[OpenMP] Use compound operators for reduction combiner if available.

[mikerice]

Summary

The OpenMP spec seems to require the compound operators be used for
+, *, &, |, and ^ reduction. So use these if a class has those operators.
If not try the simple operators as we did previously to limit the impact
to existing code.

Fixes: https://bugs.llvm.org/show_bug.cgi?id=48584

[ABataev]

clang/lib/Sema/SemaOpenMP.cpp

16858

Why only for records?

clang/test/OpenMP/reduction_compound_op.cpp

2

Use a script to generate checks

[mikerice]

clang/lib/Sema/SemaOpenMP.cpp

16858

For scalars there is no difference in functionality afaik but it does end up generating different llvm IR. The resulting IR changes in some non-trivial ways in some cases changing atomic instructions to critical sections in some of the lit tests. I didn't want to change and possibly destabilize known working codegen for no known benefit.

clang/test/OpenMP/reduction_compound_op.cpp

2

Sorry I am not familiar with such a script. Can you give me a pointer or explain?

[ABataev]

clang/test/OpenMP/reduction_compound_op.cpp

2

The script is llvm/utils/update_cc_test_checks.py. You can check test/OpenMP/nvptx_target_codegen.cpp as an example for args

[mikerice]

Used script for the test. Fixed clang-format issue.

[ABataev]

clang/test/OpenMP/reduction_compound_op.cpp

9–10

Add checks for -fopenmp-simd flags, that no _kmpc|tgt calls are emitted

[mikerice]

clang/test/OpenMP/reduction_compound_op.cpp

9–10

Thinking about this more I am not modifying any codegen so I think it's probably better to test this in the AST. The scripted codegen test makes it very difficult to tell what is being tested. Is it okay if I drop this codegen test and write an AST test instead?

[ABataev]

clang/test/OpenMP/reduction_compound_op.cpp

9–10

Better to keep it, if possible. Just the test for simd must very simole,withou check l8nes,just need to check that no kmpc|tgt calls are generated.

[mikerice]

Added -fopenmp-simd check to reduction_compound_op.cpp
Regenerated checks for nvptx_target_parallel_reduction_codegen_tbaa_PR46146.cpp since it uses complex reduction with operator+=.

[mikerice]

Ping. Anything else needed on this one?

[ABataev]

LG, sorry, just missed the last update

clang/lib/Sema/SemaOpenMP.cpp

if (!SingleElement) {
     }
   }
 
   // This array section can be privatized as a single value or as a constant
   // sized array.
   return true;
 }
 
+static BinaryOperatorKind
+getRelatedCompoundReductionOp(BinaryOperatorKind BOK) {
+  if (BOK == BO_Add)
+    return BO_AddAssign;
+  if (BOK == BO_Mul)
+    return BO_MulAssign;
+  if (BOK == BO_And)
+    return BO_AndAssign;
+  if (BOK == BO_Or)
+    return BO_OrAssign;
+  if (BOK == BO_Xor)
+    return BO_XorAssign;
+  return BOK;
+}
+
 static bool actOnOMPReductionKindClause(
     Sema &S, DSAStackTy *Stack, OpenMPClauseKind ClauseKind,
     ArrayRef<Expr *> VarList, SourceLocation StartLoc, SourceLocation LParenLoc,
     SourceLocation ColonLoc, SourceLocation EndLoc,
     CXXScopeSpec &ReductionIdScopeSpec, const DeclarationNameInfo &ReductionId,
     ArrayRef<Expr *> UnresolvedReductions, ReductionData &RD) {
   DeclarationName DN = ReductionId.getName();
   OverloadedOperatorKind OOK = DN.getCXXOverloadedOperator();
if (DeclareReductionRef.isUsable()) {
       auto *OVE = new (Context) OpaqueValueExpr(
           ELoc, Context.getPointerType(FnTy), VK_RValue, OK_Ordinary,
           S.DefaultLvalueConversion(DeclareReductionRef.get()).get());
       Expr *Args[] = {LHS.get(), RHS.get()};
       ReductionOp =
           CallExpr::Create(Context, OVE, Args, Context.VoidTy, VK_RValue, ELoc,
                            S.CurFPFeatureOverrides());
     } else {
-      ReductionOp = S.BuildBinOp(
-          Stack->getCurScope(), ReductionId.getBeginLoc(), BOK, LHSDRE, RHSDRE);
+      BinaryOperatorKind CombBOK = getRelatedCompoundReductionOp(BOK);
+      if (Type->isRecordType() && CombBOK != BOK) {
+        Sema::TentativeAnalysisScope Trap(S);
+        ReductionOp =
+            S.BuildBinOp(Stack->getCurScope(), ReductionId.getBeginLoc(),
+                         CombBOK, LHSDRE, RHSDRE);
+      }
+      if (!ReductionOp.isUsable()) {
+        ReductionOp =
+            S.BuildBinOp(Stack->getCurScope(), ReductionId.getBeginLoc(), BOK,
+                         LHSDRE, RHSDRE);
         if (ReductionOp.isUsable()) {
           if (BOK != BO_LT && BOK != BO_GT) {
             ReductionOp =
                 S.BuildBinOp(Stack->getCurScope(), ReductionId.getBeginLoc(),
                              BO_Assign, LHSDRE, ReductionOp.get());
           } else {
             auto *ConditionalOp = new (Context)
-              ConditionalOperator(ReductionOp.get(), ELoc, LHSDRE, ELoc, RHSDRE,
-                                  Type, VK_LValue, OK_Ordinary);
+                ConditionalOperator(ReductionOp.get(), ELoc, LHSDRE, ELoc,
+                                    RHSDRE, Type, VK_LValue, OK_Ordinary);
             ReductionOp =
                 S.BuildBinOp(Stack->getCurScope(), ReductionId.getBeginLoc(),
                              BO_Assign, LHSDRE, ConditionalOp);
           }
-        if (ReductionOp.isUsable())
-          ReductionOp = S.ActOnFinishFullExpr(ReductionOp.get(),
-                                              /*DiscardedValue*/ false);
-      }
+        }
+      }
       if (!ReductionOp.isUsable())
         continue;
     }
 
     // Add copy operations for inscan reductions.
     // LHS = RHS;
     ExprResult CopyOpRes, TempArrayRes, TempArrayElem;
     if (ClauseKind == OMPC_reduction &&

clang/test/OpenMP/distribute_parallel_for_reduction_messages.cpp

for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
-#pragma omp distribute parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}} expected-warning {{Type 'S4' is not trivially copyable and not guaranteed to be mapped correctly}} expected-warning {{Type 'S5' is not trivially copyable and not guaranteed to be mapped correctly}}
+#pragma omp distribute parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}} expected-warning {{Type 'S4' is not trivially copyable and not guaranteed to be mapped correctly}} expected-warning {{Type 'S5' is not trivially copyable and not guaranteed to be mapped correctly}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}} expected-warning {{Type 'S3' is not trivially copyable and not guaranteed to be mapped correctly}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target

clang/test/OpenMP/distribute_parallel_for_simd_reduction_messages.cpp

for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
-#pragma omp distribute parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp distribute parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute parallel for simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target

clang/test/OpenMP/distribute_simd_reduction_messages.cpp

for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
-#pragma omp distribute simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}} expected-warning {{Type 'S4' is not trivially copyable and not guaranteed to be mapped correctly}} expected-warning {{Type 'S5' is not trivially copyable and not guaranteed to be mapped correctly}}}
+#pragma omp distribute simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}} expected-warning {{Type 'S4' is not trivially copyable and not guaranteed to be mapped correctly}} expected-warning {{Type 'S5' is not trivially copyable and not guaranteed to be mapped correctly}}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target
 #pragma omp teams
 #pragma omp distribute simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}} expected-warning {{Type 'S3' is not trivially copyable and not guaranteed to be mapped correctly}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target

clang/test/OpenMP/for_reduction_messages.cpp

 #pragma omp for reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
 #pragma omp for reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
-#pragma omp for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
 #pragma omp for reduction(+ : h, k) // expected-error {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
 #pragma omp for reduction(+ : B::x, k) // expected-error {{threadprivate or thread local variable cannot be reduction}}

clang/test/OpenMP/for_simd_reduction_messages.cpp

 #pragma omp for simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
 #pragma omp for simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
-#pragma omp for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
 #pragma omp for simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel
 #pragma omp for simd reduction(+ : o) // expected-error {{no viable overloaded '='}}

clang/test/OpenMP/master_taskloop_in_reduction_messages.cpp

#pragma omp master taskloop in_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp master taskloop in_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp master taskloop in_reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be in_reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
-#pragma omp master taskloop in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp master taskloop in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskgroup task_reduction(+:k)
 #pragma omp master taskloop in_reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp master taskloop in_reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)

clang/test/OpenMP/master_taskloop_reduction_messages.cpp

#pragma omp master taskloop reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp master taskloop reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp master taskloop reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/master_taskloop_simd_in_reduction_messages.cpp

#pragma omp master taskloop simd in_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp master taskloop simd in_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp master taskloop simd in_reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be in_reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
-#pragma omp master taskloop simd in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp master taskloop simd in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskgroup task_reduction(+:k)
 #pragma omp master taskloop simd in_reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp master taskloop simd in_reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)

clang/test/OpenMP/master_taskloop_simd_reduction_messages.cpp

#pragma omp master taskloop simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp master taskloop simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp master taskloop simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp master taskloop simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/parallel_for_reduction_messages.cpp

#pragma omp parallel for reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/parallel_for_simd_reduction_messages.cpp

#pragma omp parallel for simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel for simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/parallel_master_reduction_messages.cpp

 #pragma omp parallel master reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   {
     foo();
   }
 #pragma omp parallel master reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   {
     foo();
   }
-#pragma omp parallel master reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp parallel master reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   {
     foo();
   }
 #pragma omp parallel master reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   {
     foo();
   }
 #pragma omp parallel master reduction(+ : o, z) // expected-error {{no viable overloaded '='}}

clang/test/OpenMP/parallel_master_taskloop_reduction_messages.cpp

#pragma omp parallel master taskloop reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp parallel master taskloop reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp parallel master taskloop reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/parallel_master_taskloop_simd_reduction_messages.cpp

#pragma omp parallel master taskloop simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp parallel master taskloop simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp parallel master taskloop simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp parallel master taskloop simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/parallel_reduction_messages.cpp

 #pragma omp parallel reduction(- : da) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
 #pragma omp parallel reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   foo();
 #pragma omp parallel reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   foo();
 #pragma omp parallel reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
-#pragma omp parallel reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp parallel reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   foo();
 #pragma omp parallel reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   foo();
 #pragma omp parallel reduction(+ : o) // expected-error {{no viable overloaded '='}}
   foo();
 #pragma omp parallel private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/parallel_sections_reduction_messages.cpp

 #pragma omp parallel sections reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   {
     foo();
   }
 #pragma omp parallel sections reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   {
     foo();
   }
-#pragma omp parallel sections reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp parallel sections reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   {
     foo();
   }
 #pragma omp parallel sections reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   {
     foo();
   }
 #pragma omp parallel sections reduction(+ : o, z) // expected-error {{no viable overloaded '='}}

clang/test/OpenMP/reduction_compound_op.cpp

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs
+//RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fopenmp -DNORM \
+//RUN:  -emit-llvm -o - %s | FileCheck %s --check-prefix NORM
+
+//RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fopenmp -DCOMP \
+//RUN:  -emit-llvm -o - %s | FileCheck %s --check-prefix COMP
+
+// Prefer compound operators since that is what the spec seems to say.
+//RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fopenmp -DNORM -DCOMP \
+//RUN:  -emit-llvm -o - %s | FileCheck %s --check-prefix COMP
+
+struct Point {
+  int x = 0;
+  int y = 0;
+#if NORM
+  Point operator+(Point const &other) const;
+  Point operator-(Point const &other) const;
+  Point operator*(Point const &other) const;
+  Point operator&(Point const &other) const;
+  Point operator|(Point const &other) const;
+  Point operator^(Point const &other) const;
+#endif
+  Point operator&&(Point const &other) const;
+  Point operator||(Point const &other) const;
+  Point &operator=(Point const &other);
+#if COMP
+  Point &operator+=(Point const &other);
+  Point &operator*=(Point const &other);
+  Point &operator&=(Point const &other);
+  Point &operator|=(Point const &other);
+  Point &operator^=(Point const &other);
+#endif
+};
+
+void work(Point &P, int N, Point const *Points);
+
+void foo(int N, Point const *Points) {
+  Point Red;
+  #pragma omp parallel for reduction(+: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+
+  #pragma omp parallel for reduction(-: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+
+  #pragma omp parallel for reduction(*: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+
+  #pragma omp parallel for reduction(&: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+
+  #pragma omp parallel for reduction(|: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+
+  #pragma omp parallel for reduction(^: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+
+  #pragma omp parallel for reduction(&&: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+
+  #pragma omp parallel for reduction(||: Red)
+  for (unsigned I = 0; I < N; ++I)
+    work(Red, I, Points);
+}
+// NORM-LABEL: define {{[^@]+}}@_Z3fooiPK5Point
+// NORM-SAME: (i32 [[N:%.*]], %struct.Point* [[POINTS:%.*]]) #[[ATTR0:[0-9]+]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[RED:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
+// NORM-NEXT:    store %struct.Point* [[POINTS]], %struct.Point** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED]]) #[[ATTR4:[0-9]+]]
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@_ZN5PointC1Ev
+// NORM-SAME: (%struct.Point* nonnull dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    store %struct.Point* [[THIS]], %struct.Point** [[THIS_ADDR]], align 8
+// NORM-NEXT:    [[THIS1:%.*]] = load %struct.Point*, %struct.Point** [[THIS_ADDR]], align 8
+// NORM-NEXT:    call void @_ZN5PointC2Ev(%struct.Point* nonnull dereferenceable(8) [[THIS1]]) #[[ATTR4]]
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined.
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2:[0-9]+]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointplERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointplERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointplERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined..1
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointplERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointplERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointplERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined..3
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.4, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointmlERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointmlERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.4
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointmlERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined..5
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.6, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointanERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointanERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.6
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointanERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined..7
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.8, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointorERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointorERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.8
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointorERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined..9
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.10, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointeoERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointeoERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.10
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointeoERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined..11
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.12, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointaaERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointaaERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.12
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointaaERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp_outlined..13
+// NORM-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// NORM-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// NORM-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[I:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// NORM-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// NORM-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// NORM-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// NORM-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// NORM-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// NORM-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// NORM-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// NORM-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 0, i32* [[I]], align 4
+// NORM-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// NORM-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// NORM-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// NORM:       omp.precond.then:
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// NORM-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// NORM-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// NORM-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// NORM-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// NORM-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// NORM:       cond.true:
+// NORM-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// NORM-NEXT:    br label [[COND_END:%.*]]
+// NORM:       cond.false:
+// NORM-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    br label [[COND_END]]
+// NORM:       cond.end:
+// NORM-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// NORM-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// NORM-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// NORM:       omp.inner.for.cond:
+// NORM-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// NORM-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// NORM-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// NORM-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// NORM:       omp.inner.for.body:
+// NORM-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// NORM-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// NORM-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// NORM-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// NORM-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// NORM-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// NORM:       omp.body.continue:
+// NORM-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// NORM:       omp.inner.for.inc:
+// NORM-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// NORM-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// NORM-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// NORM:       omp.inner.for.end:
+// NORM-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// NORM:       omp.loop.exit:
+// NORM-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// NORM-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// NORM-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// NORM-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// NORM-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// NORM-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// NORM-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// NORM-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.14, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// NORM-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// NORM-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// NORM-NEXT:    ]
+// NORM:       .omp.reduction.case1:
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointooERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// NORM-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.case2:
+// NORM-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// NORM-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// NORM-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointooERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// NORM-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// NORM-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// NORM-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// NORM-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// NORM-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// NORM:       .omp.reduction.default:
+// NORM-NEXT:    br label [[OMP_PRECOND_END]]
+// NORM:       omp.precond.end:
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.14
+// NORM-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// NORM-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// NORM-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// NORM-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// NORM-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// NORM-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// NORM-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// NORM-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// NORM-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// NORM-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// NORM-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// NORM-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// NORM-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointooERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// NORM-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// NORM-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// NORM-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// NORM-NEXT:    ret void
+//
+//
+// NORM-LABEL: define {{[^@]+}}@_ZN5PointC2Ev
+// NORM-SAME: (%struct.Point* nonnull dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
+// NORM-NEXT:  entry:
+// NORM-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.Point*, align 8
+// NORM-NEXT:    store %struct.Point* [[THIS]], %struct.Point** [[THIS_ADDR]], align 8
+// NORM-NEXT:    [[THIS1:%.*]] = load %struct.Point*, %struct.Point** [[THIS_ADDR]], align 8
+// NORM-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_POINT:%.*]], %struct.Point* [[THIS1]], i32 0, i32 0
+// NORM-NEXT:    store i32 0, i32* [[X]], align 4
+// NORM-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_POINT]], %struct.Point* [[THIS1]], i32 0, i32 1
+// NORM-NEXT:    store i32 0, i32* [[Y]], align 4
+// NORM-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@_Z3fooiPK5Point
+// COMP-SAME: (i32 [[N:%.*]], %struct.Point* [[POINTS:%.*]]) #[[ATTR0:[0-9]+]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[RED:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
+// COMP-NEXT:    store %struct.Point* [[POINTS]], %struct.Point** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED]]) #[[ATTR4:[0-9]+]]
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, %struct.Point*, %struct.Point**)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], %struct.Point* [[RED]], %struct.Point** [[POINTS_ADDR]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@_ZN5PointC1Ev
+// COMP-SAME: (%struct.Point* nonnull dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    store %struct.Point* [[THIS]], %struct.Point** [[THIS_ADDR]], align 8
+// COMP-NEXT:    [[THIS1:%.*]] = load %struct.Point*, %struct.Point** [[THIS_ADDR]], align 8
+// COMP-NEXT:    call void @_ZN5PointC2Ev(%struct.Point* nonnull dereferenceable(8) [[THIS1]]) #[[ATTR4]]
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined.
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2:[0-9]+]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointpLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointpLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointpLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined..1
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointpLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointpLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointpLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined..3
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.4, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointmLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointmLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.4
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointmLERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined..5
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.6, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaNERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaNERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.6
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaNERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined..7
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.8, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointoRERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointoRERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.8
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointoRERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined..9
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.10, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointeOERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP28:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP28]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointeOERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP29]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.10
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointeOERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined..11
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// COMP-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.12, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointaaERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// COMP-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointaaERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// COMP-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// COMP-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.12
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointaaERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// COMP-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// COMP-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp_outlined..13
+// COMP-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED:%.*]], %struct.Point** nonnull align 8 dereferenceable(8) [[POINTS:%.*]]) #[[ATTR2]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[N_ADDR:%.*]] = alloca i32*, align 8
+// COMP-NEXT:    [[RED_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    [[POINTS_ADDR:%.*]] = alloca %struct.Point**, align 8
+// COMP-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[TMP:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[I:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[RED3:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    [[I4:%.*]] = alloca i32, align 4
+// COMP-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
+// COMP-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT]], align 4
+// COMP-NEXT:    [[REF_TMP10:%.*]] = alloca [[STRUCT_POINT]], align 4
+// COMP-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
+// COMP-NEXT:    store i32* [[N]], i32** [[N_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point* [[RED]], %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    store %struct.Point** [[POINTS]], %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
+// COMP-NEXT:    [[TMP1:%.*]] = load %struct.Point*, %struct.Point** [[RED_ADDR]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load %struct.Point**, %struct.Point*** [[POINTS_ADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
+// COMP-NEXT:    store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[SUB:%.*]] = sub i32 [[TMP4]], 0
+// COMP-NEXT:    [[DIV:%.*]] = udiv i32 [[SUB]], 1
+// COMP-NEXT:    [[SUB2:%.*]] = sub i32 [[DIV]], 1
+// COMP-NEXT:    store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 0, i32* [[I]], align 4
+// COMP-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
+// COMP-NEXT:    [[CMP:%.*]] = icmp ult i32 0, [[TMP5]]
+// COMP-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
+// COMP:       omp.precond.then:
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
+// COMP-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
+// COMP-NEXT:    call void @_ZN5PointC1Ev(%struct.Point* nonnull dereferenceable(8) [[RED3]]) #[[ATTR4]]
+// COMP-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
+// COMP-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    [[CMP5:%.*]] = icmp ugt i32 [[TMP9]], [[TMP10]]
+// COMP-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
+// COMP:       cond.true:
+// COMP-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
+// COMP-NEXT:    br label [[COND_END:%.*]]
+// COMP:       cond.false:
+// COMP-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    br label [[COND_END]]
+// COMP:       cond.end:
+// COMP-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
+// COMP-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
+// COMP-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
+// COMP:       omp.inner.for.cond:
+// COMP-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
+// COMP-NEXT:    [[ADD:%.*]] = add i32 [[TMP15]], 1
+// COMP-NEXT:    [[CMP6:%.*]] = icmp ult i32 [[TMP14]], [[ADD]]
+// COMP-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
+// COMP:       omp.inner.for.body:
+// COMP-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[MUL:%.*]] = mul i32 [[TMP16]], 1
+// COMP-NEXT:    [[ADD7:%.*]] = add i32 0, [[MUL]]
+// COMP-NEXT:    store i32 [[ADD7]], i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
+// COMP-NEXT:    [[TMP18:%.*]] = load %struct.Point*, %struct.Point** [[TMP2]], align 8
+// COMP-NEXT:    call void @_Z4workR5PointiPKS_(%struct.Point* nonnull align 4 dereferenceable(8) [[RED3]], i32 [[TMP17]], %struct.Point* [[TMP18]])
+// COMP-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
+// COMP:       omp.body.continue:
+// COMP-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
+// COMP:       omp.inner.for.inc:
+// COMP-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    [[ADD8:%.*]] = add i32 [[TMP19]], 1
+// COMP-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
+// COMP-NEXT:    br label [[OMP_INNER_FOR_COND]]
+// COMP:       omp.inner.for.end:
+// COMP-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
+// COMP:       omp.loop.exit:
+// COMP-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
+// COMP-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
+// COMP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
+// COMP-NEXT:    [[TMP23:%.*]] = bitcast %struct.Point* [[RED3]] to i8*
+// COMP-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
+// COMP-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
+// COMP-NEXT:    [[TMP26:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
+// COMP-NEXT:    [[TMP27:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], i32 1, i64 8, i8* [[TMP26]], void (i8*, i8*)* @.omp.reduction.reduction_func.14, [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    switch i32 [[TMP27]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
+// COMP-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
+// COMP-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
+// COMP-NEXT:    ]
+// COMP:       .omp.reduction.case1:
+// COMP-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointooERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    [[TMP28:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// COMP-NEXT:    store i64 [[CALL]], i64* [[TMP28]], align 4
+// COMP-NEXT:    [[CALL9:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// COMP-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP25]], [8 x i32]* @.gomp_critical_user_.reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.case2:
+// COMP-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
+// COMP-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
+// COMP-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    [[CALL11:%.*]] = call i64 @_ZNK5PointooERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[RED3]])
+// COMP-NEXT:    [[TMP31:%.*]] = bitcast %struct.Point* [[REF_TMP10]] to i64*
+// COMP-NEXT:    store i64 [[CALL11]], i64* [[TMP31]], align 4
+// COMP-NEXT:    [[CALL12:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP1]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP10]])
+// COMP-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB3]], i32 [[TMP30]], [8 x i32]* @.gomp_critical_user_.atomic_reduction.var)
+// COMP-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
+// COMP:       .omp.reduction.default:
+// COMP-NEXT:    br label [[OMP_PRECOND_END]]
+// COMP:       omp.precond.end:
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.14
+// COMP-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR5]] {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
+// COMP-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_POINT:%.*]], align 4
+// COMP-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
+// COMP-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
+// COMP-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
+// COMP-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
+// COMP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
+// COMP-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
+// COMP-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.Point*
+// COMP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
+// COMP-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
+// COMP-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to %struct.Point*
+// COMP-NEXT:    [[CALL:%.*]] = call i64 @_ZNK5PointooERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[TMP8]])
+// COMP-NEXT:    [[TMP12:%.*]] = bitcast %struct.Point* [[REF_TMP]] to i64*
+// COMP-NEXT:    store i64 [[CALL]], i64* [[TMP12]], align 4
+// COMP-NEXT:    [[CALL2:%.*]] = call nonnull align 4 dereferenceable(8) %struct.Point* @_ZN5PointaSERKS_(%struct.Point* nonnull dereferenceable(8) [[TMP11]], %struct.Point* nonnull align 4 dereferenceable(8) [[REF_TMP]])
+// COMP-NEXT:    ret void
+//
+//
+// COMP-LABEL: define {{[^@]+}}@_ZN5PointC2Ev
+// COMP-SAME: (%struct.Point* nonnull dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
+// COMP-NEXT:  entry:
+// COMP-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.Point*, align 8
+// COMP-NEXT:    store %struct.Point* [[THIS]], %struct.Point** [[THIS_ADDR]], align 8
+// COMP-NEXT:    [[THIS1:%.*]] = load %struct.Point*, %struct.Point** [[THIS_ADDR]], align 8
+// COMP-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_POINT:%.*]], %struct.Point* [[THIS1]], i32 0, i32 0
+// COMP-NEXT:    store i32 0, i32* [[X]], align 4
+// COMP-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_POINT]], %struct.Point* [[THIS1]], i32 0, i32 1
+// COMP-NEXT:    store i32 0, i32* [[Y]], align 4
+// COMP-NEXT:    ret void
+//

clang/test/OpenMP/sections_reduction_messages.cpp

#pragma omp sections reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
     foo();
   }
 #pragma omp parallel
 #pragma omp sections reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   {
     foo();
   }
 #pragma omp parallel
-#pragma omp sections reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp sections reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   {
     foo();
   }
 #pragma omp parallel
 #pragma omp sections reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   {
     foo();
   }

clang/test/OpenMP/simd_reduction_messages.cpp

#pragma omp simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/target_parallel_for_reduction_messages.cpp

#pragma omp target parallel for reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp target parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/target_parallel_for_simd_reduction_messages.cpp

#pragma omp target parallel for simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp target parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target parallel for simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/target_parallel_reduction_messages.cpp

 #pragma omp target parallel reduction(- : da) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
 #pragma omp target parallel reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   foo();
 #pragma omp target parallel reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   foo();
 #pragma omp target parallel reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
-#pragma omp target parallel reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target parallel reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   foo();
 #pragma omp target parallel reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   foo();
 #pragma omp target parallel reduction(+ : o) // expected-error {{no viable overloaded '='}}
   foo();
 #pragma omp target parallel private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/target_reduction_messages.cpp

 #pragma omp target reduction(- : da) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
 #pragma omp target reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   foo();
 #pragma omp target reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   foo();
 #pragma omp target reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
-#pragma omp target reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   foo();
 #pragma omp target reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   foo();
 #pragma omp target reduction(+ : o) // expected-error {{no viable overloaded '='}}
   foo();
 #pragma omp parallel private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/target_simd_reduction_messages.cpp

#pragma omp target simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp target simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp target simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/target_teams_distribute_parallel_for_reduction_messages.cpp

 #pragma omp target teams distribute parallel for reduction(- : da, z) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
-#pragma omp target teams distribute parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target teams distribute parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   for (int j=0; j<100; j++) foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/target_teams_distribute_parallel_for_simd_reduction_messages.cpp

 #pragma omp target teams distribute parallel for simd reduction(- : da) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
-#pragma omp target teams distribute parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target teams distribute parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute parallel for simd private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   for (int j=0; j<100; j++) foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/target_teams_distribute_reduction_messages.cpp

 #pragma omp target teams distribute reduction(- : da) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
-#pragma omp target teams distribute reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target teams distribute reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   for (int j=0; j<100; j++) foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/target_teams_distribute_simd_reduction_messages.cpp

 #pragma omp target teams distribute simd reduction(- : da) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
-#pragma omp target teams distribute simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target teams distribute simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target teams distribute simd private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   for (int j=0; j<100; j++) foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/target_teams_reduction_messages.cpp

 #pragma omp target teams reduction(- : da) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
 #pragma omp target teams reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   foo();
 #pragma omp target teams reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   foo();
 #pragma omp target teams reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
-#pragma omp target teams reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp target teams reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   foo();
 #pragma omp target teams reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   foo();
 #pragma omp target teams reduction(+ : o) // expected-error {{no viable overloaded '='}}
   foo();
 #pragma omp target teams private(i), reduction(+ : j), reduction(+ : q) // expected-error 2 {{argument of OpenMP clause 'reduction' must reference the same object in all threads}}
   foo();
 #pragma omp parallel private(k)

clang/test/OpenMP/task_in_reduction_message.cpp

 #pragma omp task in_reduction(- : da) // expected-error {{const-qualified variable cannot be in_reduction}}
   foo();
 #pragma omp task in_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   foo();
 #pragma omp task in_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   foo();
 #pragma omp task in_reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be in_reduction}}
   foo();
-#pragma omp task in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp task in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   foo();
 #pragma omp taskgroup task_reduction(+:k)
 #pragma omp task in_reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   foo();
 #pragma omp task in_reduction(+ : o) // expected-error {{no viable overloaded '='}}
   foo();
 #pragma omp parallel private(k)
 #pragma omp task in_reduction(+ : p), in_reduction(+ : p) // expected-error 2 {{argument of OpenMP clause 'in_reduction' must reference the same object in all threads}}

clang/test/OpenMP/taskgroup_task_reduction_messages.cpp

 #pragma omp taskgroup task_reduction(- : da) // expected-error {{const-qualified variable cannot be task_reduction}}
   foo();
 #pragma omp taskgroup task_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   foo();
 #pragma omp taskgroup task_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   foo();
 #pragma omp taskgroup task_reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be task_reduction}}
   foo();
-#pragma omp taskgroup task_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp taskgroup task_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   foo();
 #pragma omp taskgroup task_reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   foo();
 #pragma omp taskgroup task_reduction(+ : o, z) // expected-error {{no viable overloaded '='}}
   foo();
 #pragma omp parallel private(k)
 #pragma omp taskgroup task_reduction(+ : p), task_reduction(+ : p) // expected-error 2 {{argument of OpenMP clause 'task_reduction' must reference the same object in all threads}}
   foo();

clang/test/OpenMP/taskloop_in_reduction_messages.cpp

#pragma omp taskloop in_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskloop in_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskloop in_reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be in_reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
-#pragma omp taskloop in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp taskloop in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskgroup task_reduction(+:k)
 #pragma omp taskloop in_reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskloop in_reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)

clang/test/OpenMP/taskloop_reduction_messages.cpp

#pragma omp taskloop reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp taskloop reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp taskloop reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/taskloop_simd_in_reduction_messages.cpp

#pragma omp taskloop simd in_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskloop simd in_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskloop simd in_reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be in_reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
-#pragma omp taskloop simd in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp taskloop simd in_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskgroup task_reduction(+:k)
 #pragma omp taskloop simd in_reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
   foo();
 #pragma omp taskloop simd in_reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)

clang/test/OpenMP/taskloop_simd_reduction_messages.cpp

#pragma omp taskloop simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
-#pragma omp taskloop simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp taskloop simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int i = 0; i < 10; ++i)
     foo();
 #pragma omp taskloop simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int i = 0; i < 10; ++i)
     foo();

clang/test/OpenMP/teams_distribute_parallel_for_reduction_messages.cpp

#pragma omp teams distribute parallel for reduction(^ : z, fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
-#pragma omp teams distribute parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp teams distribute parallel for reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target

clang/test/OpenMP/teams_distribute_parallel_for_simd_reduction_messages.cpp

#pragma omp teams distribute parallel for simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
-#pragma omp teams distribute parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp teams distribute parallel for simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute parallel for simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target

clang/test/OpenMP/teams_distribute_reduction_messages.cpp

#pragma omp teams distribute reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
-#pragma omp teams distribute reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp teams distribute reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target

clang/test/OpenMP/teams_distribute_simd_reduction_messages.cpp

#pragma omp teams distribute simd reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute simd reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute simd reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
-#pragma omp teams distribute simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp teams distribute simd reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute simd reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target
 #pragma omp teams distribute simd reduction(+ : o) // expected-error {{no viable overloaded '='}}
   for (int j=0; j<100; j++) foo();
 #pragma omp target

clang/test/OpenMP/teams_reduction_messages.cpp

#pragma omp teams reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
   foo();
 #pragma omp target
 #pragma omp teams reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
   foo();
 #pragma omp target
 #pragma omp teams reduction(&& : S2::S2sc) // expected-error {{const-qualified variable cannot be reduction}}
   foo();
 #pragma omp target
-#pragma omp teams reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{invalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+#pragma omp teams reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
   foo();
 #pragma omp target
 #pragma omp teams reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
   foo();
 #pragma omp target
 #pragma omp teams reduction(+ : z, o) // expected-error {{no viable overloaded '='}}
   foo();
 #pragma omp target