:%.*]] = extractvalue { i32, i1 } [[TMP2023]], 0
+// CHECK-NEXT:    [[TMP2025:%.*]] = extractvalue { i32, i1 } [[TMP2023]], 1
+// CHECK-NEXT:    br i1 [[TMP2025]], label [[IX_ATOMIC_EXIT197:%.*]], label [[IX_ATOMIC_CONT198:%.*]]
+// CHECK:       ix.atomic.cont198:
+// CHECK-NEXT:    store i32 [[TMP2024]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT197]]
+// CHECK:       ix.atomic.exit197:
+// CHECK-NEXT:    [[TMP2026:%.*]] = extractvalue { i32, i1 } [[TMP2023]], 1
+// CHECK-NEXT:    [[TMP2027:%.*]] = zext i1 [[TMP2026]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2027]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2028:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2029:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2028]] acquire, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2029]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2030:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2031:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2030]] acquire, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2031]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2032:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2033:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2032]] acquire, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2033]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2034:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2035:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2034]] acquire, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2035]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2036:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2037:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2038:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2036]], i32 [[TMP2037]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2039:%.*]] = extractvalue { i32, i1 } [[TMP2038]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2039]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2040:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2041:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2042:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2040]], i32 [[TMP2041]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2043:%.*]] = extractvalue { i32, i1 } [[TMP2042]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2043]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2044:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2045:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2044]] acquire, align 4
+// CHECK-NEXT:    [[TMP2046:%.*]] = icmp ugt i32 [[TMP2045]], [[TMP2044]]
+// CHECK-NEXT:    [[TMP2047:%.*]] = select i1 [[TMP2046]], i32 [[TMP2044]], i32 [[TMP2045]]
+// CHECK-NEXT:    store volatile i32 [[TMP2047]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2048:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2049:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2048]] acquire, align 4
+// CHECK-NEXT:    [[TMP2050:%.*]] = icmp ult i32 [[TMP2049]], [[TMP2048]]
+// CHECK-NEXT:    [[TMP2051:%.*]] = select i1 [[TMP2050]], i32 [[TMP2048]], i32 [[TMP2049]]
+// CHECK-NEXT:    store volatile i32 [[TMP2051]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2052:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2053:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2052]] acquire, align 4
+// CHECK-NEXT:    [[TMP2054:%.*]] = icmp ult i32 [[TMP2053]], [[TMP2052]]
+// CHECK-NEXT:    [[TMP2055:%.*]] = select i1 [[TMP2054]], i32 [[TMP2052]], i32 [[TMP2053]]
+// CHECK-NEXT:    store volatile i32 [[TMP2055]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2056:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2057:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2056]] acquire, align 4
+// CHECK-NEXT:    [[TMP2058:%.*]] = icmp ugt i32 [[TMP2057]], [[TMP2056]]
+// CHECK-NEXT:    [[TMP2059:%.*]] = select i1 [[TMP2058]], i32 [[TMP2056]], i32 [[TMP2057]]
+// CHECK-NEXT:    store volatile i32 [[TMP2059]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2060:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2061:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2062:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2060]], i32 [[TMP2061]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2063:%.*]] = extractvalue { i32, i1 } [[TMP2062]], 0
+// CHECK-NEXT:    [[TMP2064:%.*]] = extractvalue { i32, i1 } [[TMP2062]], 1
+// CHECK-NEXT:    [[TMP2065:%.*]] = select i1 [[TMP2064]], i32 [[TMP2060]], i32 [[TMP2063]]
+// CHECK-NEXT:    store volatile i32 [[TMP2065]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2066:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2067:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2068:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2066]], i32 [[TMP2067]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2069:%.*]] = extractvalue { i32, i1 } [[TMP2068]], 0
+// CHECK-NEXT:    [[TMP2070:%.*]] = extractvalue { i32, i1 } [[TMP2068]], 1
+// CHECK-NEXT:    [[TMP2071:%.*]] = select i1 [[TMP2070]], i32 [[TMP2066]], i32 [[TMP2069]]
+// CHECK-NEXT:    store volatile i32 [[TMP2071]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2072:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2073:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2074:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2072]], i32 [[TMP2073]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2075:%.*]] = extractvalue { i32, i1 } [[TMP2074]], 0
+// CHECK-NEXT:    [[TMP2076:%.*]] = extractvalue { i32, i1 } [[TMP2074]], 1
+// CHECK-NEXT:    br i1 [[TMP2076]], label [[IX_ATOMIC_EXIT199:%.*]], label [[IX_ATOMIC_CONT200:%.*]]
+// CHECK:       ix.atomic.cont200:
+// CHECK-NEXT:    store i32 [[TMP2075]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT199]]
+// CHECK:       ix.atomic.exit199:
+// CHECK-NEXT:    [[TMP2077:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2078:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2079:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2077]], i32 [[TMP2078]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2080:%.*]] = extractvalue { i32, i1 } [[TMP2079]], 0
+// CHECK-NEXT:    [[TMP2081:%.*]] = extractvalue { i32, i1 } [[TMP2079]], 1
+// CHECK-NEXT:    br i1 [[TMP2081]], label [[IX_ATOMIC_EXIT201:%.*]], label [[IX_ATOMIC_CONT202:%.*]]
+// CHECK:       ix.atomic.cont202:
+// CHECK-NEXT:    store i32 [[TMP2080]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT201]]
+// CHECK:       ix.atomic.exit201:
+// CHECK-NEXT:    [[TMP2082:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2083:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2084:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2082]], i32 [[TMP2083]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2085:%.*]] = extractvalue { i32, i1 } [[TMP2084]], 1
+// CHECK-NEXT:    [[TMP2086:%.*]] = zext i1 [[TMP2085]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2086]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2087:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2088:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2089:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2087]], i32 [[TMP2088]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2090:%.*]] = extractvalue { i32, i1 } [[TMP2089]], 1
+// CHECK-NEXT:    [[TMP2091:%.*]] = zext i1 [[TMP2090]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2091]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2092:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2093:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2094:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2092]], i32 [[TMP2093]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2095:%.*]] = extractvalue { i32, i1 } [[TMP2094]], 0
+// CHECK-NEXT:    [[TMP2096:%.*]] = extractvalue { i32, i1 } [[TMP2094]], 1
+// CHECK-NEXT:    br i1 [[TMP2096]], label [[IX_ATOMIC_EXIT203:%.*]], label [[IX_ATOMIC_CONT204:%.*]]
+// CHECK:       ix.atomic.cont204:
+// CHECK-NEXT:    store i32 [[TMP2095]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT203]]
+// CHECK:       ix.atomic.exit203:
+// CHECK-NEXT:    [[TMP2097:%.*]] = extractvalue { i32, i1 } [[TMP2094]], 1
+// CHECK-NEXT:    [[TMP2098:%.*]] = zext i1 [[TMP2097]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2098]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2099:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2100:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2101:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2099]], i32 [[TMP2100]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2102:%.*]] = extractvalue { i32, i1 } [[TMP2101]], 0
+// CHECK-NEXT:    [[TMP2103:%.*]] = extractvalue { i32, i1 } [[TMP2101]], 1
+// CHECK-NEXT:    br i1 [[TMP2103]], label [[IX_ATOMIC_EXIT205:%.*]], label [[IX_ATOMIC_CONT206:%.*]]
+// CHECK:       ix.atomic.cont206:
+// CHECK-NEXT:    store i32 [[TMP2102]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT205]]
+// CHECK:       ix.atomic.exit205:
+// CHECK-NEXT:    [[TMP2104:%.*]] = extractvalue { i32, i1 } [[TMP2101]], 1
+// CHECK-NEXT:    [[TMP2105:%.*]] = zext i1 [[TMP2104]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2105]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2106:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2107:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2106]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2107]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2108:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2109:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2108]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2109]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2110:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2111:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2110]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2111]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2112:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2113:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2112]] monotonic, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2113]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2114:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2115:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2116:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2114]], i32 [[TMP2115]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2117:%.*]] = extractvalue { i32, i1 } [[TMP2116]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2117]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2118:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2119:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2120:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2118]], i32 [[TMP2119]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2121:%.*]] = extractvalue { i32, i1 } [[TMP2120]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2121]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2122:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2123:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2122]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2124:%.*]] = icmp ugt i32 [[TMP2123]], [[TMP2122]]
+// CHECK-NEXT:    [[TMP2125:%.*]] = select i1 [[TMP2124]], i32 [[TMP2122]], i32 [[TMP2123]]
+// CHECK-NEXT:    store volatile i32 [[TMP2125]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2126:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2127:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2126]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2128:%.*]] = icmp ult i32 [[TMP2127]], [[TMP2126]]
+// CHECK-NEXT:    [[TMP2129:%.*]] = select i1 [[TMP2128]], i32 [[TMP2126]], i32 [[TMP2127]]
+// CHECK-NEXT:    store volatile i32 [[TMP2129]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2130:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2131:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2130]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2132:%.*]] = icmp ult i32 [[TMP2131]], [[TMP2130]]
+// CHECK-NEXT:    [[TMP2133:%.*]] = select i1 [[TMP2132]], i32 [[TMP2130]], i32 [[TMP2131]]
+// CHECK-NEXT:    store volatile i32 [[TMP2133]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2134:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2135:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2134]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2136:%.*]] = icmp ugt i32 [[TMP2135]], [[TMP2134]]
+// CHECK-NEXT:    [[TMP2137:%.*]] = select i1 [[TMP2136]], i32 [[TMP2134]], i32 [[TMP2135]]
+// CHECK-NEXT:    store volatile i32 [[TMP2137]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2138:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2139:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2140:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2138]], i32 [[TMP2139]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2141:%.*]] = extractvalue { i32, i1 } [[TMP2140]], 0
+// CHECK-NEXT:    [[TMP2142:%.*]] = extractvalue { i32, i1 } [[TMP2140]], 1
+// CHECK-NEXT:    [[TMP2143:%.*]] = select i1 [[TMP2142]], i32 [[TMP2138]], i32 [[TMP2141]]
+// CHECK-NEXT:    store volatile i32 [[TMP2143]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2144:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2145:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2146:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2144]], i32 [[TMP2145]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2147:%.*]] = extractvalue { i32, i1 } [[TMP2146]], 0
+// CHECK-NEXT:    [[TMP2148:%.*]] = extractvalue { i32, i1 } [[TMP2146]], 1
+// CHECK-NEXT:    [[TMP2149:%.*]] = select i1 [[TMP2148]], i32 [[TMP2144]], i32 [[TMP2147]]
+// CHECK-NEXT:    store volatile i32 [[TMP2149]], i32* [[IV]], align 4
+// CHECK-NEXT:    [[TMP2150:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2151:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2152:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2150]], i32 [[TMP2151]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2153:%.*]] = extractvalue { i32, i1 } [[TMP2152]], 0
+// CHECK-NEXT:    [[TMP2154:%.*]] = extractvalue { i32, i1 } [[TMP2152]], 1
+// CHECK-NEXT:    br i1 [[TMP2154]], label [[IX_ATOMIC_EXIT207:%.*]], label [[IX_ATOMIC_CONT208:%.*]]
+// CHECK:       ix.atomic.cont208:
+// CHECK-NEXT:    store i32 [[TMP2153]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT207]]
+// CHECK:       ix.atomic.exit207:
+// CHECK-NEXT:    [[TMP2155:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2156:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2157:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2155]], i32 [[TMP2156]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2158:%.*]] = extractvalue { i32, i1 } [[TMP2157]], 0
+// CHECK-NEXT:    [[TMP2159:%.*]] = extractvalue { i32, i1 } [[TMP2157]], 1
+// CHECK-NEXT:    br i1 [[TMP2159]], label [[IX_ATOMIC_EXIT209:%.*]], label [[IX_ATOMIC_CONT210:%.*]]
+// CHECK:       ix.atomic.cont210:
+// CHECK-NEXT:    store i32 [[TMP2158]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT209]]
+// CHECK:       ix.atomic.exit209:
+// CHECK-NEXT:    [[TMP2160:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2161:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2162:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2160]], i32 [[TMP2161]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2163:%.*]] = extractvalue { i32, i1 } [[TMP2162]], 1
+// CHECK-NEXT:    [[TMP2164:%.*]] = zext i1 [[TMP2163]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2164]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2165:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2166:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2167:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2165]], i32 [[TMP2166]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2168:%.*]] = extractvalue { i32, i1 } [[TMP2167]], 1
+// CHECK-NEXT:    [[TMP2169:%.*]] = zext i1 [[TMP2168]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2169]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2170:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2171:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2172:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2170]], i32 [[TMP2171]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2173:%.*]] = extractvalue { i32, i1 } [[TMP2172]], 0
+// CHECK-NEXT:    [[TMP2174:%.*]] = extractvalue { i32, i1 } [[TMP2172]], 1
+// CHECK-NEXT:    br i1 [[TMP2174]], label [[IX_ATOMIC_EXIT211:%.*]], label [[IX_ATOMIC_CONT212:%.*]]
+// CHECK:       ix.atomic.cont212:
+// CHECK-NEXT:    store i32 [[TMP2173]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT211]]
+// CHECK:       ix.atomic.exit211:
+// CHECK-NEXT:    [[TMP2175:%.*]] = extractvalue { i32, i1 } [[TMP2172]], 1
+// CHECK-NEXT:    [[TMP2176:%.*]] = zext i1 [[TMP2175]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2176]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2177:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2178:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2179:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2177]], i32 [[TMP2178]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2180:%.*]] = extractvalue { i32, i1 } [[TMP2179]], 0
+// CHECK-NEXT:    [[TMP2181:%.*]] = extractvalue { i32, i1 } [[TMP2179]], 1
+// CHECK-NEXT:    br i1 [[TMP2181]], label [[IX_ATOMIC_EXIT213:%.*]], label [[IX_ATOMIC_CONT214:%.*]]
+// CHECK:       ix.atomic.cont214:
+// CHECK-NEXT:    store i32 [[TMP2180]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT213]]
+// CHECK:       ix.atomic.exit213:
+// CHECK-NEXT:    [[TMP2182:%.*]] = extractvalue { i32, i1 } [[TMP2179]], 1
+// CHECK-NEXT:    [[TMP2183:%.*]] = zext i1 [[TMP2182]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2183]], i32* [[IR]], align 4
+// CHECK-NEXT:    [[TMP2184:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2185:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2184]] release, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2185]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2186:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2187:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2186]] release, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2187]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2188:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2189:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2188]] release, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2189]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2190:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2191:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2190]] release, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2191]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2192:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2193:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2194:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2192]], i32 [[TMP2193]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2195:%.*]] = extractvalue { i32, i1 } [[TMP2194]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2195]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2196:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2197:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2198:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2196]], i32 [[TMP2197]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2199:%.*]] = extractvalue { i32, i1 } [[TMP2198]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2199]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2200:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2201:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2200]] release, align 4
+// CHECK-NEXT:    [[TMP2202:%.*]] = icmp ugt i32 [[TMP2201]], [[TMP2200]]
+// CHECK-NEXT:    [[TMP2203:%.*]] = select i1 [[TMP2202]], i32 [[TMP2200]], i32 [[TMP2201]]
+// CHECK-NEXT:    store volatile i32 [[TMP2203]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2204:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2205:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2204]] release, align 4
+// CHECK-NEXT:    [[TMP2206:%.*]] = icmp ult i32 [[TMP2205]], [[TMP2204]]
+// CHECK-NEXT:    [[TMP2207:%.*]] = select i1 [[TMP2206]], i32 [[TMP2204]], i32 [[TMP2205]]
+// CHECK-NEXT:    store volatile i32 [[TMP2207]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2208:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2209:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2208]] release, align 4
+// CHECK-NEXT:    [[TMP2210:%.*]] = icmp ult i32 [[TMP2209]], [[TMP2208]]
+// CHECK-NEXT:    [[TMP2211:%.*]] = select i1 [[TMP2210]], i32 [[TMP2208]], i32 [[TMP2209]]
+// CHECK-NEXT:    store volatile i32 [[TMP2211]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2212:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2213:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2212]] release, align 4
+// CHECK-NEXT:    [[TMP2214:%.*]] = icmp ugt i32 [[TMP2213]], [[TMP2212]]
+// CHECK-NEXT:    [[TMP2215:%.*]] = select i1 [[TMP2214]], i32 [[TMP2212]], i32 [[TMP2213]]
+// CHECK-NEXT:    store volatile i32 [[TMP2215]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2216:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2217:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2218:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2216]], i32 [[TMP2217]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2219:%.*]] = extractvalue { i32, i1 } [[TMP2218]], 0
+// CHECK-NEXT:    [[TMP2220:%.*]] = extractvalue { i32, i1 } [[TMP2218]], 1
+// CHECK-NEXT:    [[TMP2221:%.*]] = select i1 [[TMP2220]], i32 [[TMP2216]], i32 [[TMP2219]]
+// CHECK-NEXT:    store volatile i32 [[TMP2221]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2222:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2223:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2224:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2222]], i32 [[TMP2223]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2225:%.*]] = extractvalue { i32, i1 } [[TMP2224]], 0
+// CHECK-NEXT:    [[TMP2226:%.*]] = extractvalue { i32, i1 } [[TMP2224]], 1
+// CHECK-NEXT:    [[TMP2227:%.*]] = select i1 [[TMP2226]], i32 [[TMP2222]], i32 [[TMP2225]]
+// CHECK-NEXT:    store volatile i32 [[TMP2227]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2228:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2229:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2230:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2228]], i32 [[TMP2229]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2231:%.*]] = extractvalue { i32, i1 } [[TMP2230]], 0
+// CHECK-NEXT:    [[TMP2232:%.*]] = extractvalue { i32, i1 } [[TMP2230]], 1
+// CHECK-NEXT:    br i1 [[TMP2232]], label [[IX_ATOMIC_EXIT215:%.*]], label [[IX_ATOMIC_CONT216:%.*]]
+// CHECK:       ix.atomic.cont216:
+// CHECK-NEXT:    store i32 [[TMP2231]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT215]]
+// CHECK:       ix.atomic.exit215:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2233:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2234:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2235:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2233]], i32 [[TMP2234]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2236:%.*]] = extractvalue { i32, i1 } [[TMP2235]], 0
+// CHECK-NEXT:    [[TMP2237:%.*]] = extractvalue { i32, i1 } [[TMP2235]], 1
+// CHECK-NEXT:    br i1 [[TMP2237]], label [[IX_ATOMIC_EXIT217:%.*]], label [[IX_ATOMIC_CONT218:%.*]]
+// CHECK:       ix.atomic.cont218:
+// CHECK-NEXT:    store i32 [[TMP2236]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT217]]
+// CHECK:       ix.atomic.exit217:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2238:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2239:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2240:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2238]], i32 [[TMP2239]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2241:%.*]] = extractvalue { i32, i1 } [[TMP2240]], 1
+// CHECK-NEXT:    [[TMP2242:%.*]] = zext i1 [[TMP2241]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2242]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2243:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2244:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2245:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2243]], i32 [[TMP2244]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2246:%.*]] = extractvalue { i32, i1 } [[TMP2245]], 1
+// CHECK-NEXT:    [[TMP2247:%.*]] = zext i1 [[TMP2246]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2247]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2248:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2249:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2250:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2248]], i32 [[TMP2249]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2251:%.*]] = extractvalue { i32, i1 } [[TMP2250]], 0
+// CHECK-NEXT:    [[TMP2252:%.*]] = extractvalue { i32, i1 } [[TMP2250]], 1
+// CHECK-NEXT:    br i1 [[TMP2252]], label [[IX_ATOMIC_EXIT219:%.*]], label [[IX_ATOMIC_CONT220:%.*]]
+// CHECK:       ix.atomic.cont220:
+// CHECK-NEXT:    store i32 [[TMP2251]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT219]]
+// CHECK:       ix.atomic.exit219:
+// CHECK-NEXT:    [[TMP2253:%.*]] = extractvalue { i32, i1 } [[TMP2250]], 1
+// CHECK-NEXT:    [[TMP2254:%.*]] = zext i1 [[TMP2253]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2254]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2255:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2256:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2257:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2255]], i32 [[TMP2256]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2258:%.*]] = extractvalue { i32, i1 } [[TMP2257]], 0
+// CHECK-NEXT:    [[TMP2259:%.*]] = extractvalue { i32, i1 } [[TMP2257]], 1
+// CHECK-NEXT:    br i1 [[TMP2259]], label [[IX_ATOMIC_EXIT221:%.*]], label [[IX_ATOMIC_CONT222:%.*]]
+// CHECK:       ix.atomic.cont222:
+// CHECK-NEXT:    store i32 [[TMP2258]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT221]]
+// CHECK:       ix.atomic.exit221:
+// CHECK-NEXT:    [[TMP2260:%.*]] = extractvalue { i32, i1 } [[TMP2257]], 1
+// CHECK-NEXT:    [[TMP2261:%.*]] = zext i1 [[TMP2260]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2261]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2262:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2263:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2262]] seq_cst, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2263]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2264:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2265:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2264]] seq_cst, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2265]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2266:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2267:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2266]] seq_cst, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2267]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2268:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2269:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2268]] seq_cst, align 4
+// CHECK-NEXT:    store volatile i32 [[TMP2269]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2270:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2271:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2272:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2270]], i32 [[TMP2271]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2273:%.*]] = extractvalue { i32, i1 } [[TMP2272]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2273]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2274:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2275:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2276:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2274]], i32 [[TMP2275]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2277:%.*]] = extractvalue { i32, i1 } [[TMP2276]], 0
+// CHECK-NEXT:    store volatile i32 [[TMP2277]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2278:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2279:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2278]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2280:%.*]] = icmp ugt i32 [[TMP2279]], [[TMP2278]]
+// CHECK-NEXT:    [[TMP2281:%.*]] = select i1 [[TMP2280]], i32 [[TMP2278]], i32 [[TMP2279]]
+// CHECK-NEXT:    store volatile i32 [[TMP2281]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2282:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2283:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2282]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2284:%.*]] = icmp ult i32 [[TMP2283]], [[TMP2282]]
+// CHECK-NEXT:    [[TMP2285:%.*]] = select i1 [[TMP2284]], i32 [[TMP2282]], i32 [[TMP2283]]
+// CHECK-NEXT:    store volatile i32 [[TMP2285]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2286:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2287:%.*]] = atomicrmw umin i32* [[IX]], i32 [[TMP2286]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2288:%.*]] = icmp ult i32 [[TMP2287]], [[TMP2286]]
+// CHECK-NEXT:    [[TMP2289:%.*]] = select i1 [[TMP2288]], i32 [[TMP2286]], i32 [[TMP2287]]
+// CHECK-NEXT:    store volatile i32 [[TMP2289]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2290:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2291:%.*]] = atomicrmw umax i32* [[IX]], i32 [[TMP2290]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2292:%.*]] = icmp ugt i32 [[TMP2291]], [[TMP2290]]
+// CHECK-NEXT:    [[TMP2293:%.*]] = select i1 [[TMP2292]], i32 [[TMP2290]], i32 [[TMP2291]]
+// CHECK-NEXT:    store volatile i32 [[TMP2293]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2294:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2295:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2296:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2294]], i32 [[TMP2295]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2297:%.*]] = extractvalue { i32, i1 } [[TMP2296]], 0
+// CHECK-NEXT:    [[TMP2298:%.*]] = extractvalue { i32, i1 } [[TMP2296]], 1
+// CHECK-NEXT:    [[TMP2299:%.*]] = select i1 [[TMP2298]], i32 [[TMP2294]], i32 [[TMP2297]]
+// CHECK-NEXT:    store volatile i32 [[TMP2299]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2300:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2301:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2302:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2300]], i32 [[TMP2301]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2303:%.*]] = extractvalue { i32, i1 } [[TMP2302]], 0
+// CHECK-NEXT:    [[TMP2304:%.*]] = extractvalue { i32, i1 } [[TMP2302]], 1
+// CHECK-NEXT:    [[TMP2305:%.*]] = select i1 [[TMP2304]], i32 [[TMP2300]], i32 [[TMP2303]]
+// CHECK-NEXT:    store volatile i32 [[TMP2305]], i32* [[IV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2306:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2307:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2308:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2306]], i32 [[TMP2307]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2309:%.*]] = extractvalue { i32, i1 } [[TMP2308]], 0
+// CHECK-NEXT:    [[TMP2310:%.*]] = extractvalue { i32, i1 } [[TMP2308]], 1
+// CHECK-NEXT:    br i1 [[TMP2310]], label [[IX_ATOMIC_EXIT223:%.*]], label [[IX_ATOMIC_CONT224:%.*]]
+// CHECK:       ix.atomic.cont224:
+// CHECK-NEXT:    store i32 [[TMP2309]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT223]]
+// CHECK:       ix.atomic.exit223:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2311:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2312:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2313:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2311]], i32 [[TMP2312]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2314:%.*]] = extractvalue { i32, i1 } [[TMP2313]], 0
+// CHECK-NEXT:    [[TMP2315:%.*]] = extractvalue { i32, i1 } [[TMP2313]], 1
+// CHECK-NEXT:    br i1 [[TMP2315]], label [[IX_ATOMIC_EXIT225:%.*]], label [[IX_ATOMIC_CONT226:%.*]]
+// CHECK:       ix.atomic.cont226:
+// CHECK-NEXT:    store i32 [[TMP2314]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT225]]
+// CHECK:       ix.atomic.exit225:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2316:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2317:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2318:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2316]], i32 [[TMP2317]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2319:%.*]] = extractvalue { i32, i1 } [[TMP2318]], 1
+// CHECK-NEXT:    [[TMP2320:%.*]] = zext i1 [[TMP2319]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2320]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2321:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2322:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2323:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2321]], i32 [[TMP2322]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2324:%.*]] = extractvalue { i32, i1 } [[TMP2323]], 1
+// CHECK-NEXT:    [[TMP2325:%.*]] = zext i1 [[TMP2324]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2325]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2326:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2327:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2328:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2326]], i32 [[TMP2327]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2329:%.*]] = extractvalue { i32, i1 } [[TMP2328]], 0
+// CHECK-NEXT:    [[TMP2330:%.*]] = extractvalue { i32, i1 } [[TMP2328]], 1
+// CHECK-NEXT:    br i1 [[TMP2330]], label [[IX_ATOMIC_EXIT227:%.*]], label [[IX_ATOMIC_CONT228:%.*]]
+// CHECK:       ix.atomic.cont228:
+// CHECK-NEXT:    store i32 [[TMP2329]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT227]]
+// CHECK:       ix.atomic.exit227:
+// CHECK-NEXT:    [[TMP2331:%.*]] = extractvalue { i32, i1 } [[TMP2328]], 1
+// CHECK-NEXT:    [[TMP2332:%.*]] = zext i1 [[TMP2331]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2332]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2333:%.*]] = load i32, i32* [[IE]], align 4
+// CHECK-NEXT:    [[TMP2334:%.*]] = load i32, i32* [[ID]], align 4
+// CHECK-NEXT:    [[TMP2335:%.*]] = cmpxchg i32* [[IX]], i32 [[TMP2333]], i32 [[TMP2334]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2336:%.*]] = extractvalue { i32, i1 } [[TMP2335]], 0
+// CHECK-NEXT:    [[TMP2337:%.*]] = extractvalue { i32, i1 } [[TMP2335]], 1
+// CHECK-NEXT:    br i1 [[TMP2337]], label [[IX_ATOMIC_EXIT229:%.*]], label [[IX_ATOMIC_CONT230:%.*]]
+// CHECK:       ix.atomic.cont230:
+// CHECK-NEXT:    store i32 [[TMP2336]], i32* [[IV]], align 4
+// CHECK-NEXT:    br label [[IX_ATOMIC_EXIT229]]
+// CHECK:       ix.atomic.exit229:
+// CHECK-NEXT:    [[TMP2338:%.*]] = extractvalue { i32, i1 } [[TMP2335]], 1
+// CHECK-NEXT:    [[TMP2339:%.*]] = zext i1 [[TMP2338]] to i32
+// CHECK-NEXT:    store volatile i32 [[TMP2339]], i32* [[IR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2340:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2341:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2340]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2341]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2342:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2343:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2342]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2343]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2344:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2345:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2344]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2345]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2346:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2347:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2346]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2347]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2348:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2349:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2350:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2348]], i32 [[TMP2349]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2351:%.*]] = extractvalue { i32, i1 } [[TMP2350]], 0
+// CHECK-NEXT:    store i32 [[TMP2351]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2352:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2353:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2354:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2352]], i32 [[TMP2353]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2355:%.*]] = extractvalue { i32, i1 } [[TMP2354]], 0
+// CHECK-NEXT:    store i32 [[TMP2355]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2356:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2357:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2356]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2358:%.*]] = icmp ugt i32 [[TMP2357]], [[TMP2356]]
+// CHECK-NEXT:    [[TMP2359:%.*]] = select i1 [[TMP2358]], i32 [[TMP2356]], i32 [[TMP2357]]
+// CHECK-NEXT:    store i32 [[TMP2359]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2360:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2361:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2360]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2362:%.*]] = icmp ult i32 [[TMP2361]], [[TMP2360]]
+// CHECK-NEXT:    [[TMP2363:%.*]] = select i1 [[TMP2362]], i32 [[TMP2360]], i32 [[TMP2361]]
+// CHECK-NEXT:    store i32 [[TMP2363]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2364:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2365:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2364]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2366:%.*]] = icmp ult i32 [[TMP2365]], [[TMP2364]]
+// CHECK-NEXT:    [[TMP2367:%.*]] = select i1 [[TMP2366]], i32 [[TMP2364]], i32 [[TMP2365]]
+// CHECK-NEXT:    store i32 [[TMP2367]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2368:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2369:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2368]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2370:%.*]] = icmp ugt i32 [[TMP2369]], [[TMP2368]]
+// CHECK-NEXT:    [[TMP2371:%.*]] = select i1 [[TMP2370]], i32 [[TMP2368]], i32 [[TMP2369]]
+// CHECK-NEXT:    store i32 [[TMP2371]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2372:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2373:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2374:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2372]], i32 [[TMP2373]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2375:%.*]] = extractvalue { i32, i1 } [[TMP2374]], 0
+// CHECK-NEXT:    [[TMP2376:%.*]] = extractvalue { i32, i1 } [[TMP2374]], 1
+// CHECK-NEXT:    [[TMP2377:%.*]] = select i1 [[TMP2376]], i32 [[TMP2372]], i32 [[TMP2375]]
+// CHECK-NEXT:    store i32 [[TMP2377]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2378:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2379:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2380:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2378]], i32 [[TMP2379]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2381:%.*]] = extractvalue { i32, i1 } [[TMP2380]], 0
+// CHECK-NEXT:    [[TMP2382:%.*]] = extractvalue { i32, i1 } [[TMP2380]], 1
+// CHECK-NEXT:    [[TMP2383:%.*]] = select i1 [[TMP2382]], i32 [[TMP2378]], i32 [[TMP2381]]
+// CHECK-NEXT:    store i32 [[TMP2383]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2384:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2385:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2386:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2384]], i32 [[TMP2385]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2387:%.*]] = extractvalue { i32, i1 } [[TMP2386]], 0
+// CHECK-NEXT:    [[TMP2388:%.*]] = extractvalue { i32, i1 } [[TMP2386]], 1
+// CHECK-NEXT:    br i1 [[TMP2388]], label [[UIX_ATOMIC_EXIT:%.*]], label [[UIX_ATOMIC_CONT:%.*]]
+// CHECK:       uix.atomic.cont:
+// CHECK-NEXT:    store i32 [[TMP2387]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT]]
+// CHECK:       uix.atomic.exit:
+// CHECK-NEXT:    [[TMP2389:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2390:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2391:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2389]], i32 [[TMP2390]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2392:%.*]] = extractvalue { i32, i1 } [[TMP2391]], 0
+// CHECK-NEXT:    [[TMP2393:%.*]] = extractvalue { i32, i1 } [[TMP2391]], 1
+// CHECK-NEXT:    br i1 [[TMP2393]], label [[UIX_ATOMIC_EXIT231:%.*]], label [[UIX_ATOMIC_CONT232:%.*]]
+// CHECK:       uix.atomic.cont232:
+// CHECK-NEXT:    store i32 [[TMP2392]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT231]]
+// CHECK:       uix.atomic.exit231:
+// CHECK-NEXT:    [[TMP2394:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2395:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2396:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2394]], i32 [[TMP2395]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2397:%.*]] = extractvalue { i32, i1 } [[TMP2396]], 1
+// CHECK-NEXT:    [[TMP2398:%.*]] = zext i1 [[TMP2397]] to i32
+// CHECK-NEXT:    store i32 [[TMP2398]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2399:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2400:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2401:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2399]], i32 [[TMP2400]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2402:%.*]] = extractvalue { i32, i1 } [[TMP2401]], 1
+// CHECK-NEXT:    [[TMP2403:%.*]] = zext i1 [[TMP2402]] to i32
+// CHECK-NEXT:    store i32 [[TMP2403]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2404:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2405:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2406:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2404]], i32 [[TMP2405]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2407:%.*]] = extractvalue { i32, i1 } [[TMP2406]], 0
+// CHECK-NEXT:    [[TMP2408:%.*]] = extractvalue { i32, i1 } [[TMP2406]], 1
+// CHECK-NEXT:    br i1 [[TMP2408]], label [[UIX_ATOMIC_EXIT233:%.*]], label [[UIX_ATOMIC_CONT234:%.*]]
+// CHECK:       uix.atomic.cont234:
+// CHECK-NEXT:    store i32 [[TMP2407]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT233]]
+// CHECK:       uix.atomic.exit233:
+// CHECK-NEXT:    [[TMP2409:%.*]] = extractvalue { i32, i1 } [[TMP2406]], 1
+// CHECK-NEXT:    [[TMP2410:%.*]] = zext i1 [[TMP2409]] to i32
+// CHECK-NEXT:    store i32 [[TMP2410]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2411:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2412:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2413:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2411]], i32 [[TMP2412]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2414:%.*]] = extractvalue { i32, i1 } [[TMP2413]], 0
+// CHECK-NEXT:    [[TMP2415:%.*]] = extractvalue { i32, i1 } [[TMP2413]], 1
+// CHECK-NEXT:    br i1 [[TMP2415]], label [[UIX_ATOMIC_EXIT235:%.*]], label [[UIX_ATOMIC_CONT236:%.*]]
+// CHECK:       uix.atomic.cont236:
+// CHECK-NEXT:    store i32 [[TMP2414]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT235]]
+// CHECK:       uix.atomic.exit235:
+// CHECK-NEXT:    [[TMP2416:%.*]] = extractvalue { i32, i1 } [[TMP2413]], 1
+// CHECK-NEXT:    [[TMP2417:%.*]] = zext i1 [[TMP2416]] to i32
+// CHECK-NEXT:    store i32 [[TMP2417]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2418:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2419:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2418]] acq_rel, align 4
+// CHECK-NEXT:    store i32 [[TMP2419]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2420:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2421:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2420]] acq_rel, align 4
+// CHECK-NEXT:    store i32 [[TMP2421]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2422:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2423:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2422]] acq_rel, align 4
+// CHECK-NEXT:    store i32 [[TMP2423]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2424:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2425:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2424]] acq_rel, align 4
+// CHECK-NEXT:    store i32 [[TMP2425]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2426:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2427:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2428:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2426]], i32 [[TMP2427]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2429:%.*]] = extractvalue { i32, i1 } [[TMP2428]], 0
+// CHECK-NEXT:    store i32 [[TMP2429]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2430:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2431:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2432:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2430]], i32 [[TMP2431]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2433:%.*]] = extractvalue { i32, i1 } [[TMP2432]], 0
+// CHECK-NEXT:    store i32 [[TMP2433]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2434:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2435:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2434]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP2436:%.*]] = icmp ugt i32 [[TMP2435]], [[TMP2434]]
+// CHECK-NEXT:    [[TMP2437:%.*]] = select i1 [[TMP2436]], i32 [[TMP2434]], i32 [[TMP2435]]
+// CHECK-NEXT:    store i32 [[TMP2437]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2438:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2439:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2438]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP2440:%.*]] = icmp ult i32 [[TMP2439]], [[TMP2438]]
+// CHECK-NEXT:    [[TMP2441:%.*]] = select i1 [[TMP2440]], i32 [[TMP2438]], i32 [[TMP2439]]
+// CHECK-NEXT:    store i32 [[TMP2441]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2442:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2443:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2442]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP2444:%.*]] = icmp ult i32 [[TMP2443]], [[TMP2442]]
+// CHECK-NEXT:    [[TMP2445:%.*]] = select i1 [[TMP2444]], i32 [[TMP2442]], i32 [[TMP2443]]
+// CHECK-NEXT:    store i32 [[TMP2445]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2446:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2447:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2446]] acq_rel, align 4
+// CHECK-NEXT:    [[TMP2448:%.*]] = icmp ugt i32 [[TMP2447]], [[TMP2446]]
+// CHECK-NEXT:    [[TMP2449:%.*]] = select i1 [[TMP2448]], i32 [[TMP2446]], i32 [[TMP2447]]
+// CHECK-NEXT:    store i32 [[TMP2449]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2450:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2451:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2452:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2450]], i32 [[TMP2451]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2453:%.*]] = extractvalue { i32, i1 } [[TMP2452]], 0
+// CHECK-NEXT:    [[TMP2454:%.*]] = extractvalue { i32, i1 } [[TMP2452]], 1
+// CHECK-NEXT:    [[TMP2455:%.*]] = select i1 [[TMP2454]], i32 [[TMP2450]], i32 [[TMP2453]]
+// CHECK-NEXT:    store i32 [[TMP2455]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2456:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2457:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2458:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2456]], i32 [[TMP2457]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2459:%.*]] = extractvalue { i32, i1 } [[TMP2458]], 0
+// CHECK-NEXT:    [[TMP2460:%.*]] = extractvalue { i32, i1 } [[TMP2458]], 1
+// CHECK-NEXT:    [[TMP2461:%.*]] = select i1 [[TMP2460]], i32 [[TMP2456]], i32 [[TMP2459]]
+// CHECK-NEXT:    store i32 [[TMP2461]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2462:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2463:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2464:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2462]], i32 [[TMP2463]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2465:%.*]] = extractvalue { i32, i1 } [[TMP2464]], 0
+// CHECK-NEXT:    [[TMP2466:%.*]] = extractvalue { i32, i1 } [[TMP2464]], 1
+// CHECK-NEXT:    br i1 [[TMP2466]], label [[UIX_ATOMIC_EXIT237:%.*]], label [[UIX_ATOMIC_CONT238:%.*]]
+// CHECK:       uix.atomic.cont238:
+// CHECK-NEXT:    store i32 [[TMP2465]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT237]]
+// CHECK:       uix.atomic.exit237:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2467:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2468:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2469:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2467]], i32 [[TMP2468]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2470:%.*]] = extractvalue { i32, i1 } [[TMP2469]], 0
+// CHECK-NEXT:    [[TMP2471:%.*]] = extractvalue { i32, i1 } [[TMP2469]], 1
+// CHECK-NEXT:    br i1 [[TMP2471]], label [[UIX_ATOMIC_EXIT239:%.*]], label [[UIX_ATOMIC_CONT240:%.*]]
+// CHECK:       uix.atomic.cont240:
+// CHECK-NEXT:    store i32 [[TMP2470]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT239]]
+// CHECK:       uix.atomic.exit239:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2472:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2473:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2474:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2472]], i32 [[TMP2473]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2475:%.*]] = extractvalue { i32, i1 } [[TMP2474]], 1
+// CHECK-NEXT:    [[TMP2476:%.*]] = zext i1 [[TMP2475]] to i32
+// CHECK-NEXT:    store i32 [[TMP2476]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2477:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2478:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2479:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2477]], i32 [[TMP2478]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2480:%.*]] = extractvalue { i32, i1 } [[TMP2479]], 1
+// CHECK-NEXT:    [[TMP2481:%.*]] = zext i1 [[TMP2480]] to i32
+// CHECK-NEXT:    store i32 [[TMP2481]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2482:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2483:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2484:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2482]], i32 [[TMP2483]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2485:%.*]] = extractvalue { i32, i1 } [[TMP2484]], 0
+// CHECK-NEXT:    [[TMP2486:%.*]] = extractvalue { i32, i1 } [[TMP2484]], 1
+// CHECK-NEXT:    br i1 [[TMP2486]], label [[UIX_ATOMIC_EXIT241:%.*]], label [[UIX_ATOMIC_CONT242:%.*]]
+// CHECK:       uix.atomic.cont242:
+// CHECK-NEXT:    store i32 [[TMP2485]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT241]]
+// CHECK:       uix.atomic.exit241:
+// CHECK-NEXT:    [[TMP2487:%.*]] = extractvalue { i32, i1 } [[TMP2484]], 1
+// CHECK-NEXT:    [[TMP2488:%.*]] = zext i1 [[TMP2487]] to i32
+// CHECK-NEXT:    store i32 [[TMP2488]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2489:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2490:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2491:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2489]], i32 [[TMP2490]] acq_rel acquire, align 4
+// CHECK-NEXT:    [[TMP2492:%.*]] = extractvalue { i32, i1 } [[TMP2491]], 0
+// CHECK-NEXT:    [[TMP2493:%.*]] = extractvalue { i32, i1 } [[TMP2491]], 1
+// CHECK-NEXT:    br i1 [[TMP2493]], label [[UIX_ATOMIC_EXIT243:%.*]], label [[UIX_ATOMIC_CONT244:%.*]]
+// CHECK:       uix.atomic.cont244:
+// CHECK-NEXT:    store i32 [[TMP2492]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT243]]
+// CHECK:       uix.atomic.exit243:
+// CHECK-NEXT:    [[TMP2494:%.*]] = extractvalue { i32, i1 } [[TMP2491]], 1
+// CHECK-NEXT:    [[TMP2495:%.*]] = zext i1 [[TMP2494]] to i32
+// CHECK-NEXT:    store i32 [[TMP2495]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2496:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2497:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2496]] acquire, align 4
+// CHECK-NEXT:    store i32 [[TMP2497]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2498:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2499:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2498]] acquire, align 4
+// CHECK-NEXT:    store i32 [[TMP2499]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2500:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2501:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2500]] acquire, align 4
+// CHECK-NEXT:    store i32 [[TMP2501]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2502:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2503:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2502]] acquire, align 4
+// CHECK-NEXT:    store i32 [[TMP2503]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2504:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2505:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2506:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2504]], i32 [[TMP2505]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2507:%.*]] = extractvalue { i32, i1 } [[TMP2506]], 0
+// CHECK-NEXT:    store i32 [[TMP2507]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2508:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2509:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2510:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2508]], i32 [[TMP2509]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2511:%.*]] = extractvalue { i32, i1 } [[TMP2510]], 0
+// CHECK-NEXT:    store i32 [[TMP2511]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2512:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2513:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2512]] acquire, align 4
+// CHECK-NEXT:    [[TMP2514:%.*]] = icmp ugt i32 [[TMP2513]], [[TMP2512]]
+// CHECK-NEXT:    [[TMP2515:%.*]] = select i1 [[TMP2514]], i32 [[TMP2512]], i32 [[TMP2513]]
+// CHECK-NEXT:    store i32 [[TMP2515]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2516:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2517:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2516]] acquire, align 4
+// CHECK-NEXT:    [[TMP2518:%.*]] = icmp ult i32 [[TMP2517]], [[TMP2516]]
+// CHECK-NEXT:    [[TMP2519:%.*]] = select i1 [[TMP2518]], i32 [[TMP2516]], i32 [[TMP2517]]
+// CHECK-NEXT:    store i32 [[TMP2519]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2520:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2521:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2520]] acquire, align 4
+// CHECK-NEXT:    [[TMP2522:%.*]] = icmp ult i32 [[TMP2521]], [[TMP2520]]
+// CHECK-NEXT:    [[TMP2523:%.*]] = select i1 [[TMP2522]], i32 [[TMP2520]], i32 [[TMP2521]]
+// CHECK-NEXT:    store i32 [[TMP2523]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2524:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2525:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2524]] acquire, align 4
+// CHECK-NEXT:    [[TMP2526:%.*]] = icmp ugt i32 [[TMP2525]], [[TMP2524]]
+// CHECK-NEXT:    [[TMP2527:%.*]] = select i1 [[TMP2526]], i32 [[TMP2524]], i32 [[TMP2525]]
+// CHECK-NEXT:    store i32 [[TMP2527]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2528:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2529:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2530:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2528]], i32 [[TMP2529]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2531:%.*]] = extractvalue { i32, i1 } [[TMP2530]], 0
+// CHECK-NEXT:    [[TMP2532:%.*]] = extractvalue { i32, i1 } [[TMP2530]], 1
+// CHECK-NEXT:    [[TMP2533:%.*]] = select i1 [[TMP2532]], i32 [[TMP2528]], i32 [[TMP2531]]
+// CHECK-NEXT:    store i32 [[TMP2533]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2534:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2535:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2536:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2534]], i32 [[TMP2535]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2537:%.*]] = extractvalue { i32, i1 } [[TMP2536]], 0
+// CHECK-NEXT:    [[TMP2538:%.*]] = extractvalue { i32, i1 } [[TMP2536]], 1
+// CHECK-NEXT:    [[TMP2539:%.*]] = select i1 [[TMP2538]], i32 [[TMP2534]], i32 [[TMP2537]]
+// CHECK-NEXT:    store i32 [[TMP2539]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2540:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2541:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2542:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2540]], i32 [[TMP2541]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2543:%.*]] = extractvalue { i32, i1 } [[TMP2542]], 0
+// CHECK-NEXT:    [[TMP2544:%.*]] = extractvalue { i32, i1 } [[TMP2542]], 1
+// CHECK-NEXT:    br i1 [[TMP2544]], label [[UIX_ATOMIC_EXIT245:%.*]], label [[UIX_ATOMIC_CONT246:%.*]]
+// CHECK:       uix.atomic.cont246:
+// CHECK-NEXT:    store i32 [[TMP2543]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT245]]
+// CHECK:       uix.atomic.exit245:
+// CHECK-NEXT:    [[TMP2545:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2546:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2547:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2545]], i32 [[TMP2546]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2548:%.*]] = extractvalue { i32, i1 } [[TMP2547]], 0
+// CHECK-NEXT:    [[TMP2549:%.*]] = extractvalue { i32, i1 } [[TMP2547]], 1
+// CHECK-NEXT:    br i1 [[TMP2549]], label [[UIX_ATOMIC_EXIT247:%.*]], label [[UIX_ATOMIC_CONT248:%.*]]
+// CHECK:       uix.atomic.cont248:
+// CHECK-NEXT:    store i32 [[TMP2548]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT247]]
+// CHECK:       uix.atomic.exit247:
+// CHECK-NEXT:    [[TMP2550:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2551:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2552:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2550]], i32 [[TMP2551]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2553:%.*]] = extractvalue { i32, i1 } [[TMP2552]], 1
+// CHECK-NEXT:    [[TMP2554:%.*]] = zext i1 [[TMP2553]] to i32
+// CHECK-NEXT:    store i32 [[TMP2554]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2555:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2556:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2557:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2555]], i32 [[TMP2556]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2558:%.*]] = extractvalue { i32, i1 } [[TMP2557]], 1
+// CHECK-NEXT:    [[TMP2559:%.*]] = zext i1 [[TMP2558]] to i32
+// CHECK-NEXT:    store i32 [[TMP2559]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2560:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2561:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2562:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2560]], i32 [[TMP2561]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2563:%.*]] = extractvalue { i32, i1 } [[TMP2562]], 0
+// CHECK-NEXT:    [[TMP2564:%.*]] = extractvalue { i32, i1 } [[TMP2562]], 1
+// CHECK-NEXT:    br i1 [[TMP2564]], label [[UIX_ATOMIC_EXIT249:%.*]], label [[UIX_ATOMIC_CONT250:%.*]]
+// CHECK:       uix.atomic.cont250:
+// CHECK-NEXT:    store i32 [[TMP2563]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT249]]
+// CHECK:       uix.atomic.exit249:
+// CHECK-NEXT:    [[TMP2565:%.*]] = extractvalue { i32, i1 } [[TMP2562]], 1
+// CHECK-NEXT:    [[TMP2566:%.*]] = zext i1 [[TMP2565]] to i32
+// CHECK-NEXT:    store i32 [[TMP2566]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2567:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2568:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2569:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2567]], i32 [[TMP2568]] acquire acquire, align 4
+// CHECK-NEXT:    [[TMP2570:%.*]] = extractvalue { i32, i1 } [[TMP2569]], 0
+// CHECK-NEXT:    [[TMP2571:%.*]] = extractvalue { i32, i1 } [[TMP2569]], 1
+// CHECK-NEXT:    br i1 [[TMP2571]], label [[UIX_ATOMIC_EXIT251:%.*]], label [[UIX_ATOMIC_CONT252:%.*]]
+// CHECK:       uix.atomic.cont252:
+// CHECK-NEXT:    store i32 [[TMP2570]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT251]]
+// CHECK:       uix.atomic.exit251:
+// CHECK-NEXT:    [[TMP2572:%.*]] = extractvalue { i32, i1 } [[TMP2569]], 1
+// CHECK-NEXT:    [[TMP2573:%.*]] = zext i1 [[TMP2572]] to i32
+// CHECK-NEXT:    store i32 [[TMP2573]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2574:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2575:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2574]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2575]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2576:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2577:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2576]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2577]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2578:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2579:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2578]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2579]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2580:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2581:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2580]] monotonic, align 4
+// CHECK-NEXT:    store i32 [[TMP2581]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2582:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2583:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2584:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2582]], i32 [[TMP2583]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2585:%.*]] = extractvalue { i32, i1 } [[TMP2584]], 0
+// CHECK-NEXT:    store i32 [[TMP2585]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2586:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2587:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2588:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2586]], i32 [[TMP2587]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2589:%.*]] = extractvalue { i32, i1 } [[TMP2588]], 0
+// CHECK-NEXT:    store i32 [[TMP2589]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2590:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2591:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2590]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2592:%.*]] = icmp ugt i32 [[TMP2591]], [[TMP2590]]
+// CHECK-NEXT:    [[TMP2593:%.*]] = select i1 [[TMP2592]], i32 [[TMP2590]], i32 [[TMP2591]]
+// CHECK-NEXT:    store i32 [[TMP2593]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2594:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2595:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2594]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2596:%.*]] = icmp ult i32 [[TMP2595]], [[TMP2594]]
+// CHECK-NEXT:    [[TMP2597:%.*]] = select i1 [[TMP2596]], i32 [[TMP2594]], i32 [[TMP2595]]
+// CHECK-NEXT:    store i32 [[TMP2597]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2598:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2599:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2598]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2600:%.*]] = icmp ult i32 [[TMP2599]], [[TMP2598]]
+// CHECK-NEXT:    [[TMP2601:%.*]] = select i1 [[TMP2600]], i32 [[TMP2598]], i32 [[TMP2599]]
+// CHECK-NEXT:    store i32 [[TMP2601]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2602:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2603:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2602]] monotonic, align 4
+// CHECK-NEXT:    [[TMP2604:%.*]] = icmp ugt i32 [[TMP2603]], [[TMP2602]]
+// CHECK-NEXT:    [[TMP2605:%.*]] = select i1 [[TMP2604]], i32 [[TMP2602]], i32 [[TMP2603]]
+// CHECK-NEXT:    store i32 [[TMP2605]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2606:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2607:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2608:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2606]], i32 [[TMP2607]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2609:%.*]] = extractvalue { i32, i1 } [[TMP2608]], 0
+// CHECK-NEXT:    [[TMP2610:%.*]] = extractvalue { i32, i1 } [[TMP2608]], 1
+// CHECK-NEXT:    [[TMP2611:%.*]] = select i1 [[TMP2610]], i32 [[TMP2606]], i32 [[TMP2609]]
+// CHECK-NEXT:    store i32 [[TMP2611]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2612:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2613:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2614:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2612]], i32 [[TMP2613]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2615:%.*]] = extractvalue { i32, i1 } [[TMP2614]], 0
+// CHECK-NEXT:    [[TMP2616:%.*]] = extractvalue { i32, i1 } [[TMP2614]], 1
+// CHECK-NEXT:    [[TMP2617:%.*]] = select i1 [[TMP2616]], i32 [[TMP2612]], i32 [[TMP2615]]
+// CHECK-NEXT:    store i32 [[TMP2617]], i32* [[UIV]], align 4
+// CHECK-NEXT:    [[TMP2618:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2619:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2620:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2618]], i32 [[TMP2619]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2621:%.*]] = extractvalue { i32, i1 } [[TMP2620]], 0
+// CHECK-NEXT:    [[TMP2622:%.*]] = extractvalue { i32, i1 } [[TMP2620]], 1
+// CHECK-NEXT:    br i1 [[TMP2622]], label [[UIX_ATOMIC_EXIT253:%.*]], label [[UIX_ATOMIC_CONT254:%.*]]
+// CHECK:       uix.atomic.cont254:
+// CHECK-NEXT:    store i32 [[TMP2621]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT253]]
+// CHECK:       uix.atomic.exit253:
+// CHECK-NEXT:    [[TMP2623:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2624:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2625:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2623]], i32 [[TMP2624]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2626:%.*]] = extractvalue { i32, i1 } [[TMP2625]], 0
+// CHECK-NEXT:    [[TMP2627:%.*]] = extractvalue { i32, i1 } [[TMP2625]], 1
+// CHECK-NEXT:    br i1 [[TMP2627]], label [[UIX_ATOMIC_EXIT255:%.*]], label [[UIX_ATOMIC_CONT256:%.*]]
+// CHECK:       uix.atomic.cont256:
+// CHECK-NEXT:    store i32 [[TMP2626]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT255]]
+// CHECK:       uix.atomic.exit255:
+// CHECK-NEXT:    [[TMP2628:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2629:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2630:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2628]], i32 [[TMP2629]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2631:%.*]] = extractvalue { i32, i1 } [[TMP2630]], 1
+// CHECK-NEXT:    [[TMP2632:%.*]] = zext i1 [[TMP2631]] to i32
+// CHECK-NEXT:    store i32 [[TMP2632]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2633:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2634:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2635:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2633]], i32 [[TMP2634]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2636:%.*]] = extractvalue { i32, i1 } [[TMP2635]], 1
+// CHECK-NEXT:    [[TMP2637:%.*]] = zext i1 [[TMP2636]] to i32
+// CHECK-NEXT:    store i32 [[TMP2637]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2638:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2639:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2640:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2638]], i32 [[TMP2639]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2641:%.*]] = extractvalue { i32, i1 } [[TMP2640]], 0
+// CHECK-NEXT:    [[TMP2642:%.*]] = extractvalue { i32, i1 } [[TMP2640]], 1
+// CHECK-NEXT:    br i1 [[TMP2642]], label [[UIX_ATOMIC_EXIT257:%.*]], label [[UIX_ATOMIC_CONT258:%.*]]
+// CHECK:       uix.atomic.cont258:
+// CHECK-NEXT:    store i32 [[TMP2641]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT257]]
+// CHECK:       uix.atomic.exit257:
+// CHECK-NEXT:    [[TMP2643:%.*]] = extractvalue { i32, i1 } [[TMP2640]], 1
+// CHECK-NEXT:    [[TMP2644:%.*]] = zext i1 [[TMP2643]] to i32
+// CHECK-NEXT:    store i32 [[TMP2644]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2645:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2646:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2647:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2645]], i32 [[TMP2646]] monotonic monotonic, align 4
+// CHECK-NEXT:    [[TMP2648:%.*]] = extractvalue { i32, i1 } [[TMP2647]], 0
+// CHECK-NEXT:    [[TMP2649:%.*]] = extractvalue { i32, i1 } [[TMP2647]], 1
+// CHECK-NEXT:    br i1 [[TMP2649]], label [[UIX_ATOMIC_EXIT259:%.*]], label [[UIX_ATOMIC_CONT260:%.*]]
+// CHECK:       uix.atomic.cont260:
+// CHECK-NEXT:    store i32 [[TMP2648]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT259]]
+// CHECK:       uix.atomic.exit259:
+// CHECK-NEXT:    [[TMP2650:%.*]] = extractvalue { i32, i1 } [[TMP2647]], 1
+// CHECK-NEXT:    [[TMP2651:%.*]] = zext i1 [[TMP2650]] to i32
+// CHECK-NEXT:    store i32 [[TMP2651]], i32* [[UIR]], align 4
+// CHECK-NEXT:    [[TMP2652:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2653:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2652]] release, align 4
+// CHECK-NEXT:    store i32 [[TMP2653]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2654:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2655:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2654]] release, align 4
+// CHECK-NEXT:    store i32 [[TMP2655]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2656:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2657:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2656]] release, align 4
+// CHECK-NEXT:    store i32 [[TMP2657]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2658:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2659:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2658]] release, align 4
+// CHECK-NEXT:    store i32 [[TMP2659]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2660:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2661:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2662:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2660]], i32 [[TMP2661]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2663:%.*]] = extractvalue { i32, i1 } [[TMP2662]], 0
+// CHECK-NEXT:    store i32 [[TMP2663]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2664:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2665:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2666:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2664]], i32 [[TMP2665]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2667:%.*]] = extractvalue { i32, i1 } [[TMP2666]], 0
+// CHECK-NEXT:    store i32 [[TMP2667]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2668:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2669:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2668]] release, align 4
+// CHECK-NEXT:    [[TMP2670:%.*]] = icmp ugt i32 [[TMP2669]], [[TMP2668]]
+// CHECK-NEXT:    [[TMP2671:%.*]] = select i1 [[TMP2670]], i32 [[TMP2668]], i32 [[TMP2669]]
+// CHECK-NEXT:    store i32 [[TMP2671]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2672:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2673:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2672]] release, align 4
+// CHECK-NEXT:    [[TMP2674:%.*]] = icmp ult i32 [[TMP2673]], [[TMP2672]]
+// CHECK-NEXT:    [[TMP2675:%.*]] = select i1 [[TMP2674]], i32 [[TMP2672]], i32 [[TMP2673]]
+// CHECK-NEXT:    store i32 [[TMP2675]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2676:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2677:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2676]] release, align 4
+// CHECK-NEXT:    [[TMP2678:%.*]] = icmp ult i32 [[TMP2677]], [[TMP2676]]
+// CHECK-NEXT:    [[TMP2679:%.*]] = select i1 [[TMP2678]], i32 [[TMP2676]], i32 [[TMP2677]]
+// CHECK-NEXT:    store i32 [[TMP2679]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2680:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2681:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2680]] release, align 4
+// CHECK-NEXT:    [[TMP2682:%.*]] = icmp ugt i32 [[TMP2681]], [[TMP2680]]
+// CHECK-NEXT:    [[TMP2683:%.*]] = select i1 [[TMP2682]], i32 [[TMP2680]], i32 [[TMP2681]]
+// CHECK-NEXT:    store i32 [[TMP2683]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2684:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2685:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2686:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2684]], i32 [[TMP2685]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2687:%.*]] = extractvalue { i32, i1 } [[TMP2686]], 0
+// CHECK-NEXT:    [[TMP2688:%.*]] = extractvalue { i32, i1 } [[TMP2686]], 1
+// CHECK-NEXT:    [[TMP2689:%.*]] = select i1 [[TMP2688]], i32 [[TMP2684]], i32 [[TMP2687]]
+// CHECK-NEXT:    store i32 [[TMP2689]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2690:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2691:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2692:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2690]], i32 [[TMP2691]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2693:%.*]] = extractvalue { i32, i1 } [[TMP2692]], 0
+// CHECK-NEXT:    [[TMP2694:%.*]] = extractvalue { i32, i1 } [[TMP2692]], 1
+// CHECK-NEXT:    [[TMP2695:%.*]] = select i1 [[TMP2694]], i32 [[TMP2690]], i32 [[TMP2693]]
+// CHECK-NEXT:    store i32 [[TMP2695]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2696:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2697:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2698:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2696]], i32 [[TMP2697]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2699:%.*]] = extractvalue { i32, i1 } [[TMP2698]], 0
+// CHECK-NEXT:    [[TMP2700:%.*]] = extractvalue { i32, i1 } [[TMP2698]], 1
+// CHECK-NEXT:    br i1 [[TMP2700]], label [[UIX_ATOMIC_EXIT261:%.*]], label [[UIX_ATOMIC_CONT262:%.*]]
+// CHECK:       uix.atomic.cont262:
+// CHECK-NEXT:    store i32 [[TMP2699]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT261]]
+// CHECK:       uix.atomic.exit261:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2701:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2702:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2703:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2701]], i32 [[TMP2702]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2704:%.*]] = extractvalue { i32, i1 } [[TMP2703]], 0
+// CHECK-NEXT:    [[TMP2705:%.*]] = extractvalue { i32, i1 } [[TMP2703]], 1
+// CHECK-NEXT:    br i1 [[TMP2705]], label [[UIX_ATOMIC_EXIT263:%.*]], label [[UIX_ATOMIC_CONT264:%.*]]
+// CHECK:       uix.atomic.cont264:
+// CHECK-NEXT:    store i32 [[TMP2704]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT263]]
+// CHECK:       uix.atomic.exit263:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2706:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2707:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2708:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2706]], i32 [[TMP2707]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2709:%.*]] = extractvalue { i32, i1 } [[TMP2708]], 1
+// CHECK-NEXT:    [[TMP2710:%.*]] = zext i1 [[TMP2709]] to i32
+// CHECK-NEXT:    store i32 [[TMP2710]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2711:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2712:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2713:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2711]], i32 [[TMP2712]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2714:%.*]] = extractvalue { i32, i1 } [[TMP2713]], 1
+// CHECK-NEXT:    [[TMP2715:%.*]] = zext i1 [[TMP2714]] to i32
+// CHECK-NEXT:    store i32 [[TMP2715]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2716:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2717:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2718:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2716]], i32 [[TMP2717]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2719:%.*]] = extractvalue { i32, i1 } [[TMP2718]], 0
+// CHECK-NEXT:    [[TMP2720:%.*]] = extractvalue { i32, i1 } [[TMP2718]], 1
+// CHECK-NEXT:    br i1 [[TMP2720]], label [[UIX_ATOMIC_EXIT265:%.*]], label [[UIX_ATOMIC_CONT266:%.*]]
+// CHECK:       uix.atomic.cont266:
+// CHECK-NEXT:    store i32 [[TMP2719]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT265]]
+// CHECK:       uix.atomic.exit265:
+// CHECK-NEXT:    [[TMP2721:%.*]] = extractvalue { i32, i1 } [[TMP2718]], 1
+// CHECK-NEXT:    [[TMP2722:%.*]] = zext i1 [[TMP2721]] to i32
+// CHECK-NEXT:    store i32 [[TMP2722]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2723:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2724:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2725:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2723]], i32 [[TMP2724]] release monotonic, align 4
+// CHECK-NEXT:    [[TMP2726:%.*]] = extractvalue { i32, i1 } [[TMP2725]], 0
+// CHECK-NEXT:    [[TMP2727:%.*]] = extractvalue { i32, i1 } [[TMP2725]], 1
+// CHECK-NEXT:    br i1 [[TMP2727]], label [[UIX_ATOMIC_EXIT267:%.*]], label [[UIX_ATOMIC_CONT268:%.*]]
+// CHECK:       uix.atomic.cont268:
+// CHECK-NEXT:    store i32 [[TMP2726]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT267]]
+// CHECK:       uix.atomic.exit267:
+// CHECK-NEXT:    [[TMP2728:%.*]] = extractvalue { i32, i1 } [[TMP2725]], 1
+// CHECK-NEXT:    [[TMP2729:%.*]] = zext i1 [[TMP2728]] to i32
+// CHECK-NEXT:    store i32 [[TMP2729]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2730:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2731:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2730]] seq_cst, align 4
+// CHECK-NEXT:    store i32 [[TMP2731]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2732:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2733:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2732]] seq_cst, align 4
+// CHECK-NEXT:    store i32 [[TMP2733]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2734:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2735:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2734]] seq_cst, align 4
+// CHECK-NEXT:    store i32 [[TMP2735]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2736:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2737:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2736]] seq_cst, align 4
+// CHECK-NEXT:    store i32 [[TMP2737]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2738:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2739:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2740:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2738]], i32 [[TMP2739]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2741:%.*]] = extractvalue { i32, i1 } [[TMP2740]], 0
+// CHECK-NEXT:    store i32 [[TMP2741]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2742:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2743:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2744:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2742]], i32 [[TMP2743]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2745:%.*]] = extractvalue { i32, i1 } [[TMP2744]], 0
+// CHECK-NEXT:    store i32 [[TMP2745]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2746:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2747:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2746]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2748:%.*]] = icmp ugt i32 [[TMP2747]], [[TMP2746]]
+// CHECK-NEXT:    [[TMP2749:%.*]] = select i1 [[TMP2748]], i32 [[TMP2746]], i32 [[TMP2747]]
+// CHECK-NEXT:    store i32 [[TMP2749]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2750:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2751:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2750]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2752:%.*]] = icmp ult i32 [[TMP2751]], [[TMP2750]]
+// CHECK-NEXT:    [[TMP2753:%.*]] = select i1 [[TMP2752]], i32 [[TMP2750]], i32 [[TMP2751]]
+// CHECK-NEXT:    store i32 [[TMP2753]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2754:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2755:%.*]] = atomicrmw umin i32* [[UIX]], i32 [[TMP2754]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2756:%.*]] = icmp ult i32 [[TMP2755]], [[TMP2754]]
+// CHECK-NEXT:    [[TMP2757:%.*]] = select i1 [[TMP2756]], i32 [[TMP2754]], i32 [[TMP2755]]
+// CHECK-NEXT:    store i32 [[TMP2757]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2758:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2759:%.*]] = atomicrmw umax i32* [[UIX]], i32 [[TMP2758]] seq_cst, align 4
+// CHECK-NEXT:    [[TMP2760:%.*]] = icmp ugt i32 [[TMP2759]], [[TMP2758]]
+// CHECK-NEXT:    [[TMP2761:%.*]] = select i1 [[TMP2760]], i32 [[TMP2758]], i32 [[TMP2759]]
+// CHECK-NEXT:    store i32 [[TMP2761]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2762:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2763:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2764:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2762]], i32 [[TMP2763]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2765:%.*]] = extractvalue { i32, i1 } [[TMP2764]], 0
+// CHECK-NEXT:    [[TMP2766:%.*]] = extractvalue { i32, i1 } [[TMP2764]], 1
+// CHECK-NEXT:    [[TMP2767:%.*]] = select i1 [[TMP2766]], i32 [[TMP2762]], i32 [[TMP2765]]
+// CHECK-NEXT:    store i32 [[TMP2767]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2768:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2769:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2770:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2768]], i32 [[TMP2769]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2771:%.*]] = extractvalue { i32, i1 } [[TMP2770]], 0
+// CHECK-NEXT:    [[TMP2772:%.*]] = extractvalue { i32, i1 } [[TMP2770]], 1
+// CHECK-NEXT:    [[TMP2773:%.*]] = select i1 [[TMP2772]], i32 [[TMP2768]], i32 [[TMP2771]]
+// CHECK-NEXT:    store i32 [[TMP2773]], i32* [[UIV]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2774:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2775:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2776:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2774]], i32 [[TMP2775]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2777:%.*]] = extractvalue { i32, i1 } [[TMP2776]], 0
+// CHECK-NEXT:    [[TMP2778:%.*]] = extractvalue { i32, i1 } [[TMP2776]], 1
+// CHECK-NEXT:    br i1 [[TMP2778]], label [[UIX_ATOMIC_EXIT269:%.*]], label [[UIX_ATOMIC_CONT270:%.*]]
+// CHECK:       uix.atomic.cont270:
+// CHECK-NEXT:    store i32 [[TMP2777]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT269]]
+// CHECK:       uix.atomic.exit269:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2779:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2780:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2781:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2779]], i32 [[TMP2780]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2782:%.*]] = extractvalue { i32, i1 } [[TMP2781]], 0
+// CHECK-NEXT:    [[TMP2783:%.*]] = extractvalue { i32, i1 } [[TMP2781]], 1
+// CHECK-NEXT:    br i1 [[TMP2783]], label [[UIX_ATOMIC_EXIT271:%.*]], label [[UIX_ATOMIC_CONT272:%.*]]
+// CHECK:       uix.atomic.cont272:
+// CHECK-NEXT:    store i32 [[TMP2782]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT271]]
+// CHECK:       uix.atomic.exit271:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2784:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2785:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2786:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2784]], i32 [[TMP2785]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2787:%.*]] = extractvalue { i32, i1 } [[TMP2786]], 1
+// CHECK-NEXT:    [[TMP2788:%.*]] = zext i1 [[TMP2787]] to i32
+// CHECK-NEXT:    store i32 [[TMP2788]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2789:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2790:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2791:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2789]], i32 [[TMP2790]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2792:%.*]] = extractvalue { i32, i1 } [[TMP2791]], 1
+// CHECK-NEXT:    [[TMP2793:%.*]] = zext i1 [[TMP2792]] to i32
+// CHECK-NEXT:    store i32 [[TMP2793]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2794:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2795:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2796:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2794]], i32 [[TMP2795]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2797:%.*]] = extractvalue { i32, i1 } [[TMP2796]], 0
+// CHECK-NEXT:    [[TMP2798:%.*]] = extractvalue { i32, i1 } [[TMP2796]], 1
+// CHECK-NEXT:    br i1 [[TMP2798]], label [[UIX_ATOMIC_EXIT273:%.*]], label [[UIX_ATOMIC_CONT274:%.*]]
+// CHECK:       uix.atomic.cont274:
+// CHECK-NEXT:    store i32 [[TMP2797]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT273]]
+// CHECK:       uix.atomic.exit273:
+// CHECK-NEXT:    [[TMP2799:%.*]] = extractvalue { i32, i1 } [[TMP2796]], 1
+// CHECK-NEXT:    [[TMP2800:%.*]] = zext i1 [[TMP2799]] to i32
+// CHECK-NEXT:    store i32 [[TMP2800]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2801:%.*]] = load i32, i32* [[UIE]], align 4
+// CHECK-NEXT:    [[TMP2802:%.*]] = load i32, i32* [[UID]], align 4
+// CHECK-NEXT:    [[TMP2803:%.*]] = cmpxchg i32* [[UIX]], i32 [[TMP2801]], i32 [[TMP2802]] seq_cst seq_cst, align 4
+// CHECK-NEXT:    [[TMP2804:%.*]] = extractvalue { i32, i1 } [[TMP2803]], 0
+// CHECK-NEXT:    [[TMP2805:%.*]] = extractvalue { i32, i1 } [[TMP2803]], 1
+// CHECK-NEXT:    br i1 [[TMP2805]], label [[UIX_ATOMIC_EXIT275:%.*]], label [[UIX_ATOMIC_CONT276:%.*]]
+// CHECK:       uix.atomic.cont276:
+// CHECK-NEXT:    store i32 [[TMP2804]], i32* [[UIV]], align 4
+// CHECK-NEXT:    br label [[UIX_ATOMIC_EXIT275]]
+// CHECK:       uix.atomic.exit275:
+// CHECK-NEXT:    [[TMP2806:%.*]] = extractvalue { i32, i1 } [[TMP2803]], 1
+// CHECK-NEXT:    [[TMP2807:%.*]] = zext i1 [[TMP2806]] to i32
+// CHECK-NEXT:    store i32 [[TMP2807]], i32* [[UIR]], align 4
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2808:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2809:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2808]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2809]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2810:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2811:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2810]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2811]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2812:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2813:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2812]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2813]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2814:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2815:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2814]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2815]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2816:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2817:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2818:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2816]], i64 [[TMP2817]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2819:%.*]] = extractvalue { i64, i1 } [[TMP2818]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP2819]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2820:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2821:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2822:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2820]], i64 [[TMP2821]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2823:%.*]] = extractvalue { i64, i1 } [[TMP2822]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP2823]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2824:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2825:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2824]] monotonic, align 8
+// CHECK-NEXT:    [[TMP2826:%.*]] = icmp ugt i64 [[TMP2825]], [[TMP2824]]
+// CHECK-NEXT:    [[TMP2827:%.*]] = select i1 [[TMP2826]], i64 [[TMP2824]], i64 [[TMP2825]]
+// CHECK-NEXT:    store volatile i64 [[TMP2827]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2828:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2829:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2828]] monotonic, align 8
+// CHECK-NEXT:    [[TMP2830:%.*]] = icmp ult i64 [[TMP2829]], [[TMP2828]]
+// CHECK-NEXT:    [[TMP2831:%.*]] = select i1 [[TMP2830]], i64 [[TMP2828]], i64 [[TMP2829]]
+// CHECK-NEXT:    store volatile i64 [[TMP2831]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2832:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2833:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2832]] monotonic, align 8
+// CHECK-NEXT:    [[TMP2834:%.*]] = icmp ult i64 [[TMP2833]], [[TMP2832]]
+// CHECK-NEXT:    [[TMP2835:%.*]] = select i1 [[TMP2834]], i64 [[TMP2832]], i64 [[TMP2833]]
+// CHECK-NEXT:    store volatile i64 [[TMP2835]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2836:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2837:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2836]] monotonic, align 8
+// CHECK-NEXT:    [[TMP2838:%.*]] = icmp ugt i64 [[TMP2837]], [[TMP2836]]
+// CHECK-NEXT:    [[TMP2839:%.*]] = select i1 [[TMP2838]], i64 [[TMP2836]], i64 [[TMP2837]]
+// CHECK-NEXT:    store volatile i64 [[TMP2839]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2840:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2841:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2842:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2840]], i64 [[TMP2841]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2843:%.*]] = extractvalue { i64, i1 } [[TMP2842]], 0
+// CHECK-NEXT:    [[TMP2844:%.*]] = extractvalue { i64, i1 } [[TMP2842]], 1
+// CHECK-NEXT:    [[TMP2845:%.*]] = select i1 [[TMP2844]], i64 [[TMP2840]], i64 [[TMP2843]]
+// CHECK-NEXT:    store volatile i64 [[TMP2845]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2846:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2847:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2848:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2846]], i64 [[TMP2847]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2849:%.*]] = extractvalue { i64, i1 } [[TMP2848]], 0
+// CHECK-NEXT:    [[TMP2850:%.*]] = extractvalue { i64, i1 } [[TMP2848]], 1
+// CHECK-NEXT:    [[TMP2851:%.*]] = select i1 [[TMP2850]], i64 [[TMP2846]], i64 [[TMP2849]]
+// CHECK-NEXT:    store volatile i64 [[TMP2851]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2852:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2853:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2854:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2852]], i64 [[TMP2853]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2855:%.*]] = extractvalue { i64, i1 } [[TMP2854]], 0
+// CHECK-NEXT:    [[TMP2856:%.*]] = extractvalue { i64, i1 } [[TMP2854]], 1
+// CHECK-NEXT:    br i1 [[TMP2856]], label [[LX_ATOMIC_EXIT:%.*]], label [[LX_ATOMIC_CONT:%.*]]
+// CHECK:       lx.atomic.cont:
+// CHECK-NEXT:    store i64 [[TMP2855]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT]]
+// CHECK:       lx.atomic.exit:
+// CHECK-NEXT:    [[TMP2857:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2858:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2859:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2857]], i64 [[TMP2858]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2860:%.*]] = extractvalue { i64, i1 } [[TMP2859]], 0
+// CHECK-NEXT:    [[TMP2861:%.*]] = extractvalue { i64, i1 } [[TMP2859]], 1
+// CHECK-NEXT:    br i1 [[TMP2861]], label [[LX_ATOMIC_EXIT277:%.*]], label [[LX_ATOMIC_CONT278:%.*]]
+// CHECK:       lx.atomic.cont278:
+// CHECK-NEXT:    store i64 [[TMP2860]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT277]]
+// CHECK:       lx.atomic.exit277:
+// CHECK-NEXT:    [[TMP2862:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2863:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2864:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2862]], i64 [[TMP2863]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2865:%.*]] = extractvalue { i64, i1 } [[TMP2864]], 1
+// CHECK-NEXT:    [[TMP2866:%.*]] = zext i1 [[TMP2865]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2866]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP2867:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2868:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2869:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2867]], i64 [[TMP2868]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2870:%.*]] = extractvalue { i64, i1 } [[TMP2869]], 1
+// CHECK-NEXT:    [[TMP2871:%.*]] = zext i1 [[TMP2870]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2871]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP2872:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2873:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2874:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2872]], i64 [[TMP2873]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2875:%.*]] = extractvalue { i64, i1 } [[TMP2874]], 0
+// CHECK-NEXT:    [[TMP2876:%.*]] = extractvalue { i64, i1 } [[TMP2874]], 1
+// CHECK-NEXT:    br i1 [[TMP2876]], label [[LX_ATOMIC_EXIT279:%.*]], label [[LX_ATOMIC_CONT280:%.*]]
+// CHECK:       lx.atomic.cont280:
+// CHECK-NEXT:    store i64 [[TMP2875]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT279]]
+// CHECK:       lx.atomic.exit279:
+// CHECK-NEXT:    [[TMP2877:%.*]] = extractvalue { i64, i1 } [[TMP2874]], 1
+// CHECK-NEXT:    [[TMP2878:%.*]] = zext i1 [[TMP2877]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2878]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP2879:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2880:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2881:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2879]], i64 [[TMP2880]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP2882:%.*]] = extractvalue { i64, i1 } [[TMP2881]], 0
+// CHECK-NEXT:    [[TMP2883:%.*]] = extractvalue { i64, i1 } [[TMP2881]], 1
+// CHECK-NEXT:    br i1 [[TMP2883]], label [[LX_ATOMIC_EXIT281:%.*]], label [[LX_ATOMIC_CONT282:%.*]]
+// CHECK:       lx.atomic.cont282:
+// CHECK-NEXT:    store i64 [[TMP2882]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT281]]
+// CHECK:       lx.atomic.exit281:
+// CHECK-NEXT:    [[TMP2884:%.*]] = extractvalue { i64, i1 } [[TMP2881]], 1
+// CHECK-NEXT:    [[TMP2885:%.*]] = zext i1 [[TMP2884]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2885]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP2886:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2887:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2886]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2887]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2888:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2889:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2888]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2889]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2890:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2891:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2890]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2891]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2892:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2893:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2892]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2893]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2894:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2895:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2896:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2894]], i64 [[TMP2895]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2897:%.*]] = extractvalue { i64, i1 } [[TMP2896]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP2897]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2898:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2899:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2900:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2898]], i64 [[TMP2899]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2901:%.*]] = extractvalue { i64, i1 } [[TMP2900]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP2901]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2902:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2903:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2902]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP2904:%.*]] = icmp ugt i64 [[TMP2903]], [[TMP2902]]
+// CHECK-NEXT:    [[TMP2905:%.*]] = select i1 [[TMP2904]], i64 [[TMP2902]], i64 [[TMP2903]]
+// CHECK-NEXT:    store volatile i64 [[TMP2905]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2906:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2907:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2906]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP2908:%.*]] = icmp ult i64 [[TMP2907]], [[TMP2906]]
+// CHECK-NEXT:    [[TMP2909:%.*]] = select i1 [[TMP2908]], i64 [[TMP2906]], i64 [[TMP2907]]
+// CHECK-NEXT:    store volatile i64 [[TMP2909]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2910:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2911:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2910]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP2912:%.*]] = icmp ult i64 [[TMP2911]], [[TMP2910]]
+// CHECK-NEXT:    [[TMP2913:%.*]] = select i1 [[TMP2912]], i64 [[TMP2910]], i64 [[TMP2911]]
+// CHECK-NEXT:    store volatile i64 [[TMP2913]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2914:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2915:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2914]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP2916:%.*]] = icmp ugt i64 [[TMP2915]], [[TMP2914]]
+// CHECK-NEXT:    [[TMP2917:%.*]] = select i1 [[TMP2916]], i64 [[TMP2914]], i64 [[TMP2915]]
+// CHECK-NEXT:    store volatile i64 [[TMP2917]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2918:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2919:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2920:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2918]], i64 [[TMP2919]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2921:%.*]] = extractvalue { i64, i1 } [[TMP2920]], 0
+// CHECK-NEXT:    [[TMP2922:%.*]] = extractvalue { i64, i1 } [[TMP2920]], 1
+// CHECK-NEXT:    [[TMP2923:%.*]] = select i1 [[TMP2922]], i64 [[TMP2918]], i64 [[TMP2921]]
+// CHECK-NEXT:    store volatile i64 [[TMP2923]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2924:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2925:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2926:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2924]], i64 [[TMP2925]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2927:%.*]] = extractvalue { i64, i1 } [[TMP2926]], 0
+// CHECK-NEXT:    [[TMP2928:%.*]] = extractvalue { i64, i1 } [[TMP2926]], 1
+// CHECK-NEXT:    [[TMP2929:%.*]] = select i1 [[TMP2928]], i64 [[TMP2924]], i64 [[TMP2927]]
+// CHECK-NEXT:    store volatile i64 [[TMP2929]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2930:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2931:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2932:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2930]], i64 [[TMP2931]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2933:%.*]] = extractvalue { i64, i1 } [[TMP2932]], 0
+// CHECK-NEXT:    [[TMP2934:%.*]] = extractvalue { i64, i1 } [[TMP2932]], 1
+// CHECK-NEXT:    br i1 [[TMP2934]], label [[LX_ATOMIC_EXIT283:%.*]], label [[LX_ATOMIC_CONT284:%.*]]
+// CHECK:       lx.atomic.cont284:
+// CHECK-NEXT:    store i64 [[TMP2933]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT283]]
+// CHECK:       lx.atomic.exit283:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2935:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2936:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2937:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2935]], i64 [[TMP2936]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2938:%.*]] = extractvalue { i64, i1 } [[TMP2937]], 0
+// CHECK-NEXT:    [[TMP2939:%.*]] = extractvalue { i64, i1 } [[TMP2937]], 1
+// CHECK-NEXT:    br i1 [[TMP2939]], label [[LX_ATOMIC_EXIT285:%.*]], label [[LX_ATOMIC_CONT286:%.*]]
+// CHECK:       lx.atomic.cont286:
+// CHECK-NEXT:    store i64 [[TMP2938]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT285]]
+// CHECK:       lx.atomic.exit285:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2940:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2941:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2942:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2940]], i64 [[TMP2941]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2943:%.*]] = extractvalue { i64, i1 } [[TMP2942]], 1
+// CHECK-NEXT:    [[TMP2944:%.*]] = zext i1 [[TMP2943]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2944]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2945:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2946:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2947:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2945]], i64 [[TMP2946]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2948:%.*]] = extractvalue { i64, i1 } [[TMP2947]], 1
+// CHECK-NEXT:    [[TMP2949:%.*]] = zext i1 [[TMP2948]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2949]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2950:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2951:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2952:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2950]], i64 [[TMP2951]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2953:%.*]] = extractvalue { i64, i1 } [[TMP2952]], 0
+// CHECK-NEXT:    [[TMP2954:%.*]] = extractvalue { i64, i1 } [[TMP2952]], 1
+// CHECK-NEXT:    br i1 [[TMP2954]], label [[LX_ATOMIC_EXIT287:%.*]], label [[LX_ATOMIC_CONT288:%.*]]
+// CHECK:       lx.atomic.cont288:
+// CHECK-NEXT:    store i64 [[TMP2953]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT287]]
+// CHECK:       lx.atomic.exit287:
+// CHECK-NEXT:    [[TMP2955:%.*]] = extractvalue { i64, i1 } [[TMP2952]], 1
+// CHECK-NEXT:    [[TMP2956:%.*]] = zext i1 [[TMP2955]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2956]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2957:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2958:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2959:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2957]], i64 [[TMP2958]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP2960:%.*]] = extractvalue { i64, i1 } [[TMP2959]], 0
+// CHECK-NEXT:    [[TMP2961:%.*]] = extractvalue { i64, i1 } [[TMP2959]], 1
+// CHECK-NEXT:    br i1 [[TMP2961]], label [[LX_ATOMIC_EXIT289:%.*]], label [[LX_ATOMIC_CONT290:%.*]]
+// CHECK:       lx.atomic.cont290:
+// CHECK-NEXT:    store i64 [[TMP2960]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT289]]
+// CHECK:       lx.atomic.exit289:
+// CHECK-NEXT:    [[TMP2962:%.*]] = extractvalue { i64, i1 } [[TMP2959]], 1
+// CHECK-NEXT:    [[TMP2963:%.*]] = zext i1 [[TMP2962]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP2963]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP2964:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2965:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2964]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2965]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2966:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2967:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2966]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2967]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2968:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2969:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2968]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2969]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2970:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2971:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2970]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP2971]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2972:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2973:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2974:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2972]], i64 [[TMP2973]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP2975:%.*]] = extractvalue { i64, i1 } [[TMP2974]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP2975]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2976:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2977:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2978:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2976]], i64 [[TMP2977]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP2979:%.*]] = extractvalue { i64, i1 } [[TMP2978]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP2979]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2980:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2981:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2980]] acquire, align 8
+// CHECK-NEXT:    [[TMP2982:%.*]] = icmp ugt i64 [[TMP2981]], [[TMP2980]]
+// CHECK-NEXT:    [[TMP2983:%.*]] = select i1 [[TMP2982]], i64 [[TMP2980]], i64 [[TMP2981]]
+// CHECK-NEXT:    store volatile i64 [[TMP2983]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2984:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2985:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2984]] acquire, align 8
+// CHECK-NEXT:    [[TMP2986:%.*]] = icmp ult i64 [[TMP2985]], [[TMP2984]]
+// CHECK-NEXT:    [[TMP2987:%.*]] = select i1 [[TMP2986]], i64 [[TMP2984]], i64 [[TMP2985]]
+// CHECK-NEXT:    store volatile i64 [[TMP2987]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2988:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2989:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP2988]] acquire, align 8
+// CHECK-NEXT:    [[TMP2990:%.*]] = icmp ult i64 [[TMP2989]], [[TMP2988]]
+// CHECK-NEXT:    [[TMP2991:%.*]] = select i1 [[TMP2990]], i64 [[TMP2988]], i64 [[TMP2989]]
+// CHECK-NEXT:    store volatile i64 [[TMP2991]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2992:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2993:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP2992]] acquire, align 8
+// CHECK-NEXT:    [[TMP2994:%.*]] = icmp ugt i64 [[TMP2993]], [[TMP2992]]
+// CHECK-NEXT:    [[TMP2995:%.*]] = select i1 [[TMP2994]], i64 [[TMP2992]], i64 [[TMP2993]]
+// CHECK-NEXT:    store volatile i64 [[TMP2995]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP2996:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP2997:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP2998:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP2996]], i64 [[TMP2997]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP2999:%.*]] = extractvalue { i64, i1 } [[TMP2998]], 0
+// CHECK-NEXT:    [[TMP3000:%.*]] = extractvalue { i64, i1 } [[TMP2998]], 1
+// CHECK-NEXT:    [[TMP3001:%.*]] = select i1 [[TMP3000]], i64 [[TMP2996]], i64 [[TMP2999]]
+// CHECK-NEXT:    store volatile i64 [[TMP3001]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3002:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3003:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3004:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3002]], i64 [[TMP3003]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3005:%.*]] = extractvalue { i64, i1 } [[TMP3004]], 0
+// CHECK-NEXT:    [[TMP3006:%.*]] = extractvalue { i64, i1 } [[TMP3004]], 1
+// CHECK-NEXT:    [[TMP3007:%.*]] = select i1 [[TMP3006]], i64 [[TMP3002]], i64 [[TMP3005]]
+// CHECK-NEXT:    store volatile i64 [[TMP3007]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3008:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3009:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3010:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3008]], i64 [[TMP3009]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3011:%.*]] = extractvalue { i64, i1 } [[TMP3010]], 0
+// CHECK-NEXT:    [[TMP3012:%.*]] = extractvalue { i64, i1 } [[TMP3010]], 1
+// CHECK-NEXT:    br i1 [[TMP3012]], label [[LX_ATOMIC_EXIT291:%.*]], label [[LX_ATOMIC_CONT292:%.*]]
+// CHECK:       lx.atomic.cont292:
+// CHECK-NEXT:    store i64 [[TMP3011]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT291]]
+// CHECK:       lx.atomic.exit291:
+// CHECK-NEXT:    [[TMP3013:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3014:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3015:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3013]], i64 [[TMP3014]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3016:%.*]] = extractvalue { i64, i1 } [[TMP3015]], 0
+// CHECK-NEXT:    [[TMP3017:%.*]] = extractvalue { i64, i1 } [[TMP3015]], 1
+// CHECK-NEXT:    br i1 [[TMP3017]], label [[LX_ATOMIC_EXIT293:%.*]], label [[LX_ATOMIC_CONT294:%.*]]
+// CHECK:       lx.atomic.cont294:
+// CHECK-NEXT:    store i64 [[TMP3016]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT293]]
+// CHECK:       lx.atomic.exit293:
+// CHECK-NEXT:    [[TMP3018:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3019:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3020:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3018]], i64 [[TMP3019]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3021:%.*]] = extractvalue { i64, i1 } [[TMP3020]], 1
+// CHECK-NEXT:    [[TMP3022:%.*]] = zext i1 [[TMP3021]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3022]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3023:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3024:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3025:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3023]], i64 [[TMP3024]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3026:%.*]] = extractvalue { i64, i1 } [[TMP3025]], 1
+// CHECK-NEXT:    [[TMP3027:%.*]] = zext i1 [[TMP3026]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3027]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3028:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3029:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3030:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3028]], i64 [[TMP3029]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3031:%.*]] = extractvalue { i64, i1 } [[TMP3030]], 0
+// CHECK-NEXT:    [[TMP3032:%.*]] = extractvalue { i64, i1 } [[TMP3030]], 1
+// CHECK-NEXT:    br i1 [[TMP3032]], label [[LX_ATOMIC_EXIT295:%.*]], label [[LX_ATOMIC_CONT296:%.*]]
+// CHECK:       lx.atomic.cont296:
+// CHECK-NEXT:    store i64 [[TMP3031]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT295]]
+// CHECK:       lx.atomic.exit295:
+// CHECK-NEXT:    [[TMP3033:%.*]] = extractvalue { i64, i1 } [[TMP3030]], 1
+// CHECK-NEXT:    [[TMP3034:%.*]] = zext i1 [[TMP3033]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3034]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3035:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3036:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3037:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3035]], i64 [[TMP3036]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3038:%.*]] = extractvalue { i64, i1 } [[TMP3037]], 0
+// CHECK-NEXT:    [[TMP3039:%.*]] = extractvalue { i64, i1 } [[TMP3037]], 1
+// CHECK-NEXT:    br i1 [[TMP3039]], label [[LX_ATOMIC_EXIT297:%.*]], label [[LX_ATOMIC_CONT298:%.*]]
+// CHECK:       lx.atomic.cont298:
+// CHECK-NEXT:    store i64 [[TMP3038]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT297]]
+// CHECK:       lx.atomic.exit297:
+// CHECK-NEXT:    [[TMP3040:%.*]] = extractvalue { i64, i1 } [[TMP3037]], 1
+// CHECK-NEXT:    [[TMP3041:%.*]] = zext i1 [[TMP3040]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3041]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3042:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3043:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3042]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3043]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3044:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3045:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3044]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3045]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3046:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3047:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3046]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3047]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3048:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3049:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3048]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3049]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3050:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3051:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3052:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3050]], i64 [[TMP3051]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3053:%.*]] = extractvalue { i64, i1 } [[TMP3052]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3053]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3054:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3055:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3056:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3054]], i64 [[TMP3055]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3057:%.*]] = extractvalue { i64, i1 } [[TMP3056]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3057]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3058:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3059:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3058]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3060:%.*]] = icmp ugt i64 [[TMP3059]], [[TMP3058]]
+// CHECK-NEXT:    [[TMP3061:%.*]] = select i1 [[TMP3060]], i64 [[TMP3058]], i64 [[TMP3059]]
+// CHECK-NEXT:    store volatile i64 [[TMP3061]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3062:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3063:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3062]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3064:%.*]] = icmp ult i64 [[TMP3063]], [[TMP3062]]
+// CHECK-NEXT:    [[TMP3065:%.*]] = select i1 [[TMP3064]], i64 [[TMP3062]], i64 [[TMP3063]]
+// CHECK-NEXT:    store volatile i64 [[TMP3065]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3066:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3067:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3066]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3068:%.*]] = icmp ult i64 [[TMP3067]], [[TMP3066]]
+// CHECK-NEXT:    [[TMP3069:%.*]] = select i1 [[TMP3068]], i64 [[TMP3066]], i64 [[TMP3067]]
+// CHECK-NEXT:    store volatile i64 [[TMP3069]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3070:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3071:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3070]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3072:%.*]] = icmp ugt i64 [[TMP3071]], [[TMP3070]]
+// CHECK-NEXT:    [[TMP3073:%.*]] = select i1 [[TMP3072]], i64 [[TMP3070]], i64 [[TMP3071]]
+// CHECK-NEXT:    store volatile i64 [[TMP3073]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3074:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3075:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3076:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3074]], i64 [[TMP3075]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3077:%.*]] = extractvalue { i64, i1 } [[TMP3076]], 0
+// CHECK-NEXT:    [[TMP3078:%.*]] = extractvalue { i64, i1 } [[TMP3076]], 1
+// CHECK-NEXT:    [[TMP3079:%.*]] = select i1 [[TMP3078]], i64 [[TMP3074]], i64 [[TMP3077]]
+// CHECK-NEXT:    store volatile i64 [[TMP3079]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3080:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3081:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3082:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3080]], i64 [[TMP3081]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3083:%.*]] = extractvalue { i64, i1 } [[TMP3082]], 0
+// CHECK-NEXT:    [[TMP3084:%.*]] = extractvalue { i64, i1 } [[TMP3082]], 1
+// CHECK-NEXT:    [[TMP3085:%.*]] = select i1 [[TMP3084]], i64 [[TMP3080]], i64 [[TMP3083]]
+// CHECK-NEXT:    store volatile i64 [[TMP3085]], i64* [[LV]], align 8
+// CHECK-NEXT:    [[TMP3086:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3087:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3088:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3086]], i64 [[TMP3087]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3089:%.*]] = extractvalue { i64, i1 } [[TMP3088]], 0
+// CHECK-NEXT:    [[TMP3090:%.*]] = extractvalue { i64, i1 } [[TMP3088]], 1
+// CHECK-NEXT:    br i1 [[TMP3090]], label [[LX_ATOMIC_EXIT299:%.*]], label [[LX_ATOMIC_CONT300:%.*]]
+// CHECK:       lx.atomic.cont300:
+// CHECK-NEXT:    store i64 [[TMP3089]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT299]]
+// CHECK:       lx.atomic.exit299:
+// CHECK-NEXT:    [[TMP3091:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3092:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3093:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3091]], i64 [[TMP3092]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3094:%.*]] = extractvalue { i64, i1 } [[TMP3093]], 0
+// CHECK-NEXT:    [[TMP3095:%.*]] = extractvalue { i64, i1 } [[TMP3093]], 1
+// CHECK-NEXT:    br i1 [[TMP3095]], label [[LX_ATOMIC_EXIT301:%.*]], label [[LX_ATOMIC_CONT302:%.*]]
+// CHECK:       lx.atomic.cont302:
+// CHECK-NEXT:    store i64 [[TMP3094]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT301]]
+// CHECK:       lx.atomic.exit301:
+// CHECK-NEXT:    [[TMP3096:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3097:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3098:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3096]], i64 [[TMP3097]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3099:%.*]] = extractvalue { i64, i1 } [[TMP3098]], 1
+// CHECK-NEXT:    [[TMP3100:%.*]] = zext i1 [[TMP3099]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3100]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3101:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3102:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3103:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3101]], i64 [[TMP3102]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3104:%.*]] = extractvalue { i64, i1 } [[TMP3103]], 1
+// CHECK-NEXT:    [[TMP3105:%.*]] = zext i1 [[TMP3104]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3105]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3106:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3107:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3108:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3106]], i64 [[TMP3107]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3109:%.*]] = extractvalue { i64, i1 } [[TMP3108]], 0
+// CHECK-NEXT:    [[TMP3110:%.*]] = extractvalue { i64, i1 } [[TMP3108]], 1
+// CHECK-NEXT:    br i1 [[TMP3110]], label [[LX_ATOMIC_EXIT303:%.*]], label [[LX_ATOMIC_CONT304:%.*]]
+// CHECK:       lx.atomic.cont304:
+// CHECK-NEXT:    store i64 [[TMP3109]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT303]]
+// CHECK:       lx.atomic.exit303:
+// CHECK-NEXT:    [[TMP3111:%.*]] = extractvalue { i64, i1 } [[TMP3108]], 1
+// CHECK-NEXT:    [[TMP3112:%.*]] = zext i1 [[TMP3111]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3112]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3113:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3114:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3115:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3113]], i64 [[TMP3114]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3116:%.*]] = extractvalue { i64, i1 } [[TMP3115]], 0
+// CHECK-NEXT:    [[TMP3117:%.*]] = extractvalue { i64, i1 } [[TMP3115]], 1
+// CHECK-NEXT:    br i1 [[TMP3117]], label [[LX_ATOMIC_EXIT305:%.*]], label [[LX_ATOMIC_CONT306:%.*]]
+// CHECK:       lx.atomic.cont306:
+// CHECK-NEXT:    store i64 [[TMP3116]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT305]]
+// CHECK:       lx.atomic.exit305:
+// CHECK-NEXT:    [[TMP3118:%.*]] = extractvalue { i64, i1 } [[TMP3115]], 1
+// CHECK-NEXT:    [[TMP3119:%.*]] = zext i1 [[TMP3118]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3119]], i64* [[LR]], align 8
+// CHECK-NEXT:    [[TMP3120:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3121:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3120]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3121]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3122:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3123:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3122]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3123]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3124:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3125:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3124]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3125]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3126:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3127:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3126]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3127]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3128:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3129:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3130:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3128]], i64 [[TMP3129]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3131:%.*]] = extractvalue { i64, i1 } [[TMP3130]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3131]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3132:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3133:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3134:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3132]], i64 [[TMP3133]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3135:%.*]] = extractvalue { i64, i1 } [[TMP3134]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3135]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3136:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3137:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3136]] release, align 8
+// CHECK-NEXT:    [[TMP3138:%.*]] = icmp ugt i64 [[TMP3137]], [[TMP3136]]
+// CHECK-NEXT:    [[TMP3139:%.*]] = select i1 [[TMP3138]], i64 [[TMP3136]], i64 [[TMP3137]]
+// CHECK-NEXT:    store volatile i64 [[TMP3139]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3140:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3141:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3140]] release, align 8
+// CHECK-NEXT:    [[TMP3142:%.*]] = icmp ult i64 [[TMP3141]], [[TMP3140]]
+// CHECK-NEXT:    [[TMP3143:%.*]] = select i1 [[TMP3142]], i64 [[TMP3140]], i64 [[TMP3141]]
+// CHECK-NEXT:    store volatile i64 [[TMP3143]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3144:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3145:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3144]] release, align 8
+// CHECK-NEXT:    [[TMP3146:%.*]] = icmp ult i64 [[TMP3145]], [[TMP3144]]
+// CHECK-NEXT:    [[TMP3147:%.*]] = select i1 [[TMP3146]], i64 [[TMP3144]], i64 [[TMP3145]]
+// CHECK-NEXT:    store volatile i64 [[TMP3147]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3148:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3149:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3148]] release, align 8
+// CHECK-NEXT:    [[TMP3150:%.*]] = icmp ugt i64 [[TMP3149]], [[TMP3148]]
+// CHECK-NEXT:    [[TMP3151:%.*]] = select i1 [[TMP3150]], i64 [[TMP3148]], i64 [[TMP3149]]
+// CHECK-NEXT:    store volatile i64 [[TMP3151]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3152:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3153:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3154:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3152]], i64 [[TMP3153]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3155:%.*]] = extractvalue { i64, i1 } [[TMP3154]], 0
+// CHECK-NEXT:    [[TMP3156:%.*]] = extractvalue { i64, i1 } [[TMP3154]], 1
+// CHECK-NEXT:    [[TMP3157:%.*]] = select i1 [[TMP3156]], i64 [[TMP3152]], i64 [[TMP3155]]
+// CHECK-NEXT:    store volatile i64 [[TMP3157]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3158:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3159:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3160:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3158]], i64 [[TMP3159]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3161:%.*]] = extractvalue { i64, i1 } [[TMP3160]], 0
+// CHECK-NEXT:    [[TMP3162:%.*]] = extractvalue { i64, i1 } [[TMP3160]], 1
+// CHECK-NEXT:    [[TMP3163:%.*]] = select i1 [[TMP3162]], i64 [[TMP3158]], i64 [[TMP3161]]
+// CHECK-NEXT:    store volatile i64 [[TMP3163]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3164:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3165:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3166:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3164]], i64 [[TMP3165]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3167:%.*]] = extractvalue { i64, i1 } [[TMP3166]], 0
+// CHECK-NEXT:    [[TMP3168:%.*]] = extractvalue { i64, i1 } [[TMP3166]], 1
+// CHECK-NEXT:    br i1 [[TMP3168]], label [[LX_ATOMIC_EXIT307:%.*]], label [[LX_ATOMIC_CONT308:%.*]]
+// CHECK:       lx.atomic.cont308:
+// CHECK-NEXT:    store i64 [[TMP3167]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT307]]
+// CHECK:       lx.atomic.exit307:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3169:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3170:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3171:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3169]], i64 [[TMP3170]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3172:%.*]] = extractvalue { i64, i1 } [[TMP3171]], 0
+// CHECK-NEXT:    [[TMP3173:%.*]] = extractvalue { i64, i1 } [[TMP3171]], 1
+// CHECK-NEXT:    br i1 [[TMP3173]], label [[LX_ATOMIC_EXIT309:%.*]], label [[LX_ATOMIC_CONT310:%.*]]
+// CHECK:       lx.atomic.cont310:
+// CHECK-NEXT:    store i64 [[TMP3172]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT309]]
+// CHECK:       lx.atomic.exit309:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3174:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3175:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3176:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3174]], i64 [[TMP3175]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3177:%.*]] = extractvalue { i64, i1 } [[TMP3176]], 1
+// CHECK-NEXT:    [[TMP3178:%.*]] = zext i1 [[TMP3177]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3178]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3179:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3180:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3181:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3179]], i64 [[TMP3180]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3182:%.*]] = extractvalue { i64, i1 } [[TMP3181]], 1
+// CHECK-NEXT:    [[TMP3183:%.*]] = zext i1 [[TMP3182]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3183]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3184:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3185:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3186:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3184]], i64 [[TMP3185]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3187:%.*]] = extractvalue { i64, i1 } [[TMP3186]], 0
+// CHECK-NEXT:    [[TMP3188:%.*]] = extractvalue { i64, i1 } [[TMP3186]], 1
+// CHECK-NEXT:    br i1 [[TMP3188]], label [[LX_ATOMIC_EXIT311:%.*]], label [[LX_ATOMIC_CONT312:%.*]]
+// CHECK:       lx.atomic.cont312:
+// CHECK-NEXT:    store i64 [[TMP3187]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT311]]
+// CHECK:       lx.atomic.exit311:
+// CHECK-NEXT:    [[TMP3189:%.*]] = extractvalue { i64, i1 } [[TMP3186]], 1
+// CHECK-NEXT:    [[TMP3190:%.*]] = zext i1 [[TMP3189]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3190]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3191:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3192:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3193:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3191]], i64 [[TMP3192]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3194:%.*]] = extractvalue { i64, i1 } [[TMP3193]], 0
+// CHECK-NEXT:    [[TMP3195:%.*]] = extractvalue { i64, i1 } [[TMP3193]], 1
+// CHECK-NEXT:    br i1 [[TMP3195]], label [[LX_ATOMIC_EXIT313:%.*]], label [[LX_ATOMIC_CONT314:%.*]]
+// CHECK:       lx.atomic.cont314:
+// CHECK-NEXT:    store i64 [[TMP3194]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT313]]
+// CHECK:       lx.atomic.exit313:
+// CHECK-NEXT:    [[TMP3196:%.*]] = extractvalue { i64, i1 } [[TMP3193]], 1
+// CHECK-NEXT:    [[TMP3197:%.*]] = zext i1 [[TMP3196]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3197]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3198:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3199:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3198]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3199]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3200:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3201:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3200]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3201]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3202:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3203:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3202]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3203]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3204:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3205:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3204]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3205]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3206:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3207:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3208:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3206]], i64 [[TMP3207]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3209:%.*]] = extractvalue { i64, i1 } [[TMP3208]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3209]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3210:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3211:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3212:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3210]], i64 [[TMP3211]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3213:%.*]] = extractvalue { i64, i1 } [[TMP3212]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3213]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3214:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3215:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3214]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3216:%.*]] = icmp ugt i64 [[TMP3215]], [[TMP3214]]
+// CHECK-NEXT:    [[TMP3217:%.*]] = select i1 [[TMP3216]], i64 [[TMP3214]], i64 [[TMP3215]]
+// CHECK-NEXT:    store volatile i64 [[TMP3217]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3218:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3219:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3218]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3220:%.*]] = icmp ult i64 [[TMP3219]], [[TMP3218]]
+// CHECK-NEXT:    [[TMP3221:%.*]] = select i1 [[TMP3220]], i64 [[TMP3218]], i64 [[TMP3219]]
+// CHECK-NEXT:    store volatile i64 [[TMP3221]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3222:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3223:%.*]] = atomicrmw umin i64* [[LX]], i64 [[TMP3222]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3224:%.*]] = icmp ult i64 [[TMP3223]], [[TMP3222]]
+// CHECK-NEXT:    [[TMP3225:%.*]] = select i1 [[TMP3224]], i64 [[TMP3222]], i64 [[TMP3223]]
+// CHECK-NEXT:    store volatile i64 [[TMP3225]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3226:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3227:%.*]] = atomicrmw umax i64* [[LX]], i64 [[TMP3226]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3228:%.*]] = icmp ugt i64 [[TMP3227]], [[TMP3226]]
+// CHECK-NEXT:    [[TMP3229:%.*]] = select i1 [[TMP3228]], i64 [[TMP3226]], i64 [[TMP3227]]
+// CHECK-NEXT:    store volatile i64 [[TMP3229]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3230:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3231:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3232:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3230]], i64 [[TMP3231]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3233:%.*]] = extractvalue { i64, i1 } [[TMP3232]], 0
+// CHECK-NEXT:    [[TMP3234:%.*]] = extractvalue { i64, i1 } [[TMP3232]], 1
+// CHECK-NEXT:    [[TMP3235:%.*]] = select i1 [[TMP3234]], i64 [[TMP3230]], i64 [[TMP3233]]
+// CHECK-NEXT:    store volatile i64 [[TMP3235]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3236:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3237:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3238:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3236]], i64 [[TMP3237]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3239:%.*]] = extractvalue { i64, i1 } [[TMP3238]], 0
+// CHECK-NEXT:    [[TMP3240:%.*]] = extractvalue { i64, i1 } [[TMP3238]], 1
+// CHECK-NEXT:    [[TMP3241:%.*]] = select i1 [[TMP3240]], i64 [[TMP3236]], i64 [[TMP3239]]
+// CHECK-NEXT:    store volatile i64 [[TMP3241]], i64* [[LV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3242:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3243:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3244:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3242]], i64 [[TMP3243]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3245:%.*]] = extractvalue { i64, i1 } [[TMP3244]], 0
+// CHECK-NEXT:    [[TMP3246:%.*]] = extractvalue { i64, i1 } [[TMP3244]], 1
+// CHECK-NEXT:    br i1 [[TMP3246]], label [[LX_ATOMIC_EXIT315:%.*]], label [[LX_ATOMIC_CONT316:%.*]]
+// CHECK:       lx.atomic.cont316:
+// CHECK-NEXT:    store i64 [[TMP3245]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT315]]
+// CHECK:       lx.atomic.exit315:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3247:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3248:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3249:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3247]], i64 [[TMP3248]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3250:%.*]] = extractvalue { i64, i1 } [[TMP3249]], 0
+// CHECK-NEXT:    [[TMP3251:%.*]] = extractvalue { i64, i1 } [[TMP3249]], 1
+// CHECK-NEXT:    br i1 [[TMP3251]], label [[LX_ATOMIC_EXIT317:%.*]], label [[LX_ATOMIC_CONT318:%.*]]
+// CHECK:       lx.atomic.cont318:
+// CHECK-NEXT:    store i64 [[TMP3250]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT317]]
+// CHECK:       lx.atomic.exit317:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3252:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3253:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3254:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3252]], i64 [[TMP3253]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3255:%.*]] = extractvalue { i64, i1 } [[TMP3254]], 1
+// CHECK-NEXT:    [[TMP3256:%.*]] = zext i1 [[TMP3255]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3256]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3257:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3258:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3259:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3257]], i64 [[TMP3258]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3260:%.*]] = extractvalue { i64, i1 } [[TMP3259]], 1
+// CHECK-NEXT:    [[TMP3261:%.*]] = zext i1 [[TMP3260]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3261]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3262:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3263:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3264:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3262]], i64 [[TMP3263]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3265:%.*]] = extractvalue { i64, i1 } [[TMP3264]], 0
+// CHECK-NEXT:    [[TMP3266:%.*]] = extractvalue { i64, i1 } [[TMP3264]], 1
+// CHECK-NEXT:    br i1 [[TMP3266]], label [[LX_ATOMIC_EXIT319:%.*]], label [[LX_ATOMIC_CONT320:%.*]]
+// CHECK:       lx.atomic.cont320:
+// CHECK-NEXT:    store i64 [[TMP3265]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT319]]
+// CHECK:       lx.atomic.exit319:
+// CHECK-NEXT:    [[TMP3267:%.*]] = extractvalue { i64, i1 } [[TMP3264]], 1
+// CHECK-NEXT:    [[TMP3268:%.*]] = zext i1 [[TMP3267]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3268]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3269:%.*]] = load i64, i64* [[LE]], align 8
+// CHECK-NEXT:    [[TMP3270:%.*]] = load i64, i64* [[LD]], align 8
+// CHECK-NEXT:    [[TMP3271:%.*]] = cmpxchg i64* [[LX]], i64 [[TMP3269]], i64 [[TMP3270]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3272:%.*]] = extractvalue { i64, i1 } [[TMP3271]], 0
+// CHECK-NEXT:    [[TMP3273:%.*]] = extractvalue { i64, i1 } [[TMP3271]], 1
+// CHECK-NEXT:    br i1 [[TMP3273]], label [[LX_ATOMIC_EXIT321:%.*]], label [[LX_ATOMIC_CONT322:%.*]]
+// CHECK:       lx.atomic.cont322:
+// CHECK-NEXT:    store i64 [[TMP3272]], i64* [[LV]], align 8
+// CHECK-NEXT:    br label [[LX_ATOMIC_EXIT321]]
+// CHECK:       lx.atomic.exit321:
+// CHECK-NEXT:    [[TMP3274:%.*]] = extractvalue { i64, i1 } [[TMP3271]], 1
+// CHECK-NEXT:    [[TMP3275:%.*]] = zext i1 [[TMP3274]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3275]], i64* [[LR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3276:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3277:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3276]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3277]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3278:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3279:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3278]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3279]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3280:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3281:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3280]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3281]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3282:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3283:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3282]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3283]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3284:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3285:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3286:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3284]], i64 [[TMP3285]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3287:%.*]] = extractvalue { i64, i1 } [[TMP3286]], 0
+// CHECK-NEXT:    store i64 [[TMP3287]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3288:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3289:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3290:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3288]], i64 [[TMP3289]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3291:%.*]] = extractvalue { i64, i1 } [[TMP3290]], 0
+// CHECK-NEXT:    store i64 [[TMP3291]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3292:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3293:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3292]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3294:%.*]] = icmp ugt i64 [[TMP3293]], [[TMP3292]]
+// CHECK-NEXT:    [[TMP3295:%.*]] = select i1 [[TMP3294]], i64 [[TMP3292]], i64 [[TMP3293]]
+// CHECK-NEXT:    store i64 [[TMP3295]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3296:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3297:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3296]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3298:%.*]] = icmp ult i64 [[TMP3297]], [[TMP3296]]
+// CHECK-NEXT:    [[TMP3299:%.*]] = select i1 [[TMP3298]], i64 [[TMP3296]], i64 [[TMP3297]]
+// CHECK-NEXT:    store i64 [[TMP3299]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3300:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3301:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3300]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3302:%.*]] = icmp ult i64 [[TMP3301]], [[TMP3300]]
+// CHECK-NEXT:    [[TMP3303:%.*]] = select i1 [[TMP3302]], i64 [[TMP3300]], i64 [[TMP3301]]
+// CHECK-NEXT:    store i64 [[TMP3303]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3304:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3305:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3304]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3306:%.*]] = icmp ugt i64 [[TMP3305]], [[TMP3304]]
+// CHECK-NEXT:    [[TMP3307:%.*]] = select i1 [[TMP3306]], i64 [[TMP3304]], i64 [[TMP3305]]
+// CHECK-NEXT:    store i64 [[TMP3307]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3308:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3309:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3310:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3308]], i64 [[TMP3309]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3311:%.*]] = extractvalue { i64, i1 } [[TMP3310]], 0
+// CHECK-NEXT:    [[TMP3312:%.*]] = extractvalue { i64, i1 } [[TMP3310]], 1
+// CHECK-NEXT:    [[TMP3313:%.*]] = select i1 [[TMP3312]], i64 [[TMP3308]], i64 [[TMP3311]]
+// CHECK-NEXT:    store i64 [[TMP3313]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3314:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3315:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3316:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3314]], i64 [[TMP3315]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3317:%.*]] = extractvalue { i64, i1 } [[TMP3316]], 0
+// CHECK-NEXT:    [[TMP3318:%.*]] = extractvalue { i64, i1 } [[TMP3316]], 1
+// CHECK-NEXT:    [[TMP3319:%.*]] = select i1 [[TMP3318]], i64 [[TMP3314]], i64 [[TMP3317]]
+// CHECK-NEXT:    store i64 [[TMP3319]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3320:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3321:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3322:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3320]], i64 [[TMP3321]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3323:%.*]] = extractvalue { i64, i1 } [[TMP3322]], 0
+// CHECK-NEXT:    [[TMP3324:%.*]] = extractvalue { i64, i1 } [[TMP3322]], 1
+// CHECK-NEXT:    br i1 [[TMP3324]], label [[ULX_ATOMIC_EXIT:%.*]], label [[ULX_ATOMIC_CONT:%.*]]
+// CHECK:       ulx.atomic.cont:
+// CHECK-NEXT:    store i64 [[TMP3323]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT]]
+// CHECK:       ulx.atomic.exit:
+// CHECK-NEXT:    [[TMP3325:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3326:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3327:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3325]], i64 [[TMP3326]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3328:%.*]] = extractvalue { i64, i1 } [[TMP3327]], 0
+// CHECK-NEXT:    [[TMP3329:%.*]] = extractvalue { i64, i1 } [[TMP3327]], 1
+// CHECK-NEXT:    br i1 [[TMP3329]], label [[ULX_ATOMIC_EXIT323:%.*]], label [[ULX_ATOMIC_CONT324:%.*]]
+// CHECK:       ulx.atomic.cont324:
+// CHECK-NEXT:    store i64 [[TMP3328]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT323]]
+// CHECK:       ulx.atomic.exit323:
+// CHECK-NEXT:    [[TMP3330:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3331:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3332:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3330]], i64 [[TMP3331]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3333:%.*]] = extractvalue { i64, i1 } [[TMP3332]], 1
+// CHECK-NEXT:    [[TMP3334:%.*]] = zext i1 [[TMP3333]] to i64
+// CHECK-NEXT:    store i64 [[TMP3334]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3335:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3336:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3337:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3335]], i64 [[TMP3336]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3338:%.*]] = extractvalue { i64, i1 } [[TMP3337]], 1
+// CHECK-NEXT:    [[TMP3339:%.*]] = zext i1 [[TMP3338]] to i64
+// CHECK-NEXT:    store i64 [[TMP3339]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3340:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3341:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3342:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3340]], i64 [[TMP3341]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3343:%.*]] = extractvalue { i64, i1 } [[TMP3342]], 0
+// CHECK-NEXT:    [[TMP3344:%.*]] = extractvalue { i64, i1 } [[TMP3342]], 1
+// CHECK-NEXT:    br i1 [[TMP3344]], label [[ULX_ATOMIC_EXIT325:%.*]], label [[ULX_ATOMIC_CONT326:%.*]]
+// CHECK:       ulx.atomic.cont326:
+// CHECK-NEXT:    store i64 [[TMP3343]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT325]]
+// CHECK:       ulx.atomic.exit325:
+// CHECK-NEXT:    [[TMP3345:%.*]] = extractvalue { i64, i1 } [[TMP3342]], 1
+// CHECK-NEXT:    [[TMP3346:%.*]] = zext i1 [[TMP3345]] to i64
+// CHECK-NEXT:    store i64 [[TMP3346]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3347:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3348:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3349:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3347]], i64 [[TMP3348]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3350:%.*]] = extractvalue { i64, i1 } [[TMP3349]], 0
+// CHECK-NEXT:    [[TMP3351:%.*]] = extractvalue { i64, i1 } [[TMP3349]], 1
+// CHECK-NEXT:    br i1 [[TMP3351]], label [[ULX_ATOMIC_EXIT327:%.*]], label [[ULX_ATOMIC_CONT328:%.*]]
+// CHECK:       ulx.atomic.cont328:
+// CHECK-NEXT:    store i64 [[TMP3350]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT327]]
+// CHECK:       ulx.atomic.exit327:
+// CHECK-NEXT:    [[TMP3352:%.*]] = extractvalue { i64, i1 } [[TMP3349]], 1
+// CHECK-NEXT:    [[TMP3353:%.*]] = zext i1 [[TMP3352]] to i64
+// CHECK-NEXT:    store i64 [[TMP3353]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3354:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3355:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3354]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP3355]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3356:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3357:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3356]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP3357]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3358:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3359:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3358]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP3359]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3360:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3361:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3360]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP3361]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3362:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3363:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3364:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3362]], i64 [[TMP3363]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3365:%.*]] = extractvalue { i64, i1 } [[TMP3364]], 0
+// CHECK-NEXT:    store i64 [[TMP3365]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3366:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3367:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3368:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3366]], i64 [[TMP3367]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3369:%.*]] = extractvalue { i64, i1 } [[TMP3368]], 0
+// CHECK-NEXT:    store i64 [[TMP3369]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3370:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3371:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3370]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3372:%.*]] = icmp ugt i64 [[TMP3371]], [[TMP3370]]
+// CHECK-NEXT:    [[TMP3373:%.*]] = select i1 [[TMP3372]], i64 [[TMP3370]], i64 [[TMP3371]]
+// CHECK-NEXT:    store i64 [[TMP3373]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3374:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3375:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3374]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3376:%.*]] = icmp ult i64 [[TMP3375]], [[TMP3374]]
+// CHECK-NEXT:    [[TMP3377:%.*]] = select i1 [[TMP3376]], i64 [[TMP3374]], i64 [[TMP3375]]
+// CHECK-NEXT:    store i64 [[TMP3377]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3378:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3379:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3378]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3380:%.*]] = icmp ult i64 [[TMP3379]], [[TMP3378]]
+// CHECK-NEXT:    [[TMP3381:%.*]] = select i1 [[TMP3380]], i64 [[TMP3378]], i64 [[TMP3379]]
+// CHECK-NEXT:    store i64 [[TMP3381]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3382:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3383:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3382]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3384:%.*]] = icmp ugt i64 [[TMP3383]], [[TMP3382]]
+// CHECK-NEXT:    [[TMP3385:%.*]] = select i1 [[TMP3384]], i64 [[TMP3382]], i64 [[TMP3383]]
+// CHECK-NEXT:    store i64 [[TMP3385]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3386:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3387:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3388:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3386]], i64 [[TMP3387]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3389:%.*]] = extractvalue { i64, i1 } [[TMP3388]], 0
+// CHECK-NEXT:    [[TMP3390:%.*]] = extractvalue { i64, i1 } [[TMP3388]], 1
+// CHECK-NEXT:    [[TMP3391:%.*]] = select i1 [[TMP3390]], i64 [[TMP3386]], i64 [[TMP3389]]
+// CHECK-NEXT:    store i64 [[TMP3391]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3392:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3393:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3394:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3392]], i64 [[TMP3393]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3395:%.*]] = extractvalue { i64, i1 } [[TMP3394]], 0
+// CHECK-NEXT:    [[TMP3396:%.*]] = extractvalue { i64, i1 } [[TMP3394]], 1
+// CHECK-NEXT:    [[TMP3397:%.*]] = select i1 [[TMP3396]], i64 [[TMP3392]], i64 [[TMP3395]]
+// CHECK-NEXT:    store i64 [[TMP3397]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3398:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3399:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3400:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3398]], i64 [[TMP3399]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3401:%.*]] = extractvalue { i64, i1 } [[TMP3400]], 0
+// CHECK-NEXT:    [[TMP3402:%.*]] = extractvalue { i64, i1 } [[TMP3400]], 1
+// CHECK-NEXT:    br i1 [[TMP3402]], label [[ULX_ATOMIC_EXIT329:%.*]], label [[ULX_ATOMIC_CONT330:%.*]]
+// CHECK:       ulx.atomic.cont330:
+// CHECK-NEXT:    store i64 [[TMP3401]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT329]]
+// CHECK:       ulx.atomic.exit329:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3403:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3404:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3405:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3403]], i64 [[TMP3404]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3406:%.*]] = extractvalue { i64, i1 } [[TMP3405]], 0
+// CHECK-NEXT:    [[TMP3407:%.*]] = extractvalue { i64, i1 } [[TMP3405]], 1
+// CHECK-NEXT:    br i1 [[TMP3407]], label [[ULX_ATOMIC_EXIT331:%.*]], label [[ULX_ATOMIC_CONT332:%.*]]
+// CHECK:       ulx.atomic.cont332:
+// CHECK-NEXT:    store i64 [[TMP3406]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT331]]
+// CHECK:       ulx.atomic.exit331:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3408:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3409:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3410:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3408]], i64 [[TMP3409]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3411:%.*]] = extractvalue { i64, i1 } [[TMP3410]], 1
+// CHECK-NEXT:    [[TMP3412:%.*]] = zext i1 [[TMP3411]] to i64
+// CHECK-NEXT:    store i64 [[TMP3412]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3413:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3414:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3415:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3413]], i64 [[TMP3414]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3416:%.*]] = extractvalue { i64, i1 } [[TMP3415]], 1
+// CHECK-NEXT:    [[TMP3417:%.*]] = zext i1 [[TMP3416]] to i64
+// CHECK-NEXT:    store i64 [[TMP3417]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3418:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3419:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3420:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3418]], i64 [[TMP3419]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3421:%.*]] = extractvalue { i64, i1 } [[TMP3420]], 0
+// CHECK-NEXT:    [[TMP3422:%.*]] = extractvalue { i64, i1 } [[TMP3420]], 1
+// CHECK-NEXT:    br i1 [[TMP3422]], label [[ULX_ATOMIC_EXIT333:%.*]], label [[ULX_ATOMIC_CONT334:%.*]]
+// CHECK:       ulx.atomic.cont334:
+// CHECK-NEXT:    store i64 [[TMP3421]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT333]]
+// CHECK:       ulx.atomic.exit333:
+// CHECK-NEXT:    [[TMP3423:%.*]] = extractvalue { i64, i1 } [[TMP3420]], 1
+// CHECK-NEXT:    [[TMP3424:%.*]] = zext i1 [[TMP3423]] to i64
+// CHECK-NEXT:    store i64 [[TMP3424]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3425:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3426:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3427:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3425]], i64 [[TMP3426]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3428:%.*]] = extractvalue { i64, i1 } [[TMP3427]], 0
+// CHECK-NEXT:    [[TMP3429:%.*]] = extractvalue { i64, i1 } [[TMP3427]], 1
+// CHECK-NEXT:    br i1 [[TMP3429]], label [[ULX_ATOMIC_EXIT335:%.*]], label [[ULX_ATOMIC_CONT336:%.*]]
+// CHECK:       ulx.atomic.cont336:
+// CHECK-NEXT:    store i64 [[TMP3428]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT335]]
+// CHECK:       ulx.atomic.exit335:
+// CHECK-NEXT:    [[TMP3430:%.*]] = extractvalue { i64, i1 } [[TMP3427]], 1
+// CHECK-NEXT:    [[TMP3431:%.*]] = zext i1 [[TMP3430]] to i64
+// CHECK-NEXT:    store i64 [[TMP3431]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3432:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3433:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3432]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP3433]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3434:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3435:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3434]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP3435]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3436:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3437:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3436]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP3437]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3438:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3439:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3438]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP3439]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3440:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3441:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3442:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3440]], i64 [[TMP3441]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3443:%.*]] = extractvalue { i64, i1 } [[TMP3442]], 0
+// CHECK-NEXT:    store i64 [[TMP3443]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3444:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3445:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3446:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3444]], i64 [[TMP3445]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3447:%.*]] = extractvalue { i64, i1 } [[TMP3446]], 0
+// CHECK-NEXT:    store i64 [[TMP3447]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3448:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3449:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3448]] acquire, align 8
+// CHECK-NEXT:    [[TMP3450:%.*]] = icmp ugt i64 [[TMP3449]], [[TMP3448]]
+// CHECK-NEXT:    [[TMP3451:%.*]] = select i1 [[TMP3450]], i64 [[TMP3448]], i64 [[TMP3449]]
+// CHECK-NEXT:    store i64 [[TMP3451]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3452:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3453:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3452]] acquire, align 8
+// CHECK-NEXT:    [[TMP3454:%.*]] = icmp ult i64 [[TMP3453]], [[TMP3452]]
+// CHECK-NEXT:    [[TMP3455:%.*]] = select i1 [[TMP3454]], i64 [[TMP3452]], i64 [[TMP3453]]
+// CHECK-NEXT:    store i64 [[TMP3455]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3456:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3457:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3456]] acquire, align 8
+// CHECK-NEXT:    [[TMP3458:%.*]] = icmp ult i64 [[TMP3457]], [[TMP3456]]
+// CHECK-NEXT:    [[TMP3459:%.*]] = select i1 [[TMP3458]], i64 [[TMP3456]], i64 [[TMP3457]]
+// CHECK-NEXT:    store i64 [[TMP3459]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3460:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3461:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3460]] acquire, align 8
+// CHECK-NEXT:    [[TMP3462:%.*]] = icmp ugt i64 [[TMP3461]], [[TMP3460]]
+// CHECK-NEXT:    [[TMP3463:%.*]] = select i1 [[TMP3462]], i64 [[TMP3460]], i64 [[TMP3461]]
+// CHECK-NEXT:    store i64 [[TMP3463]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3464:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3465:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3466:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3464]], i64 [[TMP3465]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3467:%.*]] = extractvalue { i64, i1 } [[TMP3466]], 0
+// CHECK-NEXT:    [[TMP3468:%.*]] = extractvalue { i64, i1 } [[TMP3466]], 1
+// CHECK-NEXT:    [[TMP3469:%.*]] = select i1 [[TMP3468]], i64 [[TMP3464]], i64 [[TMP3467]]
+// CHECK-NEXT:    store i64 [[TMP3469]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3470:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3471:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3472:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3470]], i64 [[TMP3471]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3473:%.*]] = extractvalue { i64, i1 } [[TMP3472]], 0
+// CHECK-NEXT:    [[TMP3474:%.*]] = extractvalue { i64, i1 } [[TMP3472]], 1
+// CHECK-NEXT:    [[TMP3475:%.*]] = select i1 [[TMP3474]], i64 [[TMP3470]], i64 [[TMP3473]]
+// CHECK-NEXT:    store i64 [[TMP3475]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3476:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3477:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3478:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3476]], i64 [[TMP3477]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3479:%.*]] = extractvalue { i64, i1 } [[TMP3478]], 0
+// CHECK-NEXT:    [[TMP3480:%.*]] = extractvalue { i64, i1 } [[TMP3478]], 1
+// CHECK-NEXT:    br i1 [[TMP3480]], label [[ULX_ATOMIC_EXIT337:%.*]], label [[ULX_ATOMIC_CONT338:%.*]]
+// CHECK:       ulx.atomic.cont338:
+// CHECK-NEXT:    store i64 [[TMP3479]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT337]]
+// CHECK:       ulx.atomic.exit337:
+// CHECK-NEXT:    [[TMP3481:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3482:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3483:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3481]], i64 [[TMP3482]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3484:%.*]] = extractvalue { i64, i1 } [[TMP3483]], 0
+// CHECK-NEXT:    [[TMP3485:%.*]] = extractvalue { i64, i1 } [[TMP3483]], 1
+// CHECK-NEXT:    br i1 [[TMP3485]], label [[ULX_ATOMIC_EXIT339:%.*]], label [[ULX_ATOMIC_CONT340:%.*]]
+// CHECK:       ulx.atomic.cont340:
+// CHECK-NEXT:    store i64 [[TMP3484]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT339]]
+// CHECK:       ulx.atomic.exit339:
+// CHECK-NEXT:    [[TMP3486:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3487:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3488:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3486]], i64 [[TMP3487]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3489:%.*]] = extractvalue { i64, i1 } [[TMP3488]], 1
+// CHECK-NEXT:    [[TMP3490:%.*]] = zext i1 [[TMP3489]] to i64
+// CHECK-NEXT:    store i64 [[TMP3490]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3491:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3492:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3493:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3491]], i64 [[TMP3492]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3494:%.*]] = extractvalue { i64, i1 } [[TMP3493]], 1
+// CHECK-NEXT:    [[TMP3495:%.*]] = zext i1 [[TMP3494]] to i64
+// CHECK-NEXT:    store i64 [[TMP3495]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3496:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3497:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3498:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3496]], i64 [[TMP3497]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3499:%.*]] = extractvalue { i64, i1 } [[TMP3498]], 0
+// CHECK-NEXT:    [[TMP3500:%.*]] = extractvalue { i64, i1 } [[TMP3498]], 1
+// CHECK-NEXT:    br i1 [[TMP3500]], label [[ULX_ATOMIC_EXIT341:%.*]], label [[ULX_ATOMIC_CONT342:%.*]]
+// CHECK:       ulx.atomic.cont342:
+// CHECK-NEXT:    store i64 [[TMP3499]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT341]]
+// CHECK:       ulx.atomic.exit341:
+// CHECK-NEXT:    [[TMP3501:%.*]] = extractvalue { i64, i1 } [[TMP3498]], 1
+// CHECK-NEXT:    [[TMP3502:%.*]] = zext i1 [[TMP3501]] to i64
+// CHECK-NEXT:    store i64 [[TMP3502]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3503:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3504:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3505:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3503]], i64 [[TMP3504]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3506:%.*]] = extractvalue { i64, i1 } [[TMP3505]], 0
+// CHECK-NEXT:    [[TMP3507:%.*]] = extractvalue { i64, i1 } [[TMP3505]], 1
+// CHECK-NEXT:    br i1 [[TMP3507]], label [[ULX_ATOMIC_EXIT343:%.*]], label [[ULX_ATOMIC_CONT344:%.*]]
+// CHECK:       ulx.atomic.cont344:
+// CHECK-NEXT:    store i64 [[TMP3506]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT343]]
+// CHECK:       ulx.atomic.exit343:
+// CHECK-NEXT:    [[TMP3508:%.*]] = extractvalue { i64, i1 } [[TMP3505]], 1
+// CHECK-NEXT:    [[TMP3509:%.*]] = zext i1 [[TMP3508]] to i64
+// CHECK-NEXT:    store i64 [[TMP3509]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3510:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3511:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3510]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3511]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3512:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3513:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3512]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3513]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3514:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3515:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3514]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3515]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3516:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3517:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3516]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP3517]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3518:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3519:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3520:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3518]], i64 [[TMP3519]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3521:%.*]] = extractvalue { i64, i1 } [[TMP3520]], 0
+// CHECK-NEXT:    store i64 [[TMP3521]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3522:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3523:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3524:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3522]], i64 [[TMP3523]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3525:%.*]] = extractvalue { i64, i1 } [[TMP3524]], 0
+// CHECK-NEXT:    store i64 [[TMP3525]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3526:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3527:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3526]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3528:%.*]] = icmp ugt i64 [[TMP3527]], [[TMP3526]]
+// CHECK-NEXT:    [[TMP3529:%.*]] = select i1 [[TMP3528]], i64 [[TMP3526]], i64 [[TMP3527]]
+// CHECK-NEXT:    store i64 [[TMP3529]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3530:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3531:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3530]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3532:%.*]] = icmp ult i64 [[TMP3531]], [[TMP3530]]
+// CHECK-NEXT:    [[TMP3533:%.*]] = select i1 [[TMP3532]], i64 [[TMP3530]], i64 [[TMP3531]]
+// CHECK-NEXT:    store i64 [[TMP3533]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3534:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3535:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3534]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3536:%.*]] = icmp ult i64 [[TMP3535]], [[TMP3534]]
+// CHECK-NEXT:    [[TMP3537:%.*]] = select i1 [[TMP3536]], i64 [[TMP3534]], i64 [[TMP3535]]
+// CHECK-NEXT:    store i64 [[TMP3537]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3538:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3539:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3538]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3540:%.*]] = icmp ugt i64 [[TMP3539]], [[TMP3538]]
+// CHECK-NEXT:    [[TMP3541:%.*]] = select i1 [[TMP3540]], i64 [[TMP3538]], i64 [[TMP3539]]
+// CHECK-NEXT:    store i64 [[TMP3541]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3542:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3543:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3544:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3542]], i64 [[TMP3543]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3545:%.*]] = extractvalue { i64, i1 } [[TMP3544]], 0
+// CHECK-NEXT:    [[TMP3546:%.*]] = extractvalue { i64, i1 } [[TMP3544]], 1
+// CHECK-NEXT:    [[TMP3547:%.*]] = select i1 [[TMP3546]], i64 [[TMP3542]], i64 [[TMP3545]]
+// CHECK-NEXT:    store i64 [[TMP3547]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3548:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3549:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3550:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3548]], i64 [[TMP3549]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3551:%.*]] = extractvalue { i64, i1 } [[TMP3550]], 0
+// CHECK-NEXT:    [[TMP3552:%.*]] = extractvalue { i64, i1 } [[TMP3550]], 1
+// CHECK-NEXT:    [[TMP3553:%.*]] = select i1 [[TMP3552]], i64 [[TMP3548]], i64 [[TMP3551]]
+// CHECK-NEXT:    store i64 [[TMP3553]], i64* [[ULV]], align 8
+// CHECK-NEXT:    [[TMP3554:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3555:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3556:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3554]], i64 [[TMP3555]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3557:%.*]] = extractvalue { i64, i1 } [[TMP3556]], 0
+// CHECK-NEXT:    [[TMP3558:%.*]] = extractvalue { i64, i1 } [[TMP3556]], 1
+// CHECK-NEXT:    br i1 [[TMP3558]], label [[ULX_ATOMIC_EXIT345:%.*]], label [[ULX_ATOMIC_CONT346:%.*]]
+// CHECK:       ulx.atomic.cont346:
+// CHECK-NEXT:    store i64 [[TMP3557]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT345]]
+// CHECK:       ulx.atomic.exit345:
+// CHECK-NEXT:    [[TMP3559:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3560:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3561:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3559]], i64 [[TMP3560]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3562:%.*]] = extractvalue { i64, i1 } [[TMP3561]], 0
+// CHECK-NEXT:    [[TMP3563:%.*]] = extractvalue { i64, i1 } [[TMP3561]], 1
+// CHECK-NEXT:    br i1 [[TMP3563]], label [[ULX_ATOMIC_EXIT347:%.*]], label [[ULX_ATOMIC_CONT348:%.*]]
+// CHECK:       ulx.atomic.cont348:
+// CHECK-NEXT:    store i64 [[TMP3562]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT347]]
+// CHECK:       ulx.atomic.exit347:
+// CHECK-NEXT:    [[TMP3564:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3565:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3566:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3564]], i64 [[TMP3565]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3567:%.*]] = extractvalue { i64, i1 } [[TMP3566]], 1
+// CHECK-NEXT:    [[TMP3568:%.*]] = zext i1 [[TMP3567]] to i64
+// CHECK-NEXT:    store i64 [[TMP3568]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3569:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3570:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3571:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3569]], i64 [[TMP3570]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3572:%.*]] = extractvalue { i64, i1 } [[TMP3571]], 1
+// CHECK-NEXT:    [[TMP3573:%.*]] = zext i1 [[TMP3572]] to i64
+// CHECK-NEXT:    store i64 [[TMP3573]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3574:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3575:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3576:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3574]], i64 [[TMP3575]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3577:%.*]] = extractvalue { i64, i1 } [[TMP3576]], 0
+// CHECK-NEXT:    [[TMP3578:%.*]] = extractvalue { i64, i1 } [[TMP3576]], 1
+// CHECK-NEXT:    br i1 [[TMP3578]], label [[ULX_ATOMIC_EXIT349:%.*]], label [[ULX_ATOMIC_CONT350:%.*]]
+// CHECK:       ulx.atomic.cont350:
+// CHECK-NEXT:    store i64 [[TMP3577]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT349]]
+// CHECK:       ulx.atomic.exit349:
+// CHECK-NEXT:    [[TMP3579:%.*]] = extractvalue { i64, i1 } [[TMP3576]], 1
+// CHECK-NEXT:    [[TMP3580:%.*]] = zext i1 [[TMP3579]] to i64
+// CHECK-NEXT:    store i64 [[TMP3580]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3581:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3582:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3583:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3581]], i64 [[TMP3582]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3584:%.*]] = extractvalue { i64, i1 } [[TMP3583]], 0
+// CHECK-NEXT:    [[TMP3585:%.*]] = extractvalue { i64, i1 } [[TMP3583]], 1
+// CHECK-NEXT:    br i1 [[TMP3585]], label [[ULX_ATOMIC_EXIT351:%.*]], label [[ULX_ATOMIC_CONT352:%.*]]
+// CHECK:       ulx.atomic.cont352:
+// CHECK-NEXT:    store i64 [[TMP3584]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT351]]
+// CHECK:       ulx.atomic.exit351:
+// CHECK-NEXT:    [[TMP3586:%.*]] = extractvalue { i64, i1 } [[TMP3583]], 1
+// CHECK-NEXT:    [[TMP3587:%.*]] = zext i1 [[TMP3586]] to i64
+// CHECK-NEXT:    store i64 [[TMP3587]], i64* [[ULR]], align 8
+// CHECK-NEXT:    [[TMP3588:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3589:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3588]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP3589]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3590:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3591:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3590]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP3591]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3592:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3593:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3592]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP3593]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3594:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3595:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3594]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP3595]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3596:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3597:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3598:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3596]], i64 [[TMP3597]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3599:%.*]] = extractvalue { i64, i1 } [[TMP3598]], 0
+// CHECK-NEXT:    store i64 [[TMP3599]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3600:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3601:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3602:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3600]], i64 [[TMP3601]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3603:%.*]] = extractvalue { i64, i1 } [[TMP3602]], 0
+// CHECK-NEXT:    store i64 [[TMP3603]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3604:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3605:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3604]] release, align 8
+// CHECK-NEXT:    [[TMP3606:%.*]] = icmp ugt i64 [[TMP3605]], [[TMP3604]]
+// CHECK-NEXT:    [[TMP3607:%.*]] = select i1 [[TMP3606]], i64 [[TMP3604]], i64 [[TMP3605]]
+// CHECK-NEXT:    store i64 [[TMP3607]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3608:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3609:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3608]] release, align 8
+// CHECK-NEXT:    [[TMP3610:%.*]] = icmp ult i64 [[TMP3609]], [[TMP3608]]
+// CHECK-NEXT:    [[TMP3611:%.*]] = select i1 [[TMP3610]], i64 [[TMP3608]], i64 [[TMP3609]]
+// CHECK-NEXT:    store i64 [[TMP3611]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3612:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3613:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3612]] release, align 8
+// CHECK-NEXT:    [[TMP3614:%.*]] = icmp ult i64 [[TMP3613]], [[TMP3612]]
+// CHECK-NEXT:    [[TMP3615:%.*]] = select i1 [[TMP3614]], i64 [[TMP3612]], i64 [[TMP3613]]
+// CHECK-NEXT:    store i64 [[TMP3615]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3616:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3617:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3616]] release, align 8
+// CHECK-NEXT:    [[TMP3618:%.*]] = icmp ugt i64 [[TMP3617]], [[TMP3616]]
+// CHECK-NEXT:    [[TMP3619:%.*]] = select i1 [[TMP3618]], i64 [[TMP3616]], i64 [[TMP3617]]
+// CHECK-NEXT:    store i64 [[TMP3619]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3620:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3621:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3622:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3620]], i64 [[TMP3621]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3623:%.*]] = extractvalue { i64, i1 } [[TMP3622]], 0
+// CHECK-NEXT:    [[TMP3624:%.*]] = extractvalue { i64, i1 } [[TMP3622]], 1
+// CHECK-NEXT:    [[TMP3625:%.*]] = select i1 [[TMP3624]], i64 [[TMP3620]], i64 [[TMP3623]]
+// CHECK-NEXT:    store i64 [[TMP3625]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3626:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3627:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3628:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3626]], i64 [[TMP3627]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3629:%.*]] = extractvalue { i64, i1 } [[TMP3628]], 0
+// CHECK-NEXT:    [[TMP3630:%.*]] = extractvalue { i64, i1 } [[TMP3628]], 1
+// CHECK-NEXT:    [[TMP3631:%.*]] = select i1 [[TMP3630]], i64 [[TMP3626]], i64 [[TMP3629]]
+// CHECK-NEXT:    store i64 [[TMP3631]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3632:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3633:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3634:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3632]], i64 [[TMP3633]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3635:%.*]] = extractvalue { i64, i1 } [[TMP3634]], 0
+// CHECK-NEXT:    [[TMP3636:%.*]] = extractvalue { i64, i1 } [[TMP3634]], 1
+// CHECK-NEXT:    br i1 [[TMP3636]], label [[ULX_ATOMIC_EXIT353:%.*]], label [[ULX_ATOMIC_CONT354:%.*]]
+// CHECK:       ulx.atomic.cont354:
+// CHECK-NEXT:    store i64 [[TMP3635]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT353]]
+// CHECK:       ulx.atomic.exit353:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3637:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3638:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3639:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3637]], i64 [[TMP3638]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3640:%.*]] = extractvalue { i64, i1 } [[TMP3639]], 0
+// CHECK-NEXT:    [[TMP3641:%.*]] = extractvalue { i64, i1 } [[TMP3639]], 1
+// CHECK-NEXT:    br i1 [[TMP3641]], label [[ULX_ATOMIC_EXIT355:%.*]], label [[ULX_ATOMIC_CONT356:%.*]]
+// CHECK:       ulx.atomic.cont356:
+// CHECK-NEXT:    store i64 [[TMP3640]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT355]]
+// CHECK:       ulx.atomic.exit355:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3642:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3643:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3644:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3642]], i64 [[TMP3643]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3645:%.*]] = extractvalue { i64, i1 } [[TMP3644]], 1
+// CHECK-NEXT:    [[TMP3646:%.*]] = zext i1 [[TMP3645]] to i64
+// CHECK-NEXT:    store i64 [[TMP3646]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3647:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3648:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3649:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3647]], i64 [[TMP3648]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3650:%.*]] = extractvalue { i64, i1 } [[TMP3649]], 1
+// CHECK-NEXT:    [[TMP3651:%.*]] = zext i1 [[TMP3650]] to i64
+// CHECK-NEXT:    store i64 [[TMP3651]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3652:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3653:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3654:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3652]], i64 [[TMP3653]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3655:%.*]] = extractvalue { i64, i1 } [[TMP3654]], 0
+// CHECK-NEXT:    [[TMP3656:%.*]] = extractvalue { i64, i1 } [[TMP3654]], 1
+// CHECK-NEXT:    br i1 [[TMP3656]], label [[ULX_ATOMIC_EXIT357:%.*]], label [[ULX_ATOMIC_CONT358:%.*]]
+// CHECK:       ulx.atomic.cont358:
+// CHECK-NEXT:    store i64 [[TMP3655]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT357]]
+// CHECK:       ulx.atomic.exit357:
+// CHECK-NEXT:    [[TMP3657:%.*]] = extractvalue { i64, i1 } [[TMP3654]], 1
+// CHECK-NEXT:    [[TMP3658:%.*]] = zext i1 [[TMP3657]] to i64
+// CHECK-NEXT:    store i64 [[TMP3658]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3659:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3660:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3661:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3659]], i64 [[TMP3660]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP3662:%.*]] = extractvalue { i64, i1 } [[TMP3661]], 0
+// CHECK-NEXT:    [[TMP3663:%.*]] = extractvalue { i64, i1 } [[TMP3661]], 1
+// CHECK-NEXT:    br i1 [[TMP3663]], label [[ULX_ATOMIC_EXIT359:%.*]], label [[ULX_ATOMIC_CONT360:%.*]]
+// CHECK:       ulx.atomic.cont360:
+// CHECK-NEXT:    store i64 [[TMP3662]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT359]]
+// CHECK:       ulx.atomic.exit359:
+// CHECK-NEXT:    [[TMP3664:%.*]] = extractvalue { i64, i1 } [[TMP3661]], 1
+// CHECK-NEXT:    [[TMP3665:%.*]] = zext i1 [[TMP3664]] to i64
+// CHECK-NEXT:    store i64 [[TMP3665]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3666:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3667:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3666]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP3667]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3668:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3669:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3668]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP3669]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3670:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3671:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3670]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP3671]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3672:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3673:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3672]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP3673]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3674:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3675:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3676:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3674]], i64 [[TMP3675]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3677:%.*]] = extractvalue { i64, i1 } [[TMP3676]], 0
+// CHECK-NEXT:    store i64 [[TMP3677]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3678:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3679:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3680:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3678]], i64 [[TMP3679]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3681:%.*]] = extractvalue { i64, i1 } [[TMP3680]], 0
+// CHECK-NEXT:    store i64 [[TMP3681]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3682:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3683:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3682]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3684:%.*]] = icmp ugt i64 [[TMP3683]], [[TMP3682]]
+// CHECK-NEXT:    [[TMP3685:%.*]] = select i1 [[TMP3684]], i64 [[TMP3682]], i64 [[TMP3683]]
+// CHECK-NEXT:    store i64 [[TMP3685]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3686:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3687:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3686]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3688:%.*]] = icmp ult i64 [[TMP3687]], [[TMP3686]]
+// CHECK-NEXT:    [[TMP3689:%.*]] = select i1 [[TMP3688]], i64 [[TMP3686]], i64 [[TMP3687]]
+// CHECK-NEXT:    store i64 [[TMP3689]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3690:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3691:%.*]] = atomicrmw umin i64* [[ULX]], i64 [[TMP3690]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3692:%.*]] = icmp ult i64 [[TMP3691]], [[TMP3690]]
+// CHECK-NEXT:    [[TMP3693:%.*]] = select i1 [[TMP3692]], i64 [[TMP3690]], i64 [[TMP3691]]
+// CHECK-NEXT:    store i64 [[TMP3693]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3694:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3695:%.*]] = atomicrmw umax i64* [[ULX]], i64 [[TMP3694]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP3696:%.*]] = icmp ugt i64 [[TMP3695]], [[TMP3694]]
+// CHECK-NEXT:    [[TMP3697:%.*]] = select i1 [[TMP3696]], i64 [[TMP3694]], i64 [[TMP3695]]
+// CHECK-NEXT:    store i64 [[TMP3697]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3698:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3699:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3700:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3698]], i64 [[TMP3699]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3701:%.*]] = extractvalue { i64, i1 } [[TMP3700]], 0
+// CHECK-NEXT:    [[TMP3702:%.*]] = extractvalue { i64, i1 } [[TMP3700]], 1
+// CHECK-NEXT:    [[TMP3703:%.*]] = select i1 [[TMP3702]], i64 [[TMP3698]], i64 [[TMP3701]]
+// CHECK-NEXT:    store i64 [[TMP3703]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3704:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3705:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3706:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3704]], i64 [[TMP3705]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3707:%.*]] = extractvalue { i64, i1 } [[TMP3706]], 0
+// CHECK-NEXT:    [[TMP3708:%.*]] = extractvalue { i64, i1 } [[TMP3706]], 1
+// CHECK-NEXT:    [[TMP3709:%.*]] = select i1 [[TMP3708]], i64 [[TMP3704]], i64 [[TMP3707]]
+// CHECK-NEXT:    store i64 [[TMP3709]], i64* [[ULV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3710:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3711:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3712:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3710]], i64 [[TMP3711]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3713:%.*]] = extractvalue { i64, i1 } [[TMP3712]], 0
+// CHECK-NEXT:    [[TMP3714:%.*]] = extractvalue { i64, i1 } [[TMP3712]], 1
+// CHECK-NEXT:    br i1 [[TMP3714]], label [[ULX_ATOMIC_EXIT361:%.*]], label [[ULX_ATOMIC_CONT362:%.*]]
+// CHECK:       ulx.atomic.cont362:
+// CHECK-NEXT:    store i64 [[TMP3713]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT361]]
+// CHECK:       ulx.atomic.exit361:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3715:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3716:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3717:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3715]], i64 [[TMP3716]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3718:%.*]] = extractvalue { i64, i1 } [[TMP3717]], 0
+// CHECK-NEXT:    [[TMP3719:%.*]] = extractvalue { i64, i1 } [[TMP3717]], 1
+// CHECK-NEXT:    br i1 [[TMP3719]], label [[ULX_ATOMIC_EXIT363:%.*]], label [[ULX_ATOMIC_CONT364:%.*]]
+// CHECK:       ulx.atomic.cont364:
+// CHECK-NEXT:    store i64 [[TMP3718]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT363]]
+// CHECK:       ulx.atomic.exit363:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3720:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3721:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3722:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3720]], i64 [[TMP3721]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3723:%.*]] = extractvalue { i64, i1 } [[TMP3722]], 1
+// CHECK-NEXT:    [[TMP3724:%.*]] = zext i1 [[TMP3723]] to i64
+// CHECK-NEXT:    store i64 [[TMP3724]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3725:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3726:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3727:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3725]], i64 [[TMP3726]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3728:%.*]] = extractvalue { i64, i1 } [[TMP3727]], 1
+// CHECK-NEXT:    [[TMP3729:%.*]] = zext i1 [[TMP3728]] to i64
+// CHECK-NEXT:    store i64 [[TMP3729]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3730:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3731:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3732:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3730]], i64 [[TMP3731]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3733:%.*]] = extractvalue { i64, i1 } [[TMP3732]], 0
+// CHECK-NEXT:    [[TMP3734:%.*]] = extractvalue { i64, i1 } [[TMP3732]], 1
+// CHECK-NEXT:    br i1 [[TMP3734]], label [[ULX_ATOMIC_EXIT365:%.*]], label [[ULX_ATOMIC_CONT366:%.*]]
+// CHECK:       ulx.atomic.cont366:
+// CHECK-NEXT:    store i64 [[TMP3733]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT365]]
+// CHECK:       ulx.atomic.exit365:
+// CHECK-NEXT:    [[TMP3735:%.*]] = extractvalue { i64, i1 } [[TMP3732]], 1
+// CHECK-NEXT:    [[TMP3736:%.*]] = zext i1 [[TMP3735]] to i64
+// CHECK-NEXT:    store i64 [[TMP3736]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3737:%.*]] = load i64, i64* [[ULE]], align 8
+// CHECK-NEXT:    [[TMP3738:%.*]] = load i64, i64* [[ULD]], align 8
+// CHECK-NEXT:    [[TMP3739:%.*]] = cmpxchg i64* [[ULX]], i64 [[TMP3737]], i64 [[TMP3738]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP3740:%.*]] = extractvalue { i64, i1 } [[TMP3739]], 0
+// CHECK-NEXT:    [[TMP3741:%.*]] = extractvalue { i64, i1 } [[TMP3739]], 1
+// CHECK-NEXT:    br i1 [[TMP3741]], label [[ULX_ATOMIC_EXIT367:%.*]], label [[ULX_ATOMIC_CONT368:%.*]]
+// CHECK:       ulx.atomic.cont368:
+// CHECK-NEXT:    store i64 [[TMP3740]], i64* [[ULV]], align 8
+// CHECK-NEXT:    br label [[ULX_ATOMIC_EXIT367]]
+// CHECK:       ulx.atomic.exit367:
+// CHECK-NEXT:    [[TMP3742:%.*]] = extractvalue { i64, i1 } [[TMP3739]], 1
+// CHECK-NEXT:    [[TMP3743:%.*]] = zext i1 [[TMP3742]] to i64
+// CHECK-NEXT:    store i64 [[TMP3743]], i64* [[ULR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3744:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3745:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3744]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3745]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3746:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3747:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3746]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3747]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3748:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3749:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3748]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3749]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3750:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3751:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3750]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3751]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3752:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3753:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3754:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3752]], i64 [[TMP3753]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3755:%.*]] = extractvalue { i64, i1 } [[TMP3754]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3755]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3756:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3757:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3758:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3756]], i64 [[TMP3757]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3759:%.*]] = extractvalue { i64, i1 } [[TMP3758]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3759]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3760:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3761:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3760]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3762:%.*]] = icmp ugt i64 [[TMP3761]], [[TMP3760]]
+// CHECK-NEXT:    [[TMP3763:%.*]] = select i1 [[TMP3762]], i64 [[TMP3760]], i64 [[TMP3761]]
+// CHECK-NEXT:    store volatile i64 [[TMP3763]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3764:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3765:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3764]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3766:%.*]] = icmp ult i64 [[TMP3765]], [[TMP3764]]
+// CHECK-NEXT:    [[TMP3767:%.*]] = select i1 [[TMP3766]], i64 [[TMP3764]], i64 [[TMP3765]]
+// CHECK-NEXT:    store volatile i64 [[TMP3767]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3768:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3769:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3768]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3770:%.*]] = icmp ult i64 [[TMP3769]], [[TMP3768]]
+// CHECK-NEXT:    [[TMP3771:%.*]] = select i1 [[TMP3770]], i64 [[TMP3768]], i64 [[TMP3769]]
+// CHECK-NEXT:    store volatile i64 [[TMP3771]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3772:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3773:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3772]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3774:%.*]] = icmp ugt i64 [[TMP3773]], [[TMP3772]]
+// CHECK-NEXT:    [[TMP3775:%.*]] = select i1 [[TMP3774]], i64 [[TMP3772]], i64 [[TMP3773]]
+// CHECK-NEXT:    store volatile i64 [[TMP3775]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3776:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3777:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3778:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3776]], i64 [[TMP3777]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3779:%.*]] = extractvalue { i64, i1 } [[TMP3778]], 0
+// CHECK-NEXT:    [[TMP3780:%.*]] = extractvalue { i64, i1 } [[TMP3778]], 1
+// CHECK-NEXT:    [[TMP3781:%.*]] = select i1 [[TMP3780]], i64 [[TMP3776]], i64 [[TMP3779]]
+// CHECK-NEXT:    store volatile i64 [[TMP3781]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3782:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3783:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3784:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3782]], i64 [[TMP3783]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3785:%.*]] = extractvalue { i64, i1 } [[TMP3784]], 0
+// CHECK-NEXT:    [[TMP3786:%.*]] = extractvalue { i64, i1 } [[TMP3784]], 1
+// CHECK-NEXT:    [[TMP3787:%.*]] = select i1 [[TMP3786]], i64 [[TMP3782]], i64 [[TMP3785]]
+// CHECK-NEXT:    store volatile i64 [[TMP3787]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3788:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3789:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3790:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3788]], i64 [[TMP3789]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3791:%.*]] = extractvalue { i64, i1 } [[TMP3790]], 0
+// CHECK-NEXT:    [[TMP3792:%.*]] = extractvalue { i64, i1 } [[TMP3790]], 1
+// CHECK-NEXT:    br i1 [[TMP3792]], label [[LLX_ATOMIC_EXIT:%.*]], label [[LLX_ATOMIC_CONT:%.*]]
+// CHECK:       llx.atomic.cont:
+// CHECK-NEXT:    store i64 [[TMP3791]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT]]
+// CHECK:       llx.atomic.exit:
+// CHECK-NEXT:    [[TMP3793:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3794:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3795:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3793]], i64 [[TMP3794]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3796:%.*]] = extractvalue { i64, i1 } [[TMP3795]], 0
+// CHECK-NEXT:    [[TMP3797:%.*]] = extractvalue { i64, i1 } [[TMP3795]], 1
+// CHECK-NEXT:    br i1 [[TMP3797]], label [[LLX_ATOMIC_EXIT369:%.*]], label [[LLX_ATOMIC_CONT370:%.*]]
+// CHECK:       llx.atomic.cont370:
+// CHECK-NEXT:    store i64 [[TMP3796]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT369]]
+// CHECK:       llx.atomic.exit369:
+// CHECK-NEXT:    [[TMP3798:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3799:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3800:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3798]], i64 [[TMP3799]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3801:%.*]] = extractvalue { i64, i1 } [[TMP3800]], 1
+// CHECK-NEXT:    [[TMP3802:%.*]] = zext i1 [[TMP3801]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3802]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3803:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3804:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3805:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3803]], i64 [[TMP3804]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3806:%.*]] = extractvalue { i64, i1 } [[TMP3805]], 1
+// CHECK-NEXT:    [[TMP3807:%.*]] = zext i1 [[TMP3806]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3807]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3808:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3809:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3810:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3808]], i64 [[TMP3809]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3811:%.*]] = extractvalue { i64, i1 } [[TMP3810]], 0
+// CHECK-NEXT:    [[TMP3812:%.*]] = extractvalue { i64, i1 } [[TMP3810]], 1
+// CHECK-NEXT:    br i1 [[TMP3812]], label [[LLX_ATOMIC_EXIT371:%.*]], label [[LLX_ATOMIC_CONT372:%.*]]
+// CHECK:       llx.atomic.cont372:
+// CHECK-NEXT:    store i64 [[TMP3811]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT371]]
+// CHECK:       llx.atomic.exit371:
+// CHECK-NEXT:    [[TMP3813:%.*]] = extractvalue { i64, i1 } [[TMP3810]], 1
+// CHECK-NEXT:    [[TMP3814:%.*]] = zext i1 [[TMP3813]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3814]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3815:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3816:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3817:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3815]], i64 [[TMP3816]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3818:%.*]] = extractvalue { i64, i1 } [[TMP3817]], 0
+// CHECK-NEXT:    [[TMP3819:%.*]] = extractvalue { i64, i1 } [[TMP3817]], 1
+// CHECK-NEXT:    br i1 [[TMP3819]], label [[LLX_ATOMIC_EXIT373:%.*]], label [[LLX_ATOMIC_CONT374:%.*]]
+// CHECK:       llx.atomic.cont374:
+// CHECK-NEXT:    store i64 [[TMP3818]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT373]]
+// CHECK:       llx.atomic.exit373:
+// CHECK-NEXT:    [[TMP3820:%.*]] = extractvalue { i64, i1 } [[TMP3817]], 1
+// CHECK-NEXT:    [[TMP3821:%.*]] = zext i1 [[TMP3820]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3821]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3822:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3823:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3822]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3823]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3824:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3825:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3824]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3825]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3826:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3827:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3826]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3827]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3828:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3829:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3828]] acq_rel, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3829]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3830:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3831:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3832:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3830]], i64 [[TMP3831]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3833:%.*]] = extractvalue { i64, i1 } [[TMP3832]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3833]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3834:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3835:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3836:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3834]], i64 [[TMP3835]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3837:%.*]] = extractvalue { i64, i1 } [[TMP3836]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3837]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3838:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3839:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3838]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3840:%.*]] = icmp ugt i64 [[TMP3839]], [[TMP3838]]
+// CHECK-NEXT:    [[TMP3841:%.*]] = select i1 [[TMP3840]], i64 [[TMP3838]], i64 [[TMP3839]]
+// CHECK-NEXT:    store volatile i64 [[TMP3841]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3842:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3843:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3842]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3844:%.*]] = icmp ult i64 [[TMP3843]], [[TMP3842]]
+// CHECK-NEXT:    [[TMP3845:%.*]] = select i1 [[TMP3844]], i64 [[TMP3842]], i64 [[TMP3843]]
+// CHECK-NEXT:    store volatile i64 [[TMP3845]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3846:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3847:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3846]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3848:%.*]] = icmp ult i64 [[TMP3847]], [[TMP3846]]
+// CHECK-NEXT:    [[TMP3849:%.*]] = select i1 [[TMP3848]], i64 [[TMP3846]], i64 [[TMP3847]]
+// CHECK-NEXT:    store volatile i64 [[TMP3849]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3850:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3851:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3850]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP3852:%.*]] = icmp ugt i64 [[TMP3851]], [[TMP3850]]
+// CHECK-NEXT:    [[TMP3853:%.*]] = select i1 [[TMP3852]], i64 [[TMP3850]], i64 [[TMP3851]]
+// CHECK-NEXT:    store volatile i64 [[TMP3853]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3854:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3855:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3856:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3854]], i64 [[TMP3855]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3857:%.*]] = extractvalue { i64, i1 } [[TMP3856]], 0
+// CHECK-NEXT:    [[TMP3858:%.*]] = extractvalue { i64, i1 } [[TMP3856]], 1
+// CHECK-NEXT:    [[TMP3859:%.*]] = select i1 [[TMP3858]], i64 [[TMP3854]], i64 [[TMP3857]]
+// CHECK-NEXT:    store volatile i64 [[TMP3859]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3860:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3861:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3862:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3860]], i64 [[TMP3861]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3863:%.*]] = extractvalue { i64, i1 } [[TMP3862]], 0
+// CHECK-NEXT:    [[TMP3864:%.*]] = extractvalue { i64, i1 } [[TMP3862]], 1
+// CHECK-NEXT:    [[TMP3865:%.*]] = select i1 [[TMP3864]], i64 [[TMP3860]], i64 [[TMP3863]]
+// CHECK-NEXT:    store volatile i64 [[TMP3865]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3866:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3867:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3868:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3866]], i64 [[TMP3867]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3869:%.*]] = extractvalue { i64, i1 } [[TMP3868]], 0
+// CHECK-NEXT:    [[TMP3870:%.*]] = extractvalue { i64, i1 } [[TMP3868]], 1
+// CHECK-NEXT:    br i1 [[TMP3870]], label [[LLX_ATOMIC_EXIT375:%.*]], label [[LLX_ATOMIC_CONT376:%.*]]
+// CHECK:       llx.atomic.cont376:
+// CHECK-NEXT:    store i64 [[TMP3869]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT375]]
+// CHECK:       llx.atomic.exit375:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3871:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3872:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3873:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3871]], i64 [[TMP3872]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3874:%.*]] = extractvalue { i64, i1 } [[TMP3873]], 0
+// CHECK-NEXT:    [[TMP3875:%.*]] = extractvalue { i64, i1 } [[TMP3873]], 1
+// CHECK-NEXT:    br i1 [[TMP3875]], label [[LLX_ATOMIC_EXIT377:%.*]], label [[LLX_ATOMIC_CONT378:%.*]]
+// CHECK:       llx.atomic.cont378:
+// CHECK-NEXT:    store i64 [[TMP3874]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT377]]
+// CHECK:       llx.atomic.exit377:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3876:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3877:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3878:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3876]], i64 [[TMP3877]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3879:%.*]] = extractvalue { i64, i1 } [[TMP3878]], 1
+// CHECK-NEXT:    [[TMP3880:%.*]] = zext i1 [[TMP3879]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3880]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3881:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3882:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3883:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3881]], i64 [[TMP3882]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3884:%.*]] = extractvalue { i64, i1 } [[TMP3883]], 1
+// CHECK-NEXT:    [[TMP3885:%.*]] = zext i1 [[TMP3884]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3885]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3886:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3887:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3888:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3886]], i64 [[TMP3887]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3889:%.*]] = extractvalue { i64, i1 } [[TMP3888]], 0
+// CHECK-NEXT:    [[TMP3890:%.*]] = extractvalue { i64, i1 } [[TMP3888]], 1
+// CHECK-NEXT:    br i1 [[TMP3890]], label [[LLX_ATOMIC_EXIT379:%.*]], label [[LLX_ATOMIC_CONT380:%.*]]
+// CHECK:       llx.atomic.cont380:
+// CHECK-NEXT:    store i64 [[TMP3889]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT379]]
+// CHECK:       llx.atomic.exit379:
+// CHECK-NEXT:    [[TMP3891:%.*]] = extractvalue { i64, i1 } [[TMP3888]], 1
+// CHECK-NEXT:    [[TMP3892:%.*]] = zext i1 [[TMP3891]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3892]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3893:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3894:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3895:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3893]], i64 [[TMP3894]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP3896:%.*]] = extractvalue { i64, i1 } [[TMP3895]], 0
+// CHECK-NEXT:    [[TMP3897:%.*]] = extractvalue { i64, i1 } [[TMP3895]], 1
+// CHECK-NEXT:    br i1 [[TMP3897]], label [[LLX_ATOMIC_EXIT381:%.*]], label [[LLX_ATOMIC_CONT382:%.*]]
+// CHECK:       llx.atomic.cont382:
+// CHECK-NEXT:    store i64 [[TMP3896]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT381]]
+// CHECK:       llx.atomic.exit381:
+// CHECK-NEXT:    [[TMP3898:%.*]] = extractvalue { i64, i1 } [[TMP3895]], 1
+// CHECK-NEXT:    [[TMP3899:%.*]] = zext i1 [[TMP3898]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3899]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP3900:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3901:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3900]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3901]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3902:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3903:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3902]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3903]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3904:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3905:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3904]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3905]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3906:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3907:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3906]] acquire, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3907]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3908:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3909:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3910:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3908]], i64 [[TMP3909]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3911:%.*]] = extractvalue { i64, i1 } [[TMP3910]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3911]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3912:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3913:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3914:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3912]], i64 [[TMP3913]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3915:%.*]] = extractvalue { i64, i1 } [[TMP3914]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3915]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3916:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3917:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3916]] acquire, align 8
+// CHECK-NEXT:    [[TMP3918:%.*]] = icmp ugt i64 [[TMP3917]], [[TMP3916]]
+// CHECK-NEXT:    [[TMP3919:%.*]] = select i1 [[TMP3918]], i64 [[TMP3916]], i64 [[TMP3917]]
+// CHECK-NEXT:    store volatile i64 [[TMP3919]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3920:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3921:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3920]] acquire, align 8
+// CHECK-NEXT:    [[TMP3922:%.*]] = icmp ult i64 [[TMP3921]], [[TMP3920]]
+// CHECK-NEXT:    [[TMP3923:%.*]] = select i1 [[TMP3922]], i64 [[TMP3920]], i64 [[TMP3921]]
+// CHECK-NEXT:    store volatile i64 [[TMP3923]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3924:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3925:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3924]] acquire, align 8
+// CHECK-NEXT:    [[TMP3926:%.*]] = icmp ult i64 [[TMP3925]], [[TMP3924]]
+// CHECK-NEXT:    [[TMP3927:%.*]] = select i1 [[TMP3926]], i64 [[TMP3924]], i64 [[TMP3925]]
+// CHECK-NEXT:    store volatile i64 [[TMP3927]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3928:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3929:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3928]] acquire, align 8
+// CHECK-NEXT:    [[TMP3930:%.*]] = icmp ugt i64 [[TMP3929]], [[TMP3928]]
+// CHECK-NEXT:    [[TMP3931:%.*]] = select i1 [[TMP3930]], i64 [[TMP3928]], i64 [[TMP3929]]
+// CHECK-NEXT:    store volatile i64 [[TMP3931]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3932:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3933:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3934:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3932]], i64 [[TMP3933]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3935:%.*]] = extractvalue { i64, i1 } [[TMP3934]], 0
+// CHECK-NEXT:    [[TMP3936:%.*]] = extractvalue { i64, i1 } [[TMP3934]], 1
+// CHECK-NEXT:    [[TMP3937:%.*]] = select i1 [[TMP3936]], i64 [[TMP3932]], i64 [[TMP3935]]
+// CHECK-NEXT:    store volatile i64 [[TMP3937]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3938:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3939:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3940:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3938]], i64 [[TMP3939]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3941:%.*]] = extractvalue { i64, i1 } [[TMP3940]], 0
+// CHECK-NEXT:    [[TMP3942:%.*]] = extractvalue { i64, i1 } [[TMP3940]], 1
+// CHECK-NEXT:    [[TMP3943:%.*]] = select i1 [[TMP3942]], i64 [[TMP3938]], i64 [[TMP3941]]
+// CHECK-NEXT:    store volatile i64 [[TMP3943]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3944:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3945:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3946:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3944]], i64 [[TMP3945]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3947:%.*]] = extractvalue { i64, i1 } [[TMP3946]], 0
+// CHECK-NEXT:    [[TMP3948:%.*]] = extractvalue { i64, i1 } [[TMP3946]], 1
+// CHECK-NEXT:    br i1 [[TMP3948]], label [[LLX_ATOMIC_EXIT383:%.*]], label [[LLX_ATOMIC_CONT384:%.*]]
+// CHECK:       llx.atomic.cont384:
+// CHECK-NEXT:    store i64 [[TMP3947]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT383]]
+// CHECK:       llx.atomic.exit383:
+// CHECK-NEXT:    [[TMP3949:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3950:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3951:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3949]], i64 [[TMP3950]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3952:%.*]] = extractvalue { i64, i1 } [[TMP3951]], 0
+// CHECK-NEXT:    [[TMP3953:%.*]] = extractvalue { i64, i1 } [[TMP3951]], 1
+// CHECK-NEXT:    br i1 [[TMP3953]], label [[LLX_ATOMIC_EXIT385:%.*]], label [[LLX_ATOMIC_CONT386:%.*]]
+// CHECK:       llx.atomic.cont386:
+// CHECK-NEXT:    store i64 [[TMP3952]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT385]]
+// CHECK:       llx.atomic.exit385:
+// CHECK-NEXT:    [[TMP3954:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3955:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3956:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3954]], i64 [[TMP3955]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3957:%.*]] = extractvalue { i64, i1 } [[TMP3956]], 1
+// CHECK-NEXT:    [[TMP3958:%.*]] = zext i1 [[TMP3957]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3958]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3959:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3960:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3961:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3959]], i64 [[TMP3960]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3962:%.*]] = extractvalue { i64, i1 } [[TMP3961]], 1
+// CHECK-NEXT:    [[TMP3963:%.*]] = zext i1 [[TMP3962]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3963]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3964:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3965:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3966:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3964]], i64 [[TMP3965]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3967:%.*]] = extractvalue { i64, i1 } [[TMP3966]], 0
+// CHECK-NEXT:    [[TMP3968:%.*]] = extractvalue { i64, i1 } [[TMP3966]], 1
+// CHECK-NEXT:    br i1 [[TMP3968]], label [[LLX_ATOMIC_EXIT387:%.*]], label [[LLX_ATOMIC_CONT388:%.*]]
+// CHECK:       llx.atomic.cont388:
+// CHECK-NEXT:    store i64 [[TMP3967]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT387]]
+// CHECK:       llx.atomic.exit387:
+// CHECK-NEXT:    [[TMP3969:%.*]] = extractvalue { i64, i1 } [[TMP3966]], 1
+// CHECK-NEXT:    [[TMP3970:%.*]] = zext i1 [[TMP3969]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3970]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3971:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3972:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3973:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3971]], i64 [[TMP3972]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP3974:%.*]] = extractvalue { i64, i1 } [[TMP3973]], 0
+// CHECK-NEXT:    [[TMP3975:%.*]] = extractvalue { i64, i1 } [[TMP3973]], 1
+// CHECK-NEXT:    br i1 [[TMP3975]], label [[LLX_ATOMIC_EXIT389:%.*]], label [[LLX_ATOMIC_CONT390:%.*]]
+// CHECK:       llx.atomic.cont390:
+// CHECK-NEXT:    store i64 [[TMP3974]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT389]]
+// CHECK:       llx.atomic.exit389:
+// CHECK-NEXT:    [[TMP3976:%.*]] = extractvalue { i64, i1 } [[TMP3973]], 1
+// CHECK-NEXT:    [[TMP3977:%.*]] = zext i1 [[TMP3976]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP3977]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP3978:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3979:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3978]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3979]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3980:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3981:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3980]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3981]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3982:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3983:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3982]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3983]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3984:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3985:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3984]] monotonic, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP3985]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3986:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3987:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3988:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3986]], i64 [[TMP3987]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3989:%.*]] = extractvalue { i64, i1 } [[TMP3988]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3989]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3990:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3991:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP3992:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP3990]], i64 [[TMP3991]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP3993:%.*]] = extractvalue { i64, i1 } [[TMP3992]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP3993]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3994:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3995:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP3994]] monotonic, align 8
+// CHECK-NEXT:    [[TMP3996:%.*]] = icmp ugt i64 [[TMP3995]], [[TMP3994]]
+// CHECK-NEXT:    [[TMP3997:%.*]] = select i1 [[TMP3996]], i64 [[TMP3994]], i64 [[TMP3995]]
+// CHECK-NEXT:    store volatile i64 [[TMP3997]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP3998:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP3999:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP3998]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4000:%.*]] = icmp ult i64 [[TMP3999]], [[TMP3998]]
+// CHECK-NEXT:    [[TMP4001:%.*]] = select i1 [[TMP4000]], i64 [[TMP3998]], i64 [[TMP3999]]
+// CHECK-NEXT:    store volatile i64 [[TMP4001]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP4002:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4003:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4002]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4004:%.*]] = icmp ult i64 [[TMP4003]], [[TMP4002]]
+// CHECK-NEXT:    [[TMP4005:%.*]] = select i1 [[TMP4004]], i64 [[TMP4002]], i64 [[TMP4003]]
+// CHECK-NEXT:    store volatile i64 [[TMP4005]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP4006:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4007:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4006]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4008:%.*]] = icmp ugt i64 [[TMP4007]], [[TMP4006]]
+// CHECK-NEXT:    [[TMP4009:%.*]] = select i1 [[TMP4008]], i64 [[TMP4006]], i64 [[TMP4007]]
+// CHECK-NEXT:    store volatile i64 [[TMP4009]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP4010:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4011:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4012:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4010]], i64 [[TMP4011]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4013:%.*]] = extractvalue { i64, i1 } [[TMP4012]], 0
+// CHECK-NEXT:    [[TMP4014:%.*]] = extractvalue { i64, i1 } [[TMP4012]], 1
+// CHECK-NEXT:    [[TMP4015:%.*]] = select i1 [[TMP4014]], i64 [[TMP4010]], i64 [[TMP4013]]
+// CHECK-NEXT:    store volatile i64 [[TMP4015]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP4016:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4017:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4018:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4016]], i64 [[TMP4017]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4019:%.*]] = extractvalue { i64, i1 } [[TMP4018]], 0
+// CHECK-NEXT:    [[TMP4020:%.*]] = extractvalue { i64, i1 } [[TMP4018]], 1
+// CHECK-NEXT:    [[TMP4021:%.*]] = select i1 [[TMP4020]], i64 [[TMP4016]], i64 [[TMP4019]]
+// CHECK-NEXT:    store volatile i64 [[TMP4021]], i64* [[LLV]], align 8
+// CHECK-NEXT:    [[TMP4022:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4023:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4024:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4022]], i64 [[TMP4023]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4025:%.*]] = extractvalue { i64, i1 } [[TMP4024]], 0
+// CHECK-NEXT:    [[TMP4026:%.*]] = extractvalue { i64, i1 } [[TMP4024]], 1
+// CHECK-NEXT:    br i1 [[TMP4026]], label [[LLX_ATOMIC_EXIT391:%.*]], label [[LLX_ATOMIC_CONT392:%.*]]
+// CHECK:       llx.atomic.cont392:
+// CHECK-NEXT:    store i64 [[TMP4025]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT391]]
+// CHECK:       llx.atomic.exit391:
+// CHECK-NEXT:    [[TMP4027:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4028:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4029:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4027]], i64 [[TMP4028]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4030:%.*]] = extractvalue { i64, i1 } [[TMP4029]], 0
+// CHECK-NEXT:    [[TMP4031:%.*]] = extractvalue { i64, i1 } [[TMP4029]], 1
+// CHECK-NEXT:    br i1 [[TMP4031]], label [[LLX_ATOMIC_EXIT393:%.*]], label [[LLX_ATOMIC_CONT394:%.*]]
+// CHECK:       llx.atomic.cont394:
+// CHECK-NEXT:    store i64 [[TMP4030]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT393]]
+// CHECK:       llx.atomic.exit393:
+// CHECK-NEXT:    [[TMP4032:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4033:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4034:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4032]], i64 [[TMP4033]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4035:%.*]] = extractvalue { i64, i1 } [[TMP4034]], 1
+// CHECK-NEXT:    [[TMP4036:%.*]] = zext i1 [[TMP4035]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4036]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP4037:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4038:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4039:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4037]], i64 [[TMP4038]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4040:%.*]] = extractvalue { i64, i1 } [[TMP4039]], 1
+// CHECK-NEXT:    [[TMP4041:%.*]] = zext i1 [[TMP4040]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4041]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP4042:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4043:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4044:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4042]], i64 [[TMP4043]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4045:%.*]] = extractvalue { i64, i1 } [[TMP4044]], 0
+// CHECK-NEXT:    [[TMP4046:%.*]] = extractvalue { i64, i1 } [[TMP4044]], 1
+// CHECK-NEXT:    br i1 [[TMP4046]], label [[LLX_ATOMIC_EXIT395:%.*]], label [[LLX_ATOMIC_CONT396:%.*]]
+// CHECK:       llx.atomic.cont396:
+// CHECK-NEXT:    store i64 [[TMP4045]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT395]]
+// CHECK:       llx.atomic.exit395:
+// CHECK-NEXT:    [[TMP4047:%.*]] = extractvalue { i64, i1 } [[TMP4044]], 1
+// CHECK-NEXT:    [[TMP4048:%.*]] = zext i1 [[TMP4047]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4048]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP4049:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4050:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4051:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4049]], i64 [[TMP4050]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4052:%.*]] = extractvalue { i64, i1 } [[TMP4051]], 0
+// CHECK-NEXT:    [[TMP4053:%.*]] = extractvalue { i64, i1 } [[TMP4051]], 1
+// CHECK-NEXT:    br i1 [[TMP4053]], label [[LLX_ATOMIC_EXIT397:%.*]], label [[LLX_ATOMIC_CONT398:%.*]]
+// CHECK:       llx.atomic.cont398:
+// CHECK-NEXT:    store i64 [[TMP4052]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT397]]
+// CHECK:       llx.atomic.exit397:
+// CHECK-NEXT:    [[TMP4054:%.*]] = extractvalue { i64, i1 } [[TMP4051]], 1
+// CHECK-NEXT:    [[TMP4055:%.*]] = zext i1 [[TMP4054]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4055]], i64* [[LLR]], align 8
+// CHECK-NEXT:    [[TMP4056:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4057:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4056]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4057]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4058:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4059:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4058]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4059]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4060:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4061:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4060]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4061]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4062:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4063:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4062]] release, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4063]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4064:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4065:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4066:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4064]], i64 [[TMP4065]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4067:%.*]] = extractvalue { i64, i1 } [[TMP4066]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP4067]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4068:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4069:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4070:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4068]], i64 [[TMP4069]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4071:%.*]] = extractvalue { i64, i1 } [[TMP4070]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP4071]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4072:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4073:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4072]] release, align 8
+// CHECK-NEXT:    [[TMP4074:%.*]] = icmp ugt i64 [[TMP4073]], [[TMP4072]]
+// CHECK-NEXT:    [[TMP4075:%.*]] = select i1 [[TMP4074]], i64 [[TMP4072]], i64 [[TMP4073]]
+// CHECK-NEXT:    store volatile i64 [[TMP4075]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4076:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4077:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4076]] release, align 8
+// CHECK-NEXT:    [[TMP4078:%.*]] = icmp ult i64 [[TMP4077]], [[TMP4076]]
+// CHECK-NEXT:    [[TMP4079:%.*]] = select i1 [[TMP4078]], i64 [[TMP4076]], i64 [[TMP4077]]
+// CHECK-NEXT:    store volatile i64 [[TMP4079]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4080:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4081:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4080]] release, align 8
+// CHECK-NEXT:    [[TMP4082:%.*]] = icmp ult i64 [[TMP4081]], [[TMP4080]]
+// CHECK-NEXT:    [[TMP4083:%.*]] = select i1 [[TMP4082]], i64 [[TMP4080]], i64 [[TMP4081]]
+// CHECK-NEXT:    store volatile i64 [[TMP4083]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4084:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4085:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4084]] release, align 8
+// CHECK-NEXT:    [[TMP4086:%.*]] = icmp ugt i64 [[TMP4085]], [[TMP4084]]
+// CHECK-NEXT:    [[TMP4087:%.*]] = select i1 [[TMP4086]], i64 [[TMP4084]], i64 [[TMP4085]]
+// CHECK-NEXT:    store volatile i64 [[TMP4087]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4088:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4089:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4090:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4088]], i64 [[TMP4089]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4091:%.*]] = extractvalue { i64, i1 } [[TMP4090]], 0
+// CHECK-NEXT:    [[TMP4092:%.*]] = extractvalue { i64, i1 } [[TMP4090]], 1
+// CHECK-NEXT:    [[TMP4093:%.*]] = select i1 [[TMP4092]], i64 [[TMP4088]], i64 [[TMP4091]]
+// CHECK-NEXT:    store volatile i64 [[TMP4093]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4094:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4095:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4096:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4094]], i64 [[TMP4095]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4097:%.*]] = extractvalue { i64, i1 } [[TMP4096]], 0
+// CHECK-NEXT:    [[TMP4098:%.*]] = extractvalue { i64, i1 } [[TMP4096]], 1
+// CHECK-NEXT:    [[TMP4099:%.*]] = select i1 [[TMP4098]], i64 [[TMP4094]], i64 [[TMP4097]]
+// CHECK-NEXT:    store volatile i64 [[TMP4099]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4100:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4101:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4102:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4100]], i64 [[TMP4101]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4103:%.*]] = extractvalue { i64, i1 } [[TMP4102]], 0
+// CHECK-NEXT:    [[TMP4104:%.*]] = extractvalue { i64, i1 } [[TMP4102]], 1
+// CHECK-NEXT:    br i1 [[TMP4104]], label [[LLX_ATOMIC_EXIT399:%.*]], label [[LLX_ATOMIC_CONT400:%.*]]
+// CHECK:       llx.atomic.cont400:
+// CHECK-NEXT:    store i64 [[TMP4103]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT399]]
+// CHECK:       llx.atomic.exit399:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4105:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4106:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4107:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4105]], i64 [[TMP4106]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4108:%.*]] = extractvalue { i64, i1 } [[TMP4107]], 0
+// CHECK-NEXT:    [[TMP4109:%.*]] = extractvalue { i64, i1 } [[TMP4107]], 1
+// CHECK-NEXT:    br i1 [[TMP4109]], label [[LLX_ATOMIC_EXIT401:%.*]], label [[LLX_ATOMIC_CONT402:%.*]]
+// CHECK:       llx.atomic.cont402:
+// CHECK-NEXT:    store i64 [[TMP4108]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT401]]
+// CHECK:       llx.atomic.exit401:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4110:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4111:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4112:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4110]], i64 [[TMP4111]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4113:%.*]] = extractvalue { i64, i1 } [[TMP4112]], 1
+// CHECK-NEXT:    [[TMP4114:%.*]] = zext i1 [[TMP4113]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4114]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4115:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4116:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4117:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4115]], i64 [[TMP4116]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4118:%.*]] = extractvalue { i64, i1 } [[TMP4117]], 1
+// CHECK-NEXT:    [[TMP4119:%.*]] = zext i1 [[TMP4118]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4119]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4120:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4121:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4122:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4120]], i64 [[TMP4121]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4123:%.*]] = extractvalue { i64, i1 } [[TMP4122]], 0
+// CHECK-NEXT:    [[TMP4124:%.*]] = extractvalue { i64, i1 } [[TMP4122]], 1
+// CHECK-NEXT:    br i1 [[TMP4124]], label [[LLX_ATOMIC_EXIT403:%.*]], label [[LLX_ATOMIC_CONT404:%.*]]
+// CHECK:       llx.atomic.cont404:
+// CHECK-NEXT:    store i64 [[TMP4123]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT403]]
+// CHECK:       llx.atomic.exit403:
+// CHECK-NEXT:    [[TMP4125:%.*]] = extractvalue { i64, i1 } [[TMP4122]], 1
+// CHECK-NEXT:    [[TMP4126:%.*]] = zext i1 [[TMP4125]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4126]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4127:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4128:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4129:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4127]], i64 [[TMP4128]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4130:%.*]] = extractvalue { i64, i1 } [[TMP4129]], 0
+// CHECK-NEXT:    [[TMP4131:%.*]] = extractvalue { i64, i1 } [[TMP4129]], 1
+// CHECK-NEXT:    br i1 [[TMP4131]], label [[LLX_ATOMIC_EXIT405:%.*]], label [[LLX_ATOMIC_CONT406:%.*]]
+// CHECK:       llx.atomic.cont406:
+// CHECK-NEXT:    store i64 [[TMP4130]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT405]]
+// CHECK:       llx.atomic.exit405:
+// CHECK-NEXT:    [[TMP4132:%.*]] = extractvalue { i64, i1 } [[TMP4129]], 1
+// CHECK-NEXT:    [[TMP4133:%.*]] = zext i1 [[TMP4132]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4133]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4134:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4135:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4134]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4135]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4136:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4137:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4136]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4137]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4138:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4139:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4138]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4139]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4140:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4141:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4140]] seq_cst, align 8
+// CHECK-NEXT:    store volatile i64 [[TMP4141]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4142:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4143:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4144:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4142]], i64 [[TMP4143]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4145:%.*]] = extractvalue { i64, i1 } [[TMP4144]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP4145]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4146:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4147:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4148:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4146]], i64 [[TMP4147]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4149:%.*]] = extractvalue { i64, i1 } [[TMP4148]], 0
+// CHECK-NEXT:    store volatile i64 [[TMP4149]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4150:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4151:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4150]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4152:%.*]] = icmp ugt i64 [[TMP4151]], [[TMP4150]]
+// CHECK-NEXT:    [[TMP4153:%.*]] = select i1 [[TMP4152]], i64 [[TMP4150]], i64 [[TMP4151]]
+// CHECK-NEXT:    store volatile i64 [[TMP4153]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4154:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4155:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4154]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4156:%.*]] = icmp ult i64 [[TMP4155]], [[TMP4154]]
+// CHECK-NEXT:    [[TMP4157:%.*]] = select i1 [[TMP4156]], i64 [[TMP4154]], i64 [[TMP4155]]
+// CHECK-NEXT:    store volatile i64 [[TMP4157]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4158:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4159:%.*]] = atomicrmw umin i64* [[LLX]], i64 [[TMP4158]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4160:%.*]] = icmp ult i64 [[TMP4159]], [[TMP4158]]
+// CHECK-NEXT:    [[TMP4161:%.*]] = select i1 [[TMP4160]], i64 [[TMP4158]], i64 [[TMP4159]]
+// CHECK-NEXT:    store volatile i64 [[TMP4161]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4162:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4163:%.*]] = atomicrmw umax i64* [[LLX]], i64 [[TMP4162]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4164:%.*]] = icmp ugt i64 [[TMP4163]], [[TMP4162]]
+// CHECK-NEXT:    [[TMP4165:%.*]] = select i1 [[TMP4164]], i64 [[TMP4162]], i64 [[TMP4163]]
+// CHECK-NEXT:    store volatile i64 [[TMP4165]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4166:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4167:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4168:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4166]], i64 [[TMP4167]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4169:%.*]] = extractvalue { i64, i1 } [[TMP4168]], 0
+// CHECK-NEXT:    [[TMP4170:%.*]] = extractvalue { i64, i1 } [[TMP4168]], 1
+// CHECK-NEXT:    [[TMP4171:%.*]] = select i1 [[TMP4170]], i64 [[TMP4166]], i64 [[TMP4169]]
+// CHECK-NEXT:    store volatile i64 [[TMP4171]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4172:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4173:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4174:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4172]], i64 [[TMP4173]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4175:%.*]] = extractvalue { i64, i1 } [[TMP4174]], 0
+// CHECK-NEXT:    [[TMP4176:%.*]] = extractvalue { i64, i1 } [[TMP4174]], 1
+// CHECK-NEXT:    [[TMP4177:%.*]] = select i1 [[TMP4176]], i64 [[TMP4172]], i64 [[TMP4175]]
+// CHECK-NEXT:    store volatile i64 [[TMP4177]], i64* [[LLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4178:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4179:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4180:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4178]], i64 [[TMP4179]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4181:%.*]] = extractvalue { i64, i1 } [[TMP4180]], 0
+// CHECK-NEXT:    [[TMP4182:%.*]] = extractvalue { i64, i1 } [[TMP4180]], 1
+// CHECK-NEXT:    br i1 [[TMP4182]], label [[LLX_ATOMIC_EXIT407:%.*]], label [[LLX_ATOMIC_CONT408:%.*]]
+// CHECK:       llx.atomic.cont408:
+// CHECK-NEXT:    store i64 [[TMP4181]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT407]]
+// CHECK:       llx.atomic.exit407:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4183:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4184:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4185:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4183]], i64 [[TMP4184]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4186:%.*]] = extractvalue { i64, i1 } [[TMP4185]], 0
+// CHECK-NEXT:    [[TMP4187:%.*]] = extractvalue { i64, i1 } [[TMP4185]], 1
+// CHECK-NEXT:    br i1 [[TMP4187]], label [[LLX_ATOMIC_EXIT409:%.*]], label [[LLX_ATOMIC_CONT410:%.*]]
+// CHECK:       llx.atomic.cont410:
+// CHECK-NEXT:    store i64 [[TMP4186]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT409]]
+// CHECK:       llx.atomic.exit409:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4188:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4189:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4190:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4188]], i64 [[TMP4189]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4191:%.*]] = extractvalue { i64, i1 } [[TMP4190]], 1
+// CHECK-NEXT:    [[TMP4192:%.*]] = zext i1 [[TMP4191]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4192]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4193:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4194:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4195:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4193]], i64 [[TMP4194]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4196:%.*]] = extractvalue { i64, i1 } [[TMP4195]], 1
+// CHECK-NEXT:    [[TMP4197:%.*]] = zext i1 [[TMP4196]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4197]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4198:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4199:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4200:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4198]], i64 [[TMP4199]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4201:%.*]] = extractvalue { i64, i1 } [[TMP4200]], 0
+// CHECK-NEXT:    [[TMP4202:%.*]] = extractvalue { i64, i1 } [[TMP4200]], 1
+// CHECK-NEXT:    br i1 [[TMP4202]], label [[LLX_ATOMIC_EXIT411:%.*]], label [[LLX_ATOMIC_CONT412:%.*]]
+// CHECK:       llx.atomic.cont412:
+// CHECK-NEXT:    store i64 [[TMP4201]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT411]]
+// CHECK:       llx.atomic.exit411:
+// CHECK-NEXT:    [[TMP4203:%.*]] = extractvalue { i64, i1 } [[TMP4200]], 1
+// CHECK-NEXT:    [[TMP4204:%.*]] = zext i1 [[TMP4203]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4204]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4205:%.*]] = load i64, i64* [[LLE]], align 8
+// CHECK-NEXT:    [[TMP4206:%.*]] = load i64, i64* [[LLD]], align 8
+// CHECK-NEXT:    [[TMP4207:%.*]] = cmpxchg i64* [[LLX]], i64 [[TMP4205]], i64 [[TMP4206]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4208:%.*]] = extractvalue { i64, i1 } [[TMP4207]], 0
+// CHECK-NEXT:    [[TMP4209:%.*]] = extractvalue { i64, i1 } [[TMP4207]], 1
+// CHECK-NEXT:    br i1 [[TMP4209]], label [[LLX_ATOMIC_EXIT413:%.*]], label [[LLX_ATOMIC_CONT414:%.*]]
+// CHECK:       llx.atomic.cont414:
+// CHECK-NEXT:    store i64 [[TMP4208]], i64* [[LLV]], align 8
+// CHECK-NEXT:    br label [[LLX_ATOMIC_EXIT413]]
+// CHECK:       llx.atomic.exit413:
+// CHECK-NEXT:    [[TMP4210:%.*]] = extractvalue { i64, i1 } [[TMP4207]], 1
+// CHECK-NEXT:    [[TMP4211:%.*]] = zext i1 [[TMP4210]] to i64
+// CHECK-NEXT:    store volatile i64 [[TMP4211]], i64* [[LLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4212:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4213:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4212]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4213]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4214:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4215:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4214]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4215]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4216:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4217:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4216]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4217]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4218:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4219:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4218]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4219]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4220:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4221:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4222:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4220]], i64 [[TMP4221]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4223:%.*]] = extractvalue { i64, i1 } [[TMP4222]], 0
+// CHECK-NEXT:    store i64 [[TMP4223]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4224:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4225:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4226:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4224]], i64 [[TMP4225]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4227:%.*]] = extractvalue { i64, i1 } [[TMP4226]], 0
+// CHECK-NEXT:    store i64 [[TMP4227]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4228:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4229:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4228]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4230:%.*]] = icmp ugt i64 [[TMP4229]], [[TMP4228]]
+// CHECK-NEXT:    [[TMP4231:%.*]] = select i1 [[TMP4230]], i64 [[TMP4228]], i64 [[TMP4229]]
+// CHECK-NEXT:    store i64 [[TMP4231]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4232:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4233:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4232]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4234:%.*]] = icmp ult i64 [[TMP4233]], [[TMP4232]]
+// CHECK-NEXT:    [[TMP4235:%.*]] = select i1 [[TMP4234]], i64 [[TMP4232]], i64 [[TMP4233]]
+// CHECK-NEXT:    store i64 [[TMP4235]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4236:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4237:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4236]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4238:%.*]] = icmp ult i64 [[TMP4237]], [[TMP4236]]
+// CHECK-NEXT:    [[TMP4239:%.*]] = select i1 [[TMP4238]], i64 [[TMP4236]], i64 [[TMP4237]]
+// CHECK-NEXT:    store i64 [[TMP4239]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4240:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4241:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4240]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4242:%.*]] = icmp ugt i64 [[TMP4241]], [[TMP4240]]
+// CHECK-NEXT:    [[TMP4243:%.*]] = select i1 [[TMP4242]], i64 [[TMP4240]], i64 [[TMP4241]]
+// CHECK-NEXT:    store i64 [[TMP4243]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4244:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4245:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4246:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4244]], i64 [[TMP4245]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4247:%.*]] = extractvalue { i64, i1 } [[TMP4246]], 0
+// CHECK-NEXT:    [[TMP4248:%.*]] = extractvalue { i64, i1 } [[TMP4246]], 1
+// CHECK-NEXT:    [[TMP4249:%.*]] = select i1 [[TMP4248]], i64 [[TMP4244]], i64 [[TMP4247]]
+// CHECK-NEXT:    store i64 [[TMP4249]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4250:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4251:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4252:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4250]], i64 [[TMP4251]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4253:%.*]] = extractvalue { i64, i1 } [[TMP4252]], 0
+// CHECK-NEXT:    [[TMP4254:%.*]] = extractvalue { i64, i1 } [[TMP4252]], 1
+// CHECK-NEXT:    [[TMP4255:%.*]] = select i1 [[TMP4254]], i64 [[TMP4250]], i64 [[TMP4253]]
+// CHECK-NEXT:    store i64 [[TMP4255]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4256:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4257:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4258:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4256]], i64 [[TMP4257]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4259:%.*]] = extractvalue { i64, i1 } [[TMP4258]], 0
+// CHECK-NEXT:    [[TMP4260:%.*]] = extractvalue { i64, i1 } [[TMP4258]], 1
+// CHECK-NEXT:    br i1 [[TMP4260]], label [[ULLX_ATOMIC_EXIT:%.*]], label [[ULLX_ATOMIC_CONT:%.*]]
+// CHECK:       ullx.atomic.cont:
+// CHECK-NEXT:    store i64 [[TMP4259]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT]]
+// CHECK:       ullx.atomic.exit:
+// CHECK-NEXT:    [[TMP4261:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4262:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4263:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4261]], i64 [[TMP4262]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4264:%.*]] = extractvalue { i64, i1 } [[TMP4263]], 0
+// CHECK-NEXT:    [[TMP4265:%.*]] = extractvalue { i64, i1 } [[TMP4263]], 1
+// CHECK-NEXT:    br i1 [[TMP4265]], label [[ULLX_ATOMIC_EXIT415:%.*]], label [[ULLX_ATOMIC_CONT416:%.*]]
+// CHECK:       ullx.atomic.cont416:
+// CHECK-NEXT:    store i64 [[TMP4264]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT415]]
+// CHECK:       ullx.atomic.exit415:
+// CHECK-NEXT:    [[TMP4266:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4267:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4268:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4266]], i64 [[TMP4267]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4269:%.*]] = extractvalue { i64, i1 } [[TMP4268]], 1
+// CHECK-NEXT:    [[TMP4270:%.*]] = zext i1 [[TMP4269]] to i64
+// CHECK-NEXT:    store i64 [[TMP4270]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4271:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4272:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4273:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4271]], i64 [[TMP4272]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4274:%.*]] = extractvalue { i64, i1 } [[TMP4273]], 1
+// CHECK-NEXT:    [[TMP4275:%.*]] = zext i1 [[TMP4274]] to i64
+// CHECK-NEXT:    store i64 [[TMP4275]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4276:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4277:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4278:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4276]], i64 [[TMP4277]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4279:%.*]] = extractvalue { i64, i1 } [[TMP4278]], 0
+// CHECK-NEXT:    [[TMP4280:%.*]] = extractvalue { i64, i1 } [[TMP4278]], 1
+// CHECK-NEXT:    br i1 [[TMP4280]], label [[ULLX_ATOMIC_EXIT417:%.*]], label [[ULLX_ATOMIC_CONT418:%.*]]
+// CHECK:       ullx.atomic.cont418:
+// CHECK-NEXT:    store i64 [[TMP4279]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT417]]
+// CHECK:       ullx.atomic.exit417:
+// CHECK-NEXT:    [[TMP4281:%.*]] = extractvalue { i64, i1 } [[TMP4278]], 1
+// CHECK-NEXT:    [[TMP4282:%.*]] = zext i1 [[TMP4281]] to i64
+// CHECK-NEXT:    store i64 [[TMP4282]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4283:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4284:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4285:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4283]], i64 [[TMP4284]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4286:%.*]] = extractvalue { i64, i1 } [[TMP4285]], 0
+// CHECK-NEXT:    [[TMP4287:%.*]] = extractvalue { i64, i1 } [[TMP4285]], 1
+// CHECK-NEXT:    br i1 [[TMP4287]], label [[ULLX_ATOMIC_EXIT419:%.*]], label [[ULLX_ATOMIC_CONT420:%.*]]
+// CHECK:       ullx.atomic.cont420:
+// CHECK-NEXT:    store i64 [[TMP4286]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT419]]
+// CHECK:       ullx.atomic.exit419:
+// CHECK-NEXT:    [[TMP4288:%.*]] = extractvalue { i64, i1 } [[TMP4285]], 1
+// CHECK-NEXT:    [[TMP4289:%.*]] = zext i1 [[TMP4288]] to i64
+// CHECK-NEXT:    store i64 [[TMP4289]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4290:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4291:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4290]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP4291]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4292:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4293:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4292]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP4293]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4294:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4295:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4294]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP4295]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4296:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4297:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4296]] acq_rel, align 8
+// CHECK-NEXT:    store i64 [[TMP4297]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4298:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4299:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4300:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4298]], i64 [[TMP4299]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4301:%.*]] = extractvalue { i64, i1 } [[TMP4300]], 0
+// CHECK-NEXT:    store i64 [[TMP4301]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4302:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4303:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4304:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4302]], i64 [[TMP4303]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4305:%.*]] = extractvalue { i64, i1 } [[TMP4304]], 0
+// CHECK-NEXT:    store i64 [[TMP4305]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4306:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4307:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4306]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP4308:%.*]] = icmp ugt i64 [[TMP4307]], [[TMP4306]]
+// CHECK-NEXT:    [[TMP4309:%.*]] = select i1 [[TMP4308]], i64 [[TMP4306]], i64 [[TMP4307]]
+// CHECK-NEXT:    store i64 [[TMP4309]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4310:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4311:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4310]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP4312:%.*]] = icmp ult i64 [[TMP4311]], [[TMP4310]]
+// CHECK-NEXT:    [[TMP4313:%.*]] = select i1 [[TMP4312]], i64 [[TMP4310]], i64 [[TMP4311]]
+// CHECK-NEXT:    store i64 [[TMP4313]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4314:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4315:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4314]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP4316:%.*]] = icmp ult i64 [[TMP4315]], [[TMP4314]]
+// CHECK-NEXT:    [[TMP4317:%.*]] = select i1 [[TMP4316]], i64 [[TMP4314]], i64 [[TMP4315]]
+// CHECK-NEXT:    store i64 [[TMP4317]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4318:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4319:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4318]] acq_rel, align 8
+// CHECK-NEXT:    [[TMP4320:%.*]] = icmp ugt i64 [[TMP4319]], [[TMP4318]]
+// CHECK-NEXT:    [[TMP4321:%.*]] = select i1 [[TMP4320]], i64 [[TMP4318]], i64 [[TMP4319]]
+// CHECK-NEXT:    store i64 [[TMP4321]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4322:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4323:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4324:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4322]], i64 [[TMP4323]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4325:%.*]] = extractvalue { i64, i1 } [[TMP4324]], 0
+// CHECK-NEXT:    [[TMP4326:%.*]] = extractvalue { i64, i1 } [[TMP4324]], 1
+// CHECK-NEXT:    [[TMP4327:%.*]] = select i1 [[TMP4326]], i64 [[TMP4322]], i64 [[TMP4325]]
+// CHECK-NEXT:    store i64 [[TMP4327]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4328:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4329:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4330:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4328]], i64 [[TMP4329]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4331:%.*]] = extractvalue { i64, i1 } [[TMP4330]], 0
+// CHECK-NEXT:    [[TMP4332:%.*]] = extractvalue { i64, i1 } [[TMP4330]], 1
+// CHECK-NEXT:    [[TMP4333:%.*]] = select i1 [[TMP4332]], i64 [[TMP4328]], i64 [[TMP4331]]
+// CHECK-NEXT:    store i64 [[TMP4333]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4334:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4335:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4336:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4334]], i64 [[TMP4335]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4337:%.*]] = extractvalue { i64, i1 } [[TMP4336]], 0
+// CHECK-NEXT:    [[TMP4338:%.*]] = extractvalue { i64, i1 } [[TMP4336]], 1
+// CHECK-NEXT:    br i1 [[TMP4338]], label [[ULLX_ATOMIC_EXIT421:%.*]], label [[ULLX_ATOMIC_CONT422:%.*]]
+// CHECK:       ullx.atomic.cont422:
+// CHECK-NEXT:    store i64 [[TMP4337]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT421]]
+// CHECK:       ullx.atomic.exit421:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4339:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4340:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4341:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4339]], i64 [[TMP4340]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4342:%.*]] = extractvalue { i64, i1 } [[TMP4341]], 0
+// CHECK-NEXT:    [[TMP4343:%.*]] = extractvalue { i64, i1 } [[TMP4341]], 1
+// CHECK-NEXT:    br i1 [[TMP4343]], label [[ULLX_ATOMIC_EXIT423:%.*]], label [[ULLX_ATOMIC_CONT424:%.*]]
+// CHECK:       ullx.atomic.cont424:
+// CHECK-NEXT:    store i64 [[TMP4342]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT423]]
+// CHECK:       ullx.atomic.exit423:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4344:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4345:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4346:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4344]], i64 [[TMP4345]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4347:%.*]] = extractvalue { i64, i1 } [[TMP4346]], 1
+// CHECK-NEXT:    [[TMP4348:%.*]] = zext i1 [[TMP4347]] to i64
+// CHECK-NEXT:    store i64 [[TMP4348]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4349:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4350:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4351:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4349]], i64 [[TMP4350]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4352:%.*]] = extractvalue { i64, i1 } [[TMP4351]], 1
+// CHECK-NEXT:    [[TMP4353:%.*]] = zext i1 [[TMP4352]] to i64
+// CHECK-NEXT:    store i64 [[TMP4353]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4354:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4355:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4356:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4354]], i64 [[TMP4355]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4357:%.*]] = extractvalue { i64, i1 } [[TMP4356]], 0
+// CHECK-NEXT:    [[TMP4358:%.*]] = extractvalue { i64, i1 } [[TMP4356]], 1
+// CHECK-NEXT:    br i1 [[TMP4358]], label [[ULLX_ATOMIC_EXIT425:%.*]], label [[ULLX_ATOMIC_CONT426:%.*]]
+// CHECK:       ullx.atomic.cont426:
+// CHECK-NEXT:    store i64 [[TMP4357]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT425]]
+// CHECK:       ullx.atomic.exit425:
+// CHECK-NEXT:    [[TMP4359:%.*]] = extractvalue { i64, i1 } [[TMP4356]], 1
+// CHECK-NEXT:    [[TMP4360:%.*]] = zext i1 [[TMP4359]] to i64
+// CHECK-NEXT:    store i64 [[TMP4360]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4361:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4362:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4363:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4361]], i64 [[TMP4362]] acq_rel acquire, align 8
+// CHECK-NEXT:    [[TMP4364:%.*]] = extractvalue { i64, i1 } [[TMP4363]], 0
+// CHECK-NEXT:    [[TMP4365:%.*]] = extractvalue { i64, i1 } [[TMP4363]], 1
+// CHECK-NEXT:    br i1 [[TMP4365]], label [[ULLX_ATOMIC_EXIT427:%.*]], label [[ULLX_ATOMIC_CONT428:%.*]]
+// CHECK:       ullx.atomic.cont428:
+// CHECK-NEXT:    store i64 [[TMP4364]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT427]]
+// CHECK:       ullx.atomic.exit427:
+// CHECK-NEXT:    [[TMP4366:%.*]] = extractvalue { i64, i1 } [[TMP4363]], 1
+// CHECK-NEXT:    [[TMP4367:%.*]] = zext i1 [[TMP4366]] to i64
+// CHECK-NEXT:    store i64 [[TMP4367]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4368:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4369:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4368]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP4369]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4370:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4371:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4370]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP4371]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4372:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4373:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4372]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP4373]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4374:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4375:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4374]] acquire, align 8
+// CHECK-NEXT:    store i64 [[TMP4375]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4376:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4377:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4378:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4376]], i64 [[TMP4377]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4379:%.*]] = extractvalue { i64, i1 } [[TMP4378]], 0
+// CHECK-NEXT:    store i64 [[TMP4379]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4380:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4381:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4382:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4380]], i64 [[TMP4381]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4383:%.*]] = extractvalue { i64, i1 } [[TMP4382]], 0
+// CHECK-NEXT:    store i64 [[TMP4383]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4384:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4385:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4384]] acquire, align 8
+// CHECK-NEXT:    [[TMP4386:%.*]] = icmp ugt i64 [[TMP4385]], [[TMP4384]]
+// CHECK-NEXT:    [[TMP4387:%.*]] = select i1 [[TMP4386]], i64 [[TMP4384]], i64 [[TMP4385]]
+// CHECK-NEXT:    store i64 [[TMP4387]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4388:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4389:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4388]] acquire, align 8
+// CHECK-NEXT:    [[TMP4390:%.*]] = icmp ult i64 [[TMP4389]], [[TMP4388]]
+// CHECK-NEXT:    [[TMP4391:%.*]] = select i1 [[TMP4390]], i64 [[TMP4388]], i64 [[TMP4389]]
+// CHECK-NEXT:    store i64 [[TMP4391]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4392:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4393:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4392]] acquire, align 8
+// CHECK-NEXT:    [[TMP4394:%.*]] = icmp ult i64 [[TMP4393]], [[TMP4392]]
+// CHECK-NEXT:    [[TMP4395:%.*]] = select i1 [[TMP4394]], i64 [[TMP4392]], i64 [[TMP4393]]
+// CHECK-NEXT:    store i64 [[TMP4395]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4396:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4397:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4396]] acquire, align 8
+// CHECK-NEXT:    [[TMP4398:%.*]] = icmp ugt i64 [[TMP4397]], [[TMP4396]]
+// CHECK-NEXT:    [[TMP4399:%.*]] = select i1 [[TMP4398]], i64 [[TMP4396]], i64 [[TMP4397]]
+// CHECK-NEXT:    store i64 [[TMP4399]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4400:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4401:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4402:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4400]], i64 [[TMP4401]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4403:%.*]] = extractvalue { i64, i1 } [[TMP4402]], 0
+// CHECK-NEXT:    [[TMP4404:%.*]] = extractvalue { i64, i1 } [[TMP4402]], 1
+// CHECK-NEXT:    [[TMP4405:%.*]] = select i1 [[TMP4404]], i64 [[TMP4400]], i64 [[TMP4403]]
+// CHECK-NEXT:    store i64 [[TMP4405]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4406:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4407:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4408:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4406]], i64 [[TMP4407]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4409:%.*]] = extractvalue { i64, i1 } [[TMP4408]], 0
+// CHECK-NEXT:    [[TMP4410:%.*]] = extractvalue { i64, i1 } [[TMP4408]], 1
+// CHECK-NEXT:    [[TMP4411:%.*]] = select i1 [[TMP4410]], i64 [[TMP4406]], i64 [[TMP4409]]
+// CHECK-NEXT:    store i64 [[TMP4411]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4412:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4413:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4414:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4412]], i64 [[TMP4413]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4415:%.*]] = extractvalue { i64, i1 } [[TMP4414]], 0
+// CHECK-NEXT:    [[TMP4416:%.*]] = extractvalue { i64, i1 } [[TMP4414]], 1
+// CHECK-NEXT:    br i1 [[TMP4416]], label [[ULLX_ATOMIC_EXIT429:%.*]], label [[ULLX_ATOMIC_CONT430:%.*]]
+// CHECK:       ullx.atomic.cont430:
+// CHECK-NEXT:    store i64 [[TMP4415]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT429]]
+// CHECK:       ullx.atomic.exit429:
+// CHECK-NEXT:    [[TMP4417:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4418:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4419:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4417]], i64 [[TMP4418]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4420:%.*]] = extractvalue { i64, i1 } [[TMP4419]], 0
+// CHECK-NEXT:    [[TMP4421:%.*]] = extractvalue { i64, i1 } [[TMP4419]], 1
+// CHECK-NEXT:    br i1 [[TMP4421]], label [[ULLX_ATOMIC_EXIT431:%.*]], label [[ULLX_ATOMIC_CONT432:%.*]]
+// CHECK:       ullx.atomic.cont432:
+// CHECK-NEXT:    store i64 [[TMP4420]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT431]]
+// CHECK:       ullx.atomic.exit431:
+// CHECK-NEXT:    [[TMP4422:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4423:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4424:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4422]], i64 [[TMP4423]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4425:%.*]] = extractvalue { i64, i1 } [[TMP4424]], 1
+// CHECK-NEXT:    [[TMP4426:%.*]] = zext i1 [[TMP4425]] to i64
+// CHECK-NEXT:    store i64 [[TMP4426]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4427:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4428:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4429:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4427]], i64 [[TMP4428]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4430:%.*]] = extractvalue { i64, i1 } [[TMP4429]], 1
+// CHECK-NEXT:    [[TMP4431:%.*]] = zext i1 [[TMP4430]] to i64
+// CHECK-NEXT:    store i64 [[TMP4431]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4432:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4433:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4434:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4432]], i64 [[TMP4433]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4435:%.*]] = extractvalue { i64, i1 } [[TMP4434]], 0
+// CHECK-NEXT:    [[TMP4436:%.*]] = extractvalue { i64, i1 } [[TMP4434]], 1
+// CHECK-NEXT:    br i1 [[TMP4436]], label [[ULLX_ATOMIC_EXIT433:%.*]], label [[ULLX_ATOMIC_CONT434:%.*]]
+// CHECK:       ullx.atomic.cont434:
+// CHECK-NEXT:    store i64 [[TMP4435]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT433]]
+// CHECK:       ullx.atomic.exit433:
+// CHECK-NEXT:    [[TMP4437:%.*]] = extractvalue { i64, i1 } [[TMP4434]], 1
+// CHECK-NEXT:    [[TMP4438:%.*]] = zext i1 [[TMP4437]] to i64
+// CHECK-NEXT:    store i64 [[TMP4438]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4439:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4440:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4441:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4439]], i64 [[TMP4440]] acquire acquire, align 8
+// CHECK-NEXT:    [[TMP4442:%.*]] = extractvalue { i64, i1 } [[TMP4441]], 0
+// CHECK-NEXT:    [[TMP4443:%.*]] = extractvalue { i64, i1 } [[TMP4441]], 1
+// CHECK-NEXT:    br i1 [[TMP4443]], label [[ULLX_ATOMIC_EXIT435:%.*]], label [[ULLX_ATOMIC_CONT436:%.*]]
+// CHECK:       ullx.atomic.cont436:
+// CHECK-NEXT:    store i64 [[TMP4442]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT435]]
+// CHECK:       ullx.atomic.exit435:
+// CHECK-NEXT:    [[TMP4444:%.*]] = extractvalue { i64, i1 } [[TMP4441]], 1
+// CHECK-NEXT:    [[TMP4445:%.*]] = zext i1 [[TMP4444]] to i64
+// CHECK-NEXT:    store i64 [[TMP4445]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4446:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4447:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4446]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4447]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4448:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4449:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4448]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4449]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4450:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4451:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4450]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4451]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4452:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4453:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4452]] monotonic, align 8
+// CHECK-NEXT:    store i64 [[TMP4453]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4454:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4455:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4456:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4454]], i64 [[TMP4455]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4457:%.*]] = extractvalue { i64, i1 } [[TMP4456]], 0
+// CHECK-NEXT:    store i64 [[TMP4457]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4458:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4459:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4460:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4458]], i64 [[TMP4459]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4461:%.*]] = extractvalue { i64, i1 } [[TMP4460]], 0
+// CHECK-NEXT:    store i64 [[TMP4461]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4462:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4463:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4462]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4464:%.*]] = icmp ugt i64 [[TMP4463]], [[TMP4462]]
+// CHECK-NEXT:    [[TMP4465:%.*]] = select i1 [[TMP4464]], i64 [[TMP4462]], i64 [[TMP4463]]
+// CHECK-NEXT:    store i64 [[TMP4465]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4466:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4467:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4466]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4468:%.*]] = icmp ult i64 [[TMP4467]], [[TMP4466]]
+// CHECK-NEXT:    [[TMP4469:%.*]] = select i1 [[TMP4468]], i64 [[TMP4466]], i64 [[TMP4467]]
+// CHECK-NEXT:    store i64 [[TMP4469]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4470:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4471:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4470]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4472:%.*]] = icmp ult i64 [[TMP4471]], [[TMP4470]]
+// CHECK-NEXT:    [[TMP4473:%.*]] = select i1 [[TMP4472]], i64 [[TMP4470]], i64 [[TMP4471]]
+// CHECK-NEXT:    store i64 [[TMP4473]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4474:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4475:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4474]] monotonic, align 8
+// CHECK-NEXT:    [[TMP4476:%.*]] = icmp ugt i64 [[TMP4475]], [[TMP4474]]
+// CHECK-NEXT:    [[TMP4477:%.*]] = select i1 [[TMP4476]], i64 [[TMP4474]], i64 [[TMP4475]]
+// CHECK-NEXT:    store i64 [[TMP4477]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4478:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4479:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4480:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4478]], i64 [[TMP4479]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4481:%.*]] = extractvalue { i64, i1 } [[TMP4480]], 0
+// CHECK-NEXT:    [[TMP4482:%.*]] = extractvalue { i64, i1 } [[TMP4480]], 1
+// CHECK-NEXT:    [[TMP4483:%.*]] = select i1 [[TMP4482]], i64 [[TMP4478]], i64 [[TMP4481]]
+// CHECK-NEXT:    store i64 [[TMP4483]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4484:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4485:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4486:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4484]], i64 [[TMP4485]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4487:%.*]] = extractvalue { i64, i1 } [[TMP4486]], 0
+// CHECK-NEXT:    [[TMP4488:%.*]] = extractvalue { i64, i1 } [[TMP4486]], 1
+// CHECK-NEXT:    [[TMP4489:%.*]] = select i1 [[TMP4488]], i64 [[TMP4484]], i64 [[TMP4487]]
+// CHECK-NEXT:    store i64 [[TMP4489]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    [[TMP4490:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4491:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4492:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4490]], i64 [[TMP4491]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4493:%.*]] = extractvalue { i64, i1 } [[TMP4492]], 0
+// CHECK-NEXT:    [[TMP4494:%.*]] = extractvalue { i64, i1 } [[TMP4492]], 1
+// CHECK-NEXT:    br i1 [[TMP4494]], label [[ULLX_ATOMIC_EXIT437:%.*]], label [[ULLX_ATOMIC_CONT438:%.*]]
+// CHECK:       ullx.atomic.cont438:
+// CHECK-NEXT:    store i64 [[TMP4493]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT437]]
+// CHECK:       ullx.atomic.exit437:
+// CHECK-NEXT:    [[TMP4495:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4496:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4497:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4495]], i64 [[TMP4496]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4498:%.*]] = extractvalue { i64, i1 } [[TMP4497]], 0
+// CHECK-NEXT:    [[TMP4499:%.*]] = extractvalue { i64, i1 } [[TMP4497]], 1
+// CHECK-NEXT:    br i1 [[TMP4499]], label [[ULLX_ATOMIC_EXIT439:%.*]], label [[ULLX_ATOMIC_CONT440:%.*]]
+// CHECK:       ullx.atomic.cont440:
+// CHECK-NEXT:    store i64 [[TMP4498]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT439]]
+// CHECK:       ullx.atomic.exit439:
+// CHECK-NEXT:    [[TMP4500:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4501:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4502:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4500]], i64 [[TMP4501]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4503:%.*]] = extractvalue { i64, i1 } [[TMP4502]], 1
+// CHECK-NEXT:    [[TMP4504:%.*]] = zext i1 [[TMP4503]] to i64
+// CHECK-NEXT:    store i64 [[TMP4504]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4505:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4506:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4507:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4505]], i64 [[TMP4506]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4508:%.*]] = extractvalue { i64, i1 } [[TMP4507]], 1
+// CHECK-NEXT:    [[TMP4509:%.*]] = zext i1 [[TMP4508]] to i64
+// CHECK-NEXT:    store i64 [[TMP4509]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4510:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4511:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4512:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4510]], i64 [[TMP4511]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4513:%.*]] = extractvalue { i64, i1 } [[TMP4512]], 0
+// CHECK-NEXT:    [[TMP4514:%.*]] = extractvalue { i64, i1 } [[TMP4512]], 1
+// CHECK-NEXT:    br i1 [[TMP4514]], label [[ULLX_ATOMIC_EXIT441:%.*]], label [[ULLX_ATOMIC_CONT442:%.*]]
+// CHECK:       ullx.atomic.cont442:
+// CHECK-NEXT:    store i64 [[TMP4513]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT441]]
+// CHECK:       ullx.atomic.exit441:
+// CHECK-NEXT:    [[TMP4515:%.*]] = extractvalue { i64, i1 } [[TMP4512]], 1
+// CHECK-NEXT:    [[TMP4516:%.*]] = zext i1 [[TMP4515]] to i64
+// CHECK-NEXT:    store i64 [[TMP4516]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4517:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4518:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4519:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4517]], i64 [[TMP4518]] monotonic monotonic, align 8
+// CHECK-NEXT:    [[TMP4520:%.*]] = extractvalue { i64, i1 } [[TMP4519]], 0
+// CHECK-NEXT:    [[TMP4521:%.*]] = extractvalue { i64, i1 } [[TMP4519]], 1
+// CHECK-NEXT:    br i1 [[TMP4521]], label [[ULLX_ATOMIC_EXIT443:%.*]], label [[ULLX_ATOMIC_CONT444:%.*]]
+// CHECK:       ullx.atomic.cont444:
+// CHECK-NEXT:    store i64 [[TMP4520]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT443]]
+// CHECK:       ullx.atomic.exit443:
+// CHECK-NEXT:    [[TMP4522:%.*]] = extractvalue { i64, i1 } [[TMP4519]], 1
+// CHECK-NEXT:    [[TMP4523:%.*]] = zext i1 [[TMP4522]] to i64
+// CHECK-NEXT:    store i64 [[TMP4523]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    [[TMP4524:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4525:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4524]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP4525]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4526:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4527:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4526]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP4527]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4528:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4529:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4528]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP4529]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4530:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4531:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4530]] release, align 8
+// CHECK-NEXT:    store i64 [[TMP4531]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4532:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4533:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4534:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4532]], i64 [[TMP4533]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4535:%.*]] = extractvalue { i64, i1 } [[TMP4534]], 0
+// CHECK-NEXT:    store i64 [[TMP4535]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4536:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4537:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4538:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4536]], i64 [[TMP4537]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4539:%.*]] = extractvalue { i64, i1 } [[TMP4538]], 0
+// CHECK-NEXT:    store i64 [[TMP4539]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4540:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4541:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4540]] release, align 8
+// CHECK-NEXT:    [[TMP4542:%.*]] = icmp ugt i64 [[TMP4541]], [[TMP4540]]
+// CHECK-NEXT:    [[TMP4543:%.*]] = select i1 [[TMP4542]], i64 [[TMP4540]], i64 [[TMP4541]]
+// CHECK-NEXT:    store i64 [[TMP4543]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4544:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4545:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4544]] release, align 8
+// CHECK-NEXT:    [[TMP4546:%.*]] = icmp ult i64 [[TMP4545]], [[TMP4544]]
+// CHECK-NEXT:    [[TMP4547:%.*]] = select i1 [[TMP4546]], i64 [[TMP4544]], i64 [[TMP4545]]
+// CHECK-NEXT:    store i64 [[TMP4547]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4548:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4549:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4548]] release, align 8
+// CHECK-NEXT:    [[TMP4550:%.*]] = icmp ult i64 [[TMP4549]], [[TMP4548]]
+// CHECK-NEXT:    [[TMP4551:%.*]] = select i1 [[TMP4550]], i64 [[TMP4548]], i64 [[TMP4549]]
+// CHECK-NEXT:    store i64 [[TMP4551]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4552:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4553:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4552]] release, align 8
+// CHECK-NEXT:    [[TMP4554:%.*]] = icmp ugt i64 [[TMP4553]], [[TMP4552]]
+// CHECK-NEXT:    [[TMP4555:%.*]] = select i1 [[TMP4554]], i64 [[TMP4552]], i64 [[TMP4553]]
+// CHECK-NEXT:    store i64 [[TMP4555]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4556:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4557:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4558:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4556]], i64 [[TMP4557]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4559:%.*]] = extractvalue { i64, i1 } [[TMP4558]], 0
+// CHECK-NEXT:    [[TMP4560:%.*]] = extractvalue { i64, i1 } [[TMP4558]], 1
+// CHECK-NEXT:    [[TMP4561:%.*]] = select i1 [[TMP4560]], i64 [[TMP4556]], i64 [[TMP4559]]
+// CHECK-NEXT:    store i64 [[TMP4561]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4562:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4563:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4564:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4562]], i64 [[TMP4563]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4565:%.*]] = extractvalue { i64, i1 } [[TMP4564]], 0
+// CHECK-NEXT:    [[TMP4566:%.*]] = extractvalue { i64, i1 } [[TMP4564]], 1
+// CHECK-NEXT:    [[TMP4567:%.*]] = select i1 [[TMP4566]], i64 [[TMP4562]], i64 [[TMP4565]]
+// CHECK-NEXT:    store i64 [[TMP4567]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4568:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4569:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4570:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4568]], i64 [[TMP4569]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4571:%.*]] = extractvalue { i64, i1 } [[TMP4570]], 0
+// CHECK-NEXT:    [[TMP4572:%.*]] = extractvalue { i64, i1 } [[TMP4570]], 1
+// CHECK-NEXT:    br i1 [[TMP4572]], label [[ULLX_ATOMIC_EXIT445:%.*]], label [[ULLX_ATOMIC_CONT446:%.*]]
+// CHECK:       ullx.atomic.cont446:
+// CHECK-NEXT:    store i64 [[TMP4571]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT445]]
+// CHECK:       ullx.atomic.exit445:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4573:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4574:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4575:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4573]], i64 [[TMP4574]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4576:%.*]] = extractvalue { i64, i1 } [[TMP4575]], 0
+// CHECK-NEXT:    [[TMP4577:%.*]] = extractvalue { i64, i1 } [[TMP4575]], 1
+// CHECK-NEXT:    br i1 [[TMP4577]], label [[ULLX_ATOMIC_EXIT447:%.*]], label [[ULLX_ATOMIC_CONT448:%.*]]
+// CHECK:       ullx.atomic.cont448:
+// CHECK-NEXT:    store i64 [[TMP4576]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT447]]
+// CHECK:       ullx.atomic.exit447:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4578:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4579:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4580:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4578]], i64 [[TMP4579]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4581:%.*]] = extractvalue { i64, i1 } [[TMP4580]], 1
+// CHECK-NEXT:    [[TMP4582:%.*]] = zext i1 [[TMP4581]] to i64
+// CHECK-NEXT:    store i64 [[TMP4582]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4583:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4584:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4585:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4583]], i64 [[TMP4584]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4586:%.*]] = extractvalue { i64, i1 } [[TMP4585]], 1
+// CHECK-NEXT:    [[TMP4587:%.*]] = zext i1 [[TMP4586]] to i64
+// CHECK-NEXT:    store i64 [[TMP4587]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4588:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4589:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4590:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4588]], i64 [[TMP4589]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4591:%.*]] = extractvalue { i64, i1 } [[TMP4590]], 0
+// CHECK-NEXT:    [[TMP4592:%.*]] = extractvalue { i64, i1 } [[TMP4590]], 1
+// CHECK-NEXT:    br i1 [[TMP4592]], label [[ULLX_ATOMIC_EXIT449:%.*]], label [[ULLX_ATOMIC_CONT450:%.*]]
+// CHECK:       ullx.atomic.cont450:
+// CHECK-NEXT:    store i64 [[TMP4591]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT449]]
+// CHECK:       ullx.atomic.exit449:
+// CHECK-NEXT:    [[TMP4593:%.*]] = extractvalue { i64, i1 } [[TMP4590]], 1
+// CHECK-NEXT:    [[TMP4594:%.*]] = zext i1 [[TMP4593]] to i64
+// CHECK-NEXT:    store i64 [[TMP4594]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4595:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4596:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4597:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4595]], i64 [[TMP4596]] release monotonic, align 8
+// CHECK-NEXT:    [[TMP4598:%.*]] = extractvalue { i64, i1 } [[TMP4597]], 0
+// CHECK-NEXT:    [[TMP4599:%.*]] = extractvalue { i64, i1 } [[TMP4597]], 1
+// CHECK-NEXT:    br i1 [[TMP4599]], label [[ULLX_ATOMIC_EXIT451:%.*]], label [[ULLX_ATOMIC_CONT452:%.*]]
+// CHECK:       ullx.atomic.cont452:
+// CHECK-NEXT:    store i64 [[TMP4598]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT451]]
+// CHECK:       ullx.atomic.exit451:
+// CHECK-NEXT:    [[TMP4600:%.*]] = extractvalue { i64, i1 } [[TMP4597]], 1
+// CHECK-NEXT:    [[TMP4601:%.*]] = zext i1 [[TMP4600]] to i64
+// CHECK-NEXT:    store i64 [[TMP4601]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4602:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4603:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4602]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP4603]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4604:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4605:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4604]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP4605]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4606:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4607:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4606]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP4607]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4608:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4609:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4608]] seq_cst, align 8
+// CHECK-NEXT:    store i64 [[TMP4609]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4610:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4611:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4612:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4610]], i64 [[TMP4611]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4613:%.*]] = extractvalue { i64, i1 } [[TMP4612]], 0
+// CHECK-NEXT:    store i64 [[TMP4613]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4614:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4615:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4616:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4614]], i64 [[TMP4615]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4617:%.*]] = extractvalue { i64, i1 } [[TMP4616]], 0
+// CHECK-NEXT:    store i64 [[TMP4617]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4618:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4619:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4618]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4620:%.*]] = icmp ugt i64 [[TMP4619]], [[TMP4618]]
+// CHECK-NEXT:    [[TMP4621:%.*]] = select i1 [[TMP4620]], i64 [[TMP4618]], i64 [[TMP4619]]
+// CHECK-NEXT:    store i64 [[TMP4621]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4622:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4623:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4622]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4624:%.*]] = icmp ult i64 [[TMP4623]], [[TMP4622]]
+// CHECK-NEXT:    [[TMP4625:%.*]] = select i1 [[TMP4624]], i64 [[TMP4622]], i64 [[TMP4623]]
+// CHECK-NEXT:    store i64 [[TMP4625]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4626:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4627:%.*]] = atomicrmw umin i64* [[ULLX]], i64 [[TMP4626]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4628:%.*]] = icmp ult i64 [[TMP4627]], [[TMP4626]]
+// CHECK-NEXT:    [[TMP4629:%.*]] = select i1 [[TMP4628]], i64 [[TMP4626]], i64 [[TMP4627]]
+// CHECK-NEXT:    store i64 [[TMP4629]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4630:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4631:%.*]] = atomicrmw umax i64* [[ULLX]], i64 [[TMP4630]] seq_cst, align 8
+// CHECK-NEXT:    [[TMP4632:%.*]] = icmp ugt i64 [[TMP4631]], [[TMP4630]]
+// CHECK-NEXT:    [[TMP4633:%.*]] = select i1 [[TMP4632]], i64 [[TMP4630]], i64 [[TMP4631]]
+// CHECK-NEXT:    store i64 [[TMP4633]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4634:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4635:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4636:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4634]], i64 [[TMP4635]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4637:%.*]] = extractvalue { i64, i1 } [[TMP4636]], 0
+// CHECK-NEXT:    [[TMP4638:%.*]] = extractvalue { i64, i1 } [[TMP4636]], 1
+// CHECK-NEXT:    [[TMP4639:%.*]] = select i1 [[TMP4638]], i64 [[TMP4634]], i64 [[TMP4637]]
+// CHECK-NEXT:    store i64 [[TMP4639]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4640:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4641:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4642:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4640]], i64 [[TMP4641]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4643:%.*]] = extractvalue { i64, i1 } [[TMP4642]], 0
+// CHECK-NEXT:    [[TMP4644:%.*]] = extractvalue { i64, i1 } [[TMP4642]], 1
+// CHECK-NEXT:    [[TMP4645:%.*]] = select i1 [[TMP4644]], i64 [[TMP4640]], i64 [[TMP4643]]
+// CHECK-NEXT:    store i64 [[TMP4645]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4646:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4647:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4648:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4646]], i64 [[TMP4647]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4649:%.*]] = extractvalue { i64, i1 } [[TMP4648]], 0
+// CHECK-NEXT:    [[TMP4650:%.*]] = extractvalue { i64, i1 } [[TMP4648]], 1
+// CHECK-NEXT:    br i1 [[TMP4650]], label [[ULLX_ATOMIC_EXIT453:%.*]], label [[ULLX_ATOMIC_CONT454:%.*]]
+// CHECK:       ullx.atomic.cont454:
+// CHECK-NEXT:    store i64 [[TMP4649]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT453]]
+// CHECK:       ullx.atomic.exit453:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4651:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4652:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4653:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4651]], i64 [[TMP4652]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4654:%.*]] = extractvalue { i64, i1 } [[TMP4653]], 0
+// CHECK-NEXT:    [[TMP4655:%.*]] = extractvalue { i64, i1 } [[TMP4653]], 1
+// CHECK-NEXT:    br i1 [[TMP4655]], label [[ULLX_ATOMIC_EXIT455:%.*]], label [[ULLX_ATOMIC_CONT456:%.*]]
+// CHECK:       ullx.atomic.cont456:
+// CHECK-NEXT:    store i64 [[TMP4654]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT455]]
+// CHECK:       ullx.atomic.exit455:
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4656:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4657:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4658:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4656]], i64 [[TMP4657]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4659:%.*]] = extractvalue { i64, i1 } [[TMP4658]], 1
+// CHECK-NEXT:    [[TMP4660:%.*]] = zext i1 [[TMP4659]] to i64
+// CHECK-NEXT:    store i64 [[TMP4660]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4661:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4662:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4663:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4661]], i64 [[TMP4662]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4664:%.*]] = extractvalue { i64, i1 } [[TMP4663]], 1
+// CHECK-NEXT:    [[TMP4665:%.*]] = zext i1 [[TMP4664]] to i64
+// CHECK-NEXT:    store i64 [[TMP4665]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4666:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4667:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4668:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4666]], i64 [[TMP4667]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4669:%.*]] = extractvalue { i64, i1 } [[TMP4668]], 0
+// CHECK-NEXT:    [[TMP4670:%.*]] = extractvalue { i64, i1 } [[TMP4668]], 1
+// CHECK-NEXT:    br i1 [[TMP4670]], label [[ULLX_ATOMIC_EXIT457:%.*]], label [[ULLX_ATOMIC_CONT458:%.*]]
+// CHECK:       ullx.atomic.cont458:
+// CHECK-NEXT:    store i64 [[TMP4669]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT457]]
+// CHECK:       ullx.atomic.exit457:
+// CHECK-NEXT:    [[TMP4671:%.*]] = extractvalue { i64, i1 } [[TMP4668]], 1
+// CHECK-NEXT:    [[TMP4672:%.*]] = zext i1 [[TMP4671]] to i64
+// CHECK-NEXT:    store i64 [[TMP4672]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    [[TMP4673:%.*]] = load i64, i64* [[ULLE]], align 8
+// CHECK-NEXT:    [[TMP4674:%.*]] = load i64, i64* [[ULLD]], align 8
+// CHECK-NEXT:    [[TMP4675:%.*]] = cmpxchg i64* [[ULLX]], i64 [[TMP4673]], i64 [[TMP4674]] seq_cst seq_cst, align 8
+// CHECK-NEXT:    [[TMP4676:%.*]] = extractvalue { i64, i1 } [[TMP4675]], 0
+// CHECK-NEXT:    [[TMP4677:%.*]] = extractvalue { i64, i1 } [[TMP4675]], 1
+// CHECK-NEXT:    br i1 [[TMP4677]], label [[ULLX_ATOMIC_EXIT459:%.*]], label [[ULLX_ATOMIC_CONT460:%.*]]
+// CHECK:       ullx.atomic.cont460:
+// CHECK-NEXT:    store i64 [[TMP4676]], i64* [[ULLV]], align 8
+// CHECK-NEXT:    br label [[ULLX_ATOMIC_EXIT459]]
+// CHECK:       ullx.atomic.exit459:
+// CHECK-NEXT:    [[TMP4678:%.*]] = extractvalue { i64, i1 } [[TMP4675]], 1
+// CHECK-NEXT:    [[TMP4679:%.*]] = zext i1 [[TMP4678]] to i64
+// CHECK-NEXT:    store i64 [[TMP4679]], i64* [[ULLR]], align 8
+// CHECK-NEXT:    call void @__kmpc_flush(%struct.ident_t* @[[GLOB1]])
+// CHECK-NEXT:    ret void