[OPENMP] Improved codegen for implicit/explicit 'barrier' constructs.

[ABataev]

Summary

Add particular methods to OpenMP runtime support module for explicit/implicit sync point emission. Also adds processing of 'nowait' clause on worksharing directives.

[rjmccall]

lib/CodeGen/CGOpenMPRuntime.h

375

I think you should be able to just pass OMPD_barrier for explicit barriers.

Don't have this take a default argument; make the caller think about it and opt in to OMPD_unknown if they want that.

[ABataev]

John, tanks for the review!

lib/CodeGen/CGOpenMPRuntime.h

375

Ok, I'll do this.

lib/CodeGen/CGOpenMPRuntime.h

class CGOpenMPRuntime {
 
   /// \brief Returns corresponding lock object for the specified critical region
   /// name. If the lock object does not exist it is created, otherwise the
   /// reference to the existing copy is returned.
   /// \param CriticalName Name of the critical region.
   ///
   llvm::Value *getCriticalRegionLock(StringRef CriticalName);
 
+  /// \brief Emit barrier call with the specified flags \a Flags.
+  virtual void emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
+                               OpenMPLocationFlags Flags);
+
 public:
   explicit CGOpenMPRuntime(CodeGenModule &CGM);
   virtual ~CGOpenMPRuntime() {}
   virtual void clear();
 
   /// \brief Emits outlined function for the specified OpenMP directive \a D.
   /// This outlined function has type void(*)(kmp_int32 *ThreadID, kmp_int32
   /// BoundID, struct context_vars*).
public:
   virtual void emitSingleRegion(CodeGenFunction &CGF,
                                 const std::function<void()> &SingleOpGen,
                                 SourceLocation Loc,
                                 ArrayRef<const Expr *> CopyprivateVars,
                                 ArrayRef<const Expr *> SrcExprs,
                                 ArrayRef<const Expr *> DstExprs,
                                 ArrayRef<const Expr *> AssignmentOps);
 
-  /// \brief Emits explicit barrier for OpenMP threads.
-  /// \param IsExplicit true, if it is explicitly specified barrier.
+  /// \brief Emit an explicit barrier for OpenMP threads.
   ///
-  virtual void emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
-                               bool IsExplicit = true);
+  virtual void emitExplicitBarrierCall(CodeGenFunction &CGF,
+                                       SourceLocation Loc);
+
+  /// \brief Emit an implicit barrier for OpenMP threads.
+  /// \param Kind Directive for which this implicit barrier call must be
+  /// generated.
+  ///
+  virtual void emitImplicitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
+                                       OpenMPDirectiveKind Kind = OMPD_unknown);
 
   /// \brief Check if the specified \a ScheduleKind is static non-chunked.
   /// This kind of worksharing directive is emitted without outer loop.
   /// \param ScheduleKind Schedule kind specified in the 'schedule' clause.
   /// \param Chunked True if chunk is specified in the clause.
   ///
   virtual bool isStaticNonchunked(OpenMPScheduleClauseKind ScheduleKind,
                                   bool Chunked) const;

lib/CodeGen/CGOpenMPRuntime.cpp

 }
 
 void CGOpenMPRegionInfo::EmitBody(CodeGenFunction &CGF, const Stmt *S) {
   CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
   CGF.EmitOMPPrivateClause(Directive, PrivateScope);
   CGF.EmitOMPFirstprivateClause(Directive, PrivateScope);
   if (PrivateScope.Privatize())
     // Emit implicit barrier to synchronize threads and avoid data races.
-    CGF.CGM.getOpenMPRuntime().emitBarrierCall(CGF, Directive.getLocStart(),
-                                               /*IsExplicit=*/false);
+    CGF.CGM.getOpenMPRuntime().emitImplicitBarrierCall(CGF,
+                                                       Directive.getLocStart());
   CGCapturedStmtInfo::EmitBody(CGF, S);
 }
 
 LValue CGOpenMPTaskOutlinedRegionInfo::getThreadIDVariableLValue(
     CodeGenFunction &CGF) {
   return CGF.MakeNaturalAlignAddrLValue(
       CGF.GetAddrOfLocalVar(getThreadIDVariable()),
       getThreadIDVariable()->getType());
llvm::Value *Args[] = {
         CpyFn,                        // void (*) (void *, void *) <copy_func>
         DidItVal                      // i32 did_it
     };
     CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_copyprivate), Args);
   }
 }
 
 void CGOpenMPRuntime::emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
-                                      bool IsExplicit) {
-  // Build call __kmpc_cancel_barrier(loc, thread_id);
-  auto Flags = static_cast<OpenMPLocationFlags>(
-      OMP_IDENT_KMPC |
-      (IsExplicit ? OMP_IDENT_BARRIER_EXPL : OMP_IDENT_BARRIER_IMPL));
+                                      OpenMPLocationFlags Flags) {
   // Build call __kmpc_cancel_barrier(loc, thread_id);
   // Replace __kmpc_barrier() function by __kmpc_cancel_barrier() because this
   // one provides the same functionality and adds initial support for
   // cancellation constructs introduced in OpenMP 4.0. __kmpc_cancel_barrier()
   // is provided default by the runtime library so it safe to make such
   // replacement.
   llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc, Flags),
                          getThreadID(CGF, Loc)};
   CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_cancel_barrier), Args);
 }
 
+void CGOpenMPRuntime::emitExplicitBarrierCall(CodeGenFunction &CGF,
+                                              SourceLocation Loc) {
+  // Build call __kmpc_cancel_barrier(loc, thread_id);
+  OpenMPLocationFlags Flags =
+      static_cast<OpenMPLocationFlags>(OMP_IDENT_KMPC | OMP_IDENT_BARRIER_EXPL);
+  emitBarrierCall(CGF, Loc, Flags);
+}
+
+void CGOpenMPRuntime::emitImplicitBarrierCall(CodeGenFunction &CGF,
+                                              SourceLocation Loc,
+                                              OpenMPDirectiveKind Kind) {
+  // Build call __kmpc_cancel_barrier(loc, thread_id);
+  OpenMPLocationFlags Flags = OMP_IDENT_KMPC;
+  if (Kind == OMPD_for) {
+    Flags = static_cast<OpenMPLocationFlags>(Flags |
+                                             OMP_IDENT_BARRIER_IMPL_FOR);
+  } else if (Kind == OMPD_sections) {
+    Flags = static_cast<OpenMPLocationFlags>(Flags |
+                                             OMP_IDENT_BARRIER_IMPL_SECTIONS);
+  } else if (Kind == OMPD_single) {
+    Flags =
+        static_cast<OpenMPLocationFlags>(Flags | OMP_IDENT_BARRIER_IMPL_SINGLE);
+  } else {
+    Flags = static_cast<OpenMPLocationFlags>(Flags | OMP_IDENT_BARRIER_IMPL);
+  }
+  emitBarrierCall(CGF, Loc, Flags);
+}
+
 /// \brief Schedule types for 'omp for' loops (these enumerators are taken from
 /// the enum sched_type in kmp.h).
 enum OpenMPSchedType {
   /// \brief Lower bound for default (unordered) versions.
   OMP_sch_lower = 32,
   OMP_sch_static_chunked = 33,
   OMP_sch_static = 34,
   OMP_sch_dynamic_chunked = 35,

lib/CodeGen/CGStmtOpenMP.cpp

 }
 
 void CodeGenFunction::EmitOMPForDirective(const OMPForDirective &S) {
   InlinedOpenMPRegionScopeRAII Region(*this, S);
 
   EmitOMPWorksharingLoop(S);
 
   // Emit an implicit barrier at the end.
-  CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(),
-                                         /*IsExplicit*/ false);
+  if (!S.getSingleClause(OMPC_nowait))
+    CGM.getOpenMPRuntime().emitImplicitBarrierCall(*this, S.getLocStart(),
+                                                   OMPD_for);
 }
 
 void CodeGenFunction::EmitOMPForSimdDirective(const OMPForSimdDirective &) {
   llvm_unreachable("CodeGen for 'omp for simd' is not supported yet.");
 }
 
 static LValue createSectionLVal(CodeGenFunction &CGF, QualType Ty,
                                 const Twine &Name,
CGM.getOpenMPRuntime().emitSingleRegion(*this, [&]() -> void {
       InlinedOpenMPRegionScopeRAII Region(*this, S);
       EmitStmt(Stmt);
       EnsureInsertPoint();
     }, S.getLocStart(), llvm::None, llvm::None, llvm::None, llvm::None);
   }
 
   // Emit an implicit barrier at the end.
   if (!S.getSingleClause(OMPC_nowait))
-    CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(),
-                                           /*IsExplicit=*/false);
+    CGM.getOpenMPRuntime().emitImplicitBarrierCall(
+        *this, S.getLocStart(),
+        (CS && CS->size() > 1) ? OMPD_sections : OMPD_single);
 }
 
 void CodeGenFunction::EmitOMPSectionDirective(const OMPSectionDirective &S) {
   InlinedOpenMPRegionScopeRAII Region(*this, S);
   EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
   EnsureInsertPoint();
 }
 
void CodeGenFunction::EmitOMPSingleDirective(const OMPSingleDirective &S) {
   // Emit code for 'single' region along with 'copyprivate' clauses
   CGM.getOpenMPRuntime().emitSingleRegion(*this, [&]() -> void {
     InlinedOpenMPRegionScopeRAII Region(*this, S);
     EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
     EnsureInsertPoint();
   }, S.getLocStart(), CopyprivateVars, SrcExprs, DstExprs, AssignmentOps);
   // Emit an implicit barrier at the end.
   if (!S.getSingleClause(OMPC_nowait))
-    CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart(),
-                                           /*IsExplicit=*/false);
+    CGM.getOpenMPRuntime().emitImplicitBarrierCall(*this, S.getLocStart(),
+                                                   OMPD_single);
 }
 
 void CodeGenFunction::EmitOMPMasterDirective(const OMPMasterDirective &S) {
   CGM.getOpenMPRuntime().emitMasterRegion(*this, [&]() -> void {
     InlinedOpenMPRegionScopeRAII Region(*this, S);
     EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
     EnsureInsertPoint();
   }, S.getLocStart());

 }
 
 void CodeGenFunction::EmitOMPTaskyieldDirective(
     const OMPTaskyieldDirective &S) {
   CGM.getOpenMPRuntime().emitTaskyieldCall(*this, S.getLocStart());
 }
 
 void CodeGenFunction::EmitOMPBarrierDirective(const OMPBarrierDirective &S) {
-  CGM.getOpenMPRuntime().emitBarrierCall(*this, S.getLocStart());
+  CGM.getOpenMPRuntime().emitExplicitBarrierCall(*this, S.getLocStart());
 }
 
 void CodeGenFunction::EmitOMPTaskwaitDirective(const OMPTaskwaitDirective &) {
   llvm_unreachable("CodeGen for 'omp taskwait' is not supported yet.");
 }
 
 void CodeGenFunction::EmitOMPFlushDirective(const OMPFlushDirective &S) {
   CGM.getOpenMPRuntime().emitFlush(*this, [&]() -> ArrayRef<const Expr *> {

test/OpenMP/for_codegen.cpp

 // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s
 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp=libiomp5 -fexceptions -fcxx-exceptions -gline-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=TERM_DEBUG
 //
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
 
 // CHECK: [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
+// CHECK: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8*
 // CHECK-LABEL: define {{.*void}} @{{.*}}without_schedule_clause{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void without_schedule_clause(float *a, float *b, float *c, float *d) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
-  #pragma omp for
+  #pragma omp for nowait
 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1)
 // UB = min(UB, GlobalUB)
 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
 // CHECK-NEXT: [[UBCMP:%.+]] = icmp sgt i32 [[UB]], 4571423
 // CHECK-NEXT: br i1 [[UBCMP]], label [[UB_TRUE:%[^,]+]], label [[UB_FALSE:%[^,]+]]
 // CHECK: [[UBRESULT:%.+]] = phi i32 [ 4571423, [[UB_TRUE]] ], [ [[UBVAL:%[^,]+]], [[UB_FALSE]] ]
 // CHECK-NEXT: store i32 [[UBRESULT]], i32* [[OMP_UB]]
 // CHECK-NEXT: [[LB:%.+]] = load i32, i32* [[OMP_LB]]
// CHECK: store float [[RESULT:%.+]], float* {{%.+}}
     a[i] = b[i] * c[i] * d[i];
 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
 // CHECK-NEXT: br label %{{.+}}
   }
 // CHECK: [[LOOP1_END]]
 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
-// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
+// CHECK-NOT: __kmpc_cancel_barrier
 // CHECK: ret void
 }
 
 // CHECK-LABEL: define {{.*void}} @{{.*}}static_not_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void static_not_chunked(float *a, float *b, float *c, float *d) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
   #pragma omp for schedule(static)
 // CHECK: call void @__kmpc_for_static_init_4([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 34, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 1)
// CHECK: store float [[RESULT:%.+]], float* {{%.+}}
     a[i] = b[i] * c[i] * d[i];
 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
 // CHECK-NEXT: br label %{{.+}}
   }
 // CHECK: [[LOOP1_END]]
 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
-// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
+// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
 
 // CHECK-LABEL: define {{.*void}} @{{.*}}static_chunked{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void static_chunked(float *a, float *b, float *c, float *d) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
   #pragma omp for schedule(static, 5)
 // CHECK: call void @__kmpc_for_static_init_4u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 33, i32* [[IS_LAST:%[^,]+]], i32* [[OMP_LB:%[^,]+]], i32* [[OMP_UB:%[^,]+]], i32* [[OMP_ST:%[^,]+]], i32 1, i32 5)

 // CHECK-NEXT: store i32 [[ADD_LB]], i32* [[OMP_LB]]
 // CHECK-NEXT: [[UB:%.+]] = load i32, i32* [[OMP_UB]]
 // CHECK-NEXT: [[ST:%.+]] = load i32, i32* [[OMP_ST]]
 // CHECK-NEXT: [[ADD_UB:%.+]] = add i32 [[UB]], [[ST]]
 // CHECK-NEXT: store i32 [[ADD_UB]], i32* [[OMP_UB]]
 
 // CHECK: [[O_LOOP1_END]]
 // CHECK: call void @__kmpc_for_static_fini([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
-// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
+// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
 
 // CHECK-LABEL: define {{.*void}} @{{.*}}dynamic1{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void dynamic1(float *a, float *b, float *c, float *d) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
   #pragma omp for schedule(dynamic)
 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 35, i64 0, i64 16908287, i64 1, i64 1)
// CHECK: store float [[RESULT:%.+]], float* {{%.+}}
     a[i] = b[i] * c[i] * d[i];
 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
 // CHECK-NEXT: br label %{{.+}}
   }
 // CHECK: [[LOOP1_END]]
 // CHECK: [[O_LOOP1_END]]
-// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
+// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
 
 // CHECK-LABEL: define {{.*void}} @{{.*}}guided7{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void guided7(float *a, float *b, float *c, float *d) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
   #pragma omp for schedule(guided, 7)
 // CHECK: call void @__kmpc_dispatch_init_8u([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
// CHECK: store float [[RESULT:%.+]], float* {{%.+}}
     a[i] = b[i] * c[i] * d[i];
 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
 // CHECK-NEXT: [[ADD1_2:%.+]] = add i64 [[IV1_2]], 1
 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
 // CHECK-NEXT: br label %{{.+}}
   }
 // CHECK: [[LOOP1_END]]
 // CHECK: [[O_LOOP1_END]]
-// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
+// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
 
 // CHECK-LABEL: define {{.*void}} @{{.*}}test_auto{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void test_auto(float *a, float *b, float *c, float *d) {
   unsigned int x = 0;
   unsigned int y = 0;
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
// CHECK: store float [[RESULT:%.+]], float* {{%.+}}
     a[i] = b[i] * c[i] * d[i];
 // CHECK: [[IV1_2:%.+]] = load i64, i64* [[OMP_IV]]{{.*}}
 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i64 [[IV1_2]], 1
 // CHECK-NEXT: store i64 [[ADD1_2]], i64* [[OMP_IV]]
 // CHECK-NEXT: br label %{{.+}}
   }
 // CHECK: [[LOOP1_END]]
 // CHECK: [[O_LOOP1_END]]
-// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
+// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
 
 // CHECK-LABEL: define {{.*void}} @{{.*}}runtime{{.*}}(float* {{.+}}, float* {{.+}}, float* {{.+}}, float* {{.+}})
 void runtime(float *a, float *b, float *c, float *d) {
   int x = 0;
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num([[IDENT_T_TY]]* [[DEFAULT_LOC:[@%].+]])
   #pragma omp for collapse(2) schedule(runtime)
// CHECK: store float [[RESULT:%.+]], float* {{%.+}}
     a[i] = b[i] * c[i] * d[i];
 // CHECK: [[IV1_2:%.+]] = load i32, i32* [[OMP_IV]]{{.*}}
 // CHECK-NEXT: [[ADD1_2:%.+]] = add nsw i32 [[IV1_2]], 1
 // CHECK-NEXT: store i32 [[ADD1_2]], i32* [[OMP_IV]]
 // CHECK-NEXT: br label %{{.+}}
   }
 // CHECK: [[LOOP1_END]]
 // CHECK: [[O_LOOP1_END]]
-// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[DEFAULT_LOC_BARRIER:[@%].+]], i32 [[GTID]])
+// CHECK: call {{.+}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_LOC]], i32 [[GTID]])
 // CHECK: ret void
 }
 
 // TERM_DEBUG-LABEL: foo
 int foo() {return 0;};
 
 // TERM_DEBUG-LABEL: parallel_for
 void parallel_for(float *a) {

test/OpenMP/sections_codegen.cpp

 // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -emit-llvm -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -o - %s | FileCheck %s
 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -fexceptions -fcxx-exceptions -triple x86_64-unknown-unknown -emit-pch -o %t %s
 // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -include-pch %t -fsyntax-only -verify %s -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-llvm -o - | FileCheck %s
 // expected-no-diagnostics
 
 #ifndef HEADER
 #define HEADER
-
+// CHECK: [[IMPLICIT_BARRIER_SECTIONS_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 194, i32 0, i32 0, i8*
+// CHECK: [[IMPLICIT_BARRIER_SINGLE_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 322, i32 0, i32 0, i8*
 // CHECK-LABEL: foo
 void foo() {};
 // CHECK-LABEL: bar
 void bar() {};
 
 template <class T>
 T tmain() {
 #pragma omp parallel
 #pragma omp sections
   {
     foo();
   }
   return T();
 }
 
 // CHECK-LABEL: @main
 int main() {
   float l = 0.0; // Used as a base point in checks.
 // CHECK: [[GTID:%.+]] = call{{.*}} i32 @__kmpc_global_thread_num({{.*}})
 // CHECK: store float
-#pragma omp sections nowait
+#pragma omp sections
   {
 // CHECK:      store i32 0, i32* [[LB_PTR:%.+]],
 // CHECK:      store i32 1, i32* [[UB_PTR:%.+]],
 // CHECK:      call void @__kmpc_for_static_init_4(%{{.+}}* @{{.+}}, i32 [[GTID]], i32 34, i32* [[IS_LAST_PTR:%.+]], i32* [[LB_PTR]], i32* [[UB_PTR]], i32* [[STRIDE_PTR:%.+]], i32 1, i32 1)
 // <<UB = min(UB, GlobalUB);>>
 // CHECK:      [[UB:%.+]] = load i32, i32* [[UB_PTR]]
 // CHECK:      [[CMP:%.+]] = icmp slt i32 [[UB]], 1
 // CHECK:      [[MIN_UB_GLOBALUB:%.+]] = select i1 [[CMP]], i32 [[UB]], i32 1

 // <<++IV;>>
 // CHECK:      [[IV:%.+]] = load i32, i32* [[IV_PTR]]
 // CHECK-NEXT: [[INC:%.+]] = add nsw i32 [[IV]], 1
 // CHECK-NEXT: store i32 [[INC]], i32* [[IV_PTR]]
 // CHECK-NEXT: br label %[[INNER_FOR_COND]]
 // CHECK:      [[INNER_LOOP_END]]
   }
 // CHECK:      call void @__kmpc_for_static_fini(%{{.+}}* @{{.+}}, i32 [[GTID]])
+// CHECK:      call i32 @__kmpc_cancel_barrier(%{{.+}}* [[IMPLICIT_BARRIER_SECTIONS_LOC]],
+#pragma omp sections nowait
+  {
+    foo();
+#pragma omp section
+    bar();
+  }
 // CHECK-NOT:  __kmpc_cancel_barrier
   return tmain<int>();
 }
 
 // CHECK-LABEL: tmain
 // CHECK:       call void {{.*}} @__kmpc_fork_call(
 // CHECK-NOT:   __kmpc_global_thread_num
 // CHECK:       [[RES:%.+]] = call i32 @__kmpc_single(
 // CHECK-NEXT:  [[BOOLRES:%.+]] = icmp ne i32 [[RES]], 0
 // CHECK-NEXT:  br i1 [[BOOLRES]], label %[[THEN:.+]], label %[[END:.+]]
 // CHECK:       [[THEN]]
 // CHECK-NEXT:  invoke void @{{.*}}foo{{.*}}()
 // CHECK-NEXT:  unwind label %[[TERM_LPAD:.+]]
 // CHECK:       call void @__kmpc_end_single(
 // CHECK-NEXT:  br label %[[END]]
 // CHECK:       [[END]]
-// CHECK-NEXT:  call i32 @__kmpc_cancel_barrier(
+// CHECK-NEXT:  call i32 @__kmpc_cancel_barrier(%{{.+}}* [[IMPLICIT_BARRIER_SINGLE_LOC]],
 // CHECK-NEXT:  ret
 // CHECK:       [[TERM_LPAD]]
 // CHECK:       call void @__clang_call_terminate(i8*
 // CHECK-NEXT:  unreachable
 
 #endif

test/OpenMP/single_codegen.cpp

public:
   TestClass() : a(0) {}
   TestClass(const TestClass &C) : a(C.a) {}
   TestClass &operator=(const TestClass &) { return *this;}
   ~TestClass(){};
 };
 
 // CHECK-DAG:   [[TEST_CLASS_TY:%.+]] = type { i{{[0-9]+}} }
 // CHECK:       [[IDENT_T_TY:%.+]] = type { i32, i32, i32, i32, i8* }
+// CHECK:       [[IMPLICIT_BARRIER_SINGLE_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 322, i32 0, i32 0, i8*
 
 // CHECK:       define void [[FOO:@.+]]()
 
 TestClass tc;
 #pragma omp threadprivate(tc)
 
 void foo() {}
 

 // CHECK:       [[RES:%.+]] = call i32 @__kmpc_single([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
 // CHECK-NEXT:  [[IS_SINGLE:%.+]] = icmp ne i32 [[RES]], 0
 // CHECK-NEXT:  br i1 [[IS_SINGLE]], label {{%?}}[[THEN:.+]], label {{%?}}[[EXIT:.+]]
 // CHECK:       [[THEN]]
 // CHECK-NEXT:  store i8 2, i8* [[A_ADDR]]
 // CHECK-NEXT:  call void @__kmpc_end_single([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
 // CHECK-NEXT:  br label {{%?}}[[EXIT]]
 // CHECK:       [[EXIT]]
+// CHECK-NOT:   __kmpc_cancel_barrier
 #pragma omp single nowait
   a = 2;
 // CHECK:       [[RES:%.+]] = call i32 @__kmpc_single([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]])
 // CHECK-NEXT:  [[IS_SINGLE:%.+]] = icmp ne i32 [[RES]], 0
 // CHECK-NEXT:  br i1 [[IS_SINGLE]], label {{%?}}[[THEN:.+]], label {{%?}}[[EXIT:.+]]
 // CHECK:       [[THEN]]
 // CHECK-NEXT:  invoke void [[FOO]]()
 // CHECK:       to label {{%?}}[[CONT:.+]] unwind

 // CHECK:       [[TC_PTR_REF:%.+]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[COPY_LIST]], i{{[0-9]+}} 0, i{{[0-9]+}} 2
 // CHECK:       [[TC_THREADPRIVATE_ADDR_VOID_PTR:%.+]] = call{{.*}} i8* @__kmpc_threadprivate_cached
 // CHECK:       [[TC_THREADPRIVATE_ADDR:%.+]] = bitcast i8* [[TC_THREADPRIVATE_ADDR_VOID_PTR]] to [[TEST_CLASS_TY]]*
 // CHECK:       [[TC_PTR_REF_VOID_PTR:%.+]] = bitcast [[TEST_CLASS_TY]]* [[TC_THREADPRIVATE_ADDR]] to i8*
 // CHECK:       store i8* [[TC_PTR_REF_VOID_PTR]], i8** [[TC_PTR_REF]],
 // CHECK:       [[COPY_LIST_VOID_PTR:%.+]] = bitcast [3 x i8*]* [[COPY_LIST]] to i8*
 // CHECK:       [[DID_IT_VAL:%.+]] = load i32, i32* [[DID_IT]],
 // CHECK:       call void @__kmpc_copyprivate([[IDENT_T_TY]]* [[DEFAULT_LOC]], i32 [[GTID]], i32 24, i8* [[COPY_LIST_VOID_PTR]], void (i8*, i8*)* [[COPY_FUNC:@.+]], i32 [[DID_IT_VAL]])
-// CHECK:       call{{.*}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* {{@.+}}, i32 [[GTID]])
+// CHECK:       call{{.*}} @__kmpc_cancel_barrier([[IDENT_T_TY]]* [[IMPLICIT_BARRIER_SINGLE_LOC]], i32 [[GTID]])
 #pragma omp single copyprivate(a, c, tc)
   foo();
 // CHECK-NOT:   call i32 @__kmpc_single
 // CHECK-NOT:   call void @__kmpc_end_single
   return a;
 }
 
 // CHECK: void [[COPY_FUNC]](i8*, i8*)