diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -518,8 +518,10 @@
   F->setDoesNotRecurse();
 
   // Always inline the outlined function if optimizations are enabled.
-  if (CGM.getCodeGenOpts().OptimizationLevel != 0)
+  if (CGM.getCodeGenOpts().OptimizationLevel != 0) {
+    F->removeFnAttr(llvm::Attribute::NoInline);
     F->addFnAttr(llvm::Attribute::AlwaysInline);
+  }
 
   // Generate the function.
   CGF.StartFunction(CD, Ctx.VoidTy, F, FuncInfo, TargetArgs,
@@ -5364,8 +5366,6 @@
   CGF.CapturedStmtInfo = &CapStmtInfo;
   llvm::Function *Fn = CGF.GenerateOpenMPCapturedStmtFunction(*S, Loc);
   Fn->setDoesNotRecurse();
-  if (CGM.getCodeGenOpts().OptimizationLevel != 0)
-    Fn->addFnAttr(llvm::Attribute::AlwaysInline);
   return Fn;
 }
 
diff --git a/clang/test/OpenMP/parallel_if_codegen_PR51349.cpp b/clang/test/OpenMP/parallel_if_codegen_PR51349.cpp
--- a/clang/test/OpenMP/parallel_if_codegen_PR51349.cpp
+++ b/clang/test/OpenMP/parallel_if_codegen_PR51349.cpp
@@ -1,5 +1,6 @@
 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --check-attributes --include-generated-funcs
 // RUN: %clang_cc1 -x c++ -O1 -fopenmp-version=45 -disable-llvm-optzns -verify -fopenmp -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK
+// RUN: %clang_cc1 -x c++ -O1 -fopenmp-version=45 -disable-llvm-optzns -verify -fopenmp -triple x86_64-unknown-linux -emit-llvm -fno-inline %s -o - | FileCheck %s --check-prefix=CHECK-NOINLINE
 // expected-no-diagnostics
 
 #ifndef HEADER
@@ -8,6 +9,8 @@
 void foo() {
 #pragma omp parallel if(0)
   ;
+#pragma omp parallel
+  ;
 }
 
 #endif
@@ -23,6 +26,7 @@
 // CHECK-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
 // CHECK-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]]) #[[ATTR2:[0-9]+]]
 // CHECK-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
+// CHECK-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
 // CHECK-NEXT:    ret void
 //
 //
@@ -36,3 +40,52 @@
 // CHECK-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8, !tbaa [[TBAA7]]
 // CHECK-NEXT:    ret void
 //
+//
+// CHECK: Function Attrs: alwaysinline norecurse nounwind
+// CHECK-LABEL: define {{[^@]+}}@.omp_outlined..1
+// CHECK-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8, !tbaa [[TBAA7]]
+// CHECK-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8, !tbaa [[TBAA7]]
+// CHECK-NEXT:    ret void
+//
+//
+// CHECK-NOINLINE: Function Attrs: mustprogress noinline nounwind
+// CHECK-NOINLINE-LABEL: define {{[^@]+}}@_Z3foov
+// CHECK-NOINLINE-SAME: () #[[ATTR0:[0-9]+]] {
+// CHECK-NOINLINE-NEXT:  entry:
+// CHECK-NOINLINE-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
+// CHECK-NOINLINE-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
+// CHECK-NOINLINE-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
+// CHECK-NOINLINE-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
+// CHECK-NOINLINE-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA3:![0-9]+]]
+// CHECK-NOINLINE-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
+// CHECK-NOINLINE-NEXT:    call void @.omp_outlined.(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]]) #[[ATTR2:[0-9]+]]
+// CHECK-NOINLINE-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
+// CHECK-NOINLINE-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*))
+// CHECK-NOINLINE-NEXT:    ret void
+//
+//
+// CHECK-NOINLINE: Function Attrs: noinline norecurse nounwind
+// CHECK-NOINLINE-LABEL: define {{[^@]+}}@.omp_outlined.
+// CHECK-NOINLINE-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
+// CHECK-NOINLINE-NEXT:  entry:
+// CHECK-NOINLINE-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-NOINLINE-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-NOINLINE-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8, !tbaa [[TBAA7:![0-9]+]]
+// CHECK-NOINLINE-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8, !tbaa [[TBAA7]]
+// CHECK-NOINLINE-NEXT:    ret void
+//
+//
+// CHECK-NOINLINE: Function Attrs: alwaysinline norecurse nounwind
+// CHECK-NOINLINE-LABEL: define {{[^@]+}}@.omp_outlined..1
+// CHECK-NOINLINE-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
+// CHECK-NOINLINE-NEXT:  entry:
+// CHECK-NOINLINE-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-NOINLINE-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
+// CHECK-NOINLINE-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8, !tbaa [[TBAA7]]
+// CHECK-NOINLINE-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8, !tbaa [[TBAA7]]
+// CHECK-NOINLINE-NEXT:    ret void
+//