test/OpenMP/teams_distribute_parallel_for_simd_firstprivate_codegen.cpp - llvm-project/clang - Git at Google

 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK1
 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK1
 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK3
 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK3

 // RUN: %clang_cc1 -DCHECK -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK5
 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK5
 // RUN: %clang_cc1 -DCHECK -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK7
 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK7

 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK9
 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++  -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK9

 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK11
 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s -Wno-openmp-mapping
 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++  -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK11

 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER

 struct St {
   int a, b;
   St() : a(0), b(0) {}
   St(const St &st) : a(st.a + st.b), b(0) {}
   ~St() {}
 };

 volatile int g = 1212;
 volatile int &g1 = g;

 template <class T>
 struct S {
   T f;
   S(T a) : f(a + g) {}
   S() : f(g) {}
   S(const S &s, St t = St()) : f(s.f + t.a) {}
   operator T() { return T(); }
   ~S() {}
 };


 template <typename T>
 T tmain() {
   S<T> test;
   T t_var = T();
   T vec[] = {1, 2};
   S<T> s_arr[] = {1, 2};
   S<T> &var = test;
 #pragma omp target
 #pragma omp teams distribute parallel for simd firstprivate(t_var, vec, s_arr, var)
   for (int i = 0; i < 2; ++i) {
     vec[i] = t_var;
     s_arr[i] = var;
   }
   return T();
 }

 S<float> test;
 int t_var = 333;
 int vec[] = {1, 2};
 S<float> s_arr[] = {1, 2};
 S<float> var(3);

 int main() {
   static int sivar;
 #ifdef LAMBDA
   [&]() {
 #pragma omp target
 #pragma omp teams distribute parallel for simd firstprivate(g, g1, sivar)
   for (int i = 0; i < 2; ++i) {

     // Skip global and bound tid vars
     // skip loop vars
     g = 1;
     g1 = 1;
     sivar = 2;

     // Skip global and bound tid vars, and prev lb and ub vars
     // skip loop vars

     // use of private vars
     [&]() {
       g = 2;
       g1 = 2;
       sivar = 4;

     }();
   }
   }();
   return 0;


 #else
 #pragma omp target
 #pragma omp teams distribute parallel for simd firstprivate(t_var, vec, s_arr, var, sivar)
   for (int i = 0; i < 2; ++i) {
     vec[i] = t_var;
     s_arr[i] = var;
     sivar += i;
   }
   return tmain<int>();
 #endif
 }


 // Skip global and bound tid vars
 // Skip temp vars for loop

 // param copy

 // T_VAR and SIVAR

 // preparation vars

 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2

 // firstprivate(s_arr)

 // firstprivate(var)


 // Skip global and bound tid vars, and prev lb ub vars
 // Skip temp vars for loop

 // param copy

 // T_VAR and SIVAR

 // preparation vars

 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2

 // firstprivate(s_arr)

 // firstprivate(var)


 // Skip global and bound tid vars
 // Skip temp vars for loop

 // param copy

 // T_VAR and preparation variables

 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2

 // firstprivate(s_arr)

 // firstprivate(var)


 // Skip global and bound tid vars
 // Skip temp vars for loop

 // param copy

 // T_VAR and preparation variables

 // firstprivate vec(vec): copy from *_addr into priv1 and then from priv1 into priv2

 // firstprivate(s_arr)

 // firstprivate(var)


 #endif
 // CHECK1-LABEL: define {{[^@]+}}@__cxx_global_var_init
 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) @test)
 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @test, ptr @__dso_handle) #[[ATTR2:[0-9]+]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
 // CHECK1-NEXT:    store float [[CONV]], ptr [[F]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__cxx_global_var_init.1
 // CHECK1-SAME: () #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @s_arr, float noundef 1.000000e+00)
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 1), float noundef 2.000000e+00)
 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @s_arr
 // CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK1:       arraydestroy.done1:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 // CHECK1-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
 // CHECK1-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@__cxx_global_var_init.2
 // CHECK1-SAME: () #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @var, float noundef 3.000000e+00)
 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @var, ptr @__dso_handle) #[[ATTR2]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@main
 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr @t_var, align 4
 // CHECK1-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr @_ZZ4mainE5sivar, align 4
 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[SIVAR_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP1]], ptr [[TMP4]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP1]], ptr [[TMP5]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 // CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 // CHECK1-NEXT:    store ptr @vec, ptr [[TMP7]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 // CHECK1-NEXT:    store ptr @vec, ptr [[TMP8]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
 // CHECK1-NEXT:    store ptr null, ptr [[TMP9]], align 8
 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr @s_arr, ptr [[TMP10]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr @s_arr, ptr [[TMP11]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr @var, ptr [[TMP13]], align 8
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr @var, ptr [[TMP14]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
 // CHECK1-NEXT:    store ptr null, ptr [[TMP15]], align 8
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 // CHECK1-NEXT:    store i64 [[TMP3]], ptr [[TMP16]], align 8
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 // CHECK1-NEXT:    store i64 [[TMP3]], ptr [[TMP17]], align 8
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
 // CHECK1-NEXT:    store ptr null, ptr [[TMP18]], align 8
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK1-NEXT:    store i32 3, ptr [[TMP21]], align 4
 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK1-NEXT:    store i32 5, ptr [[TMP22]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 8
 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr [[TMP20]], ptr [[TMP24]], align 8
 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP25]], align 8
 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP26]], align 8
 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK1-NEXT:    store ptr null, ptr [[TMP27]], align 8
 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK1-NEXT:    store ptr null, ptr [[TMP28]], align 8
 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK1-NEXT:    store i64 2, ptr [[TMP29]], align 8
 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK1-NEXT:    store i64 0, ptr [[TMP30]], align 8
 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP32]], align 4
 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP33]], align 4
 // CHECK1-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
 // CHECK1-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101(i64 [[TMP1]], ptr @vec, ptr @s_arr, ptr @var, i64 [[TMP3]]) #[[ATTR2]]
 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK1:       omp_offload.cont:
 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
 // CHECK1-NEXT:    ret i32 [[CALL]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101
 // CHECK1-SAME: (i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[SIVAR_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined, ptr [[TMP0]], i64 [[TMP4]], ptr [[TMP1]], ptr [[TMP2]], i64 [[TMP6]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR4]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[VEC1:%.*]] = alloca [2 x i32], align 4
 // CHECK1-NEXT:    [[S_ARR2:%.*]] = alloca [2 x %struct.S], align 4
 // CHECK1-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK1-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S:%.*]], align 4
 // CHECK1-NEXT:    [[AGG_TMP5:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC1]], ptr align 4 [[TMP0]], i64 8, i1 false)
 // CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK1:       omp.arraycpy.body:
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK1:       omp.arraycpy.done3:
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP5]])
 // CHECK1-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR4]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP2]], ptr noundef [[AGG_TMP5]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP5]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP5:![0-9]+]]
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP13]] to i64
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    store i32 [[TMP15]], ptr [[T_VAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    store i32 [[TMP17]], ptr [[SIVAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined.omp_outlined, i64 [[TMP12]], i64 [[TMP14]], ptr [[VEC1]], i64 [[TMP16]], ptr [[S_ARR2]], ptr [[VAR4]], i64 [[TMP18]]), !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP5]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK1-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP25]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
 // CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK1:       arraydestroy.done8:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN2StC1Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN2StC2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC1ERKS0_2St
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN1SIfEC2ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0]], ptr noundef [[T]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN2StD1Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN2StD2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]]) #[[ATTR2]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR4]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
 // CHECK1-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S], align 4
 // CHECK1-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK1-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S:%.*]], align 4
 // CHECK1-NEXT:    [[AGG_TMP6:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP3]] to i32
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP4]] to i32
 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC2]], ptr align 4 [[TMP0]], i64 8, i1 false)
 // CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK1:       omp.arraycpy.body:
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK1:       omp.arraycpy.done4:
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]])
 // CHECK1-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP2]], ptr noundef [[AGG_TMP6]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 1
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
 // CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC2]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 [[TMP14]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR3]], i64 0, i64 [[IDXPROM8]]
 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX9]], ptr align 4 [[VAR5]], i64 4, i1 false), !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP18]], [[TMP17]]
 // CHECK1-NEXT:    store i32 [[ADD10]], ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK1-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK1-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAY_BEGIN12:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN12]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP24]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN12]]
 // CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE13:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK1:       arraydestroy.done13:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
 // CHECK1-SAME: () #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK1-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
 // CHECK1-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK1-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 8
 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 8
 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK1-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
 // CHECK1-NEXT:    store i32 0, ptr [[T_VAR]], align 4
 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i64 8, i1 false)
 // CHECK1-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 0
 // CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef signext 1)
 // CHECK1-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i64 1
 // CHECK1-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef signext 2)
 // CHECK1-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[T_VAR_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP2]], ptr [[TMP6]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    store i64 [[TMP2]], ptr [[TMP7]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 // CHECK1-NEXT:    store ptr null, ptr [[TMP8]], align 8
 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP9]], align 8
 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP10]], align 8
 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[TMP12]], align 8
 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[TMP13]], align 8
 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
 // CHECK1-NEXT:    store ptr null, ptr [[TMP14]], align 8
 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP15]], align 8
 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr [[TMP5]], ptr [[TMP16]], align 8
 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
 // CHECK1-NEXT:    store ptr null, ptr [[TMP17]], align 8
 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK1-NEXT:    store i32 3, ptr [[TMP20]], align 4
 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK1-NEXT:    store i32 4, ptr [[TMP21]], align 4
 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK1-NEXT:    store ptr [[TMP18]], ptr [[TMP22]], align 8
 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK1-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 8
 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK1-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP24]], align 8
 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK1-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP25]], align 8
 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK1-NEXT:    store ptr null, ptr [[TMP26]], align 8
 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK1-NEXT:    store ptr null, ptr [[TMP27]], align 8
 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK1-NEXT:    store i64 2, ptr [[TMP28]], align 8
 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK1-NEXT:    store i64 0, ptr [[TMP29]], align 8
 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP30]], align 4
 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK1-NEXT:    store i32 0, ptr [[TMP32]], align 4
 // CHECK1-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.region_id, ptr [[KERNEL_ARGS]])
 // CHECK1-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
 // CHECK1-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK1:       omp_offload.failed:
 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56(i64 [[TMP2]], ptr [[VEC]], ptr [[S_ARR]], ptr [[TMP3]]) #[[ATTR2]]
 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK1:       omp_offload.cont:
 // CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP35]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
 // CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK1:       arraydestroy.done2:
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[TMP36:%.*]] = load i32, ptr [[RETVAL]], align 4
 // CHECK1-NEXT:    ret i32 [[TMP36]]
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN2StC2Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK1-NEXT:    store i32 0, ptr [[A]], align 4
 // CHECK1-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
 // CHECK1-NEXT:    store i32 0, ptr [[B]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIfEC2ERKS0_2St
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    [[F2:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP1:%.*]] = load float, ptr [[F2]], align 4
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
 // CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP2]] to float
 // CHECK1-NEXT:    [[ADD:%.*]] = fadd float [[TMP1]], [[CONV]]
 // CHECK1-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN2StD2Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef signext [[TMP0]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56
 // CHECK1-SAME: (i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 8
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
 // CHECK1-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined, ptr [[TMP0]], i64 [[TMP4]], ptr [[TMP1]], ptr [[TMP5]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
 // CHECK1-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK1-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK1-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK1-NEXT:    [[AGG_TMP6:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK1-NEXT:    [[_TMP7:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC2]], ptr align 4 [[TMP0]], i64 8, i1 false)
 // CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK1:       omp.arraycpy.body:
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK1:       omp.arraycpy.done4:
 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]])
 // CHECK1-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP4]], ptr noundef [[AGG_TMP6]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]]) #[[ATTR2]]
 // CHECK1-NEXT:    store ptr [[VAR5]], ptr [[_TMP7]], align 8
 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14:![0-9]+]]
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK1-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP14]] to i64
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    store i32 [[TMP16]], ptr [[T_VAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[_TMP7]], align 8, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined, i64 [[TMP13]], i64 [[TMP15]], ptr [[VEC2]], i64 [[TMP17]], ptr [[S_ARR3]], ptr [[TMP18]]), !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP22]])
 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
 // CHECK1-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAY_BEGIN9:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN9]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP25]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN9]]
 // CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE10:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK1:       arraydestroy.done10:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC1ERKS0_2St
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN1SIiEC2ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0]], ptr noundef [[T]])
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined
 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK1-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    [[VEC3:%.*]] = alloca [2 x i32], align 4
 // CHECK1-NEXT:    [[S_ARR4:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK1-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK1-NEXT:    [[VAR6:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK1-NEXT:    [[AGG_TMP7:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK1-NEXT:    [[_TMP8:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 8
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP3]] to i32
 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP4]] to i32
 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC3]], ptr align 4 [[TMP0]], i64 8, i1 false)
 // CHECK1-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK1:       omp.arraycpy.body:
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK1:       omp.arraycpy.done5:
 // CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK1-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP7]])
 // CHECK1-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR6]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP6]], ptr noundef [[AGG_TMP7]])
 // CHECK1-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP7]]) #[[ATTR2]]
 // CHECK1-NEXT:    store ptr [[VAR6]], ptr [[_TMP8]], align 8
 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1
 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK1:       cond.true:
 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 // CHECK1:       cond.false:
 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    br label [[COND_END]]
 // CHECK1:       cond.end:
 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK1-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK1:       omp.inner.for.cond:
 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17:![0-9]+]]
 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK1-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK1:       omp.inner.for.cond.cleanup:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK1:       omp.inner.for.body:
 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC3]], i64 0, i64 [[IDXPROM]]
 // CHECK1-NEXT:    store i32 [[TMP15]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[_TMP8]], align 8, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[IDXPROM10:%.*]] = sext i32 [[TMP18]] to i64
 // CHECK1-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i64 0, i64 [[IDXPROM10]]
 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX11]], ptr align 4 [[TMP17]], i64 4, i1 false), !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK1:       omp.body.continue:
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK1:       omp.inner.for.inc:
 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK1-NEXT:    store i32 [[ADD12]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP17]]
 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
 // CHECK1:       omp.inner.for.end:
 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK1:       omp.loop.exit:
 // CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK1-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK1:       .omp.final.then:
 // CHECK1-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK1:       .omp.final.done:
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR6]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAY_BEGIN13:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN13]], i64 2
 // CHECK1-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK1:       arraydestroy.body:
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP24]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK1-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK1-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK1-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN13]]
 // CHECK1-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE14:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK1:       arraydestroy.done14:
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK1-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK1-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP1]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiEC2ERKS0_2St
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 8
 // CHECK1-NEXT:    [[F2:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[TMP0]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[F2]], align 4
 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[TMP2]]
 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK1-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_teams_distribute_parallel_for_simd_firstprivate_codegen.cpp
 // CHECK1-SAME: () #[[ATTR0]] {
 // CHECK1-NEXT:  entry:
 // CHECK1-NEXT:    call void @__cxx_global_var_init()
 // CHECK1-NEXT:    call void @__cxx_global_var_init.1()
 // CHECK1-NEXT:    call void @__cxx_global_var_init.2()
 // CHECK1-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__cxx_global_var_init
 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) @test)
 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @test, ptr @__dso_handle) #[[ATTR2:[0-9]+]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
 // CHECK3-NEXT:    store float [[CONV]], ptr [[F]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__cxx_global_var_init.1
 // CHECK3-SAME: () #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @s_arr, float noundef 1.000000e+00)
 // CHECK3-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i32 1), float noundef 2.000000e+00)
 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i32 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @s_arr
 // CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK3:       arraydestroy.done1:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 // CHECK3-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
 // CHECK3-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@__cxx_global_var_init.2
 // CHECK3-SAME: () #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @var, float noundef 3.000000e+00)
 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @var, ptr @__dso_handle) #[[ATTR2]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@main
 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr @t_var, align 4
 // CHECK3-NEXT:    store i32 [[TMP0]], ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr @_ZZ4mainE5sivar, align 4
 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[SIVAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[SIVAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[TMP4]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[TMP5]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 // CHECK3-NEXT:    store ptr null, ptr [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr @vec, ptr [[TMP7]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr @vec, ptr [[TMP8]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr null, ptr [[TMP9]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr @s_arr, ptr [[TMP10]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr @s_arr, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr @var, ptr [[TMP13]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr @var, ptr [[TMP14]], align 4
 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr null, ptr [[TMP15]], align 4
 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[TMP16]], align 4
 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[TMP17]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
 // CHECK3-NEXT:    store ptr null, ptr [[TMP18]], align 4
 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 3, ptr [[TMP21]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 5, ptr [[TMP22]], align 4
 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr [[TMP20]], ptr [[TMP24]], align 4
 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP25]], align 4
 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP26]], align 4
 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK3-NEXT:    store ptr null, ptr [[TMP27]], align 4
 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK3-NEXT:    store ptr null, ptr [[TMP28]], align 4
 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK3-NEXT:    store i64 2, ptr [[TMP29]], align 8
 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK3-NEXT:    store i64 0, ptr [[TMP30]], align 8
 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP32]], align 4
 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP33]], align 4
 // CHECK3-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
 // CHECK3-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101(i32 [[TMP1]], ptr @vec, ptr @s_arr, ptr @var, i32 [[TMP3]]) #[[ATTR2]]
 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK3:       omp_offload.cont:
 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v()
 // CHECK3-NEXT:    ret i32 [[CALL]]
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101
 // CHECK3-SAME: (i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 noundef [[SIVAR:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[SIVAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[SIVAR_CASTED]], align 4
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined, ptr [[TMP0]], i32 [[TMP4]], ptr [[TMP1]], ptr [[TMP2]], i32 [[TMP6]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 noundef [[SIVAR:%.*]]) #[[ATTR4]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC1:%.*]] = alloca [2 x i32], align 4
 // CHECK3-NEXT:    [[S_ARR2:%.*]] = alloca [2 x %struct.S], align 4
 // CHECK3-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK3-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S:%.*]], align 4
 // CHECK3-NEXT:    [[AGG_TMP5:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC1]], ptr align 4 [[TMP0]], i32 8, i1 false)
 // CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK3:       omp.arraycpy.body:
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK3:       omp.arraycpy.done3:
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP5]])
 // CHECK3-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR4]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP2]], ptr noundef [[AGG_TMP5]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP5]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    store i32 [[TMP13]], ptr [[T_VAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    store i32 [[TMP15]], ptr [[SIVAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined.omp_outlined, i32 [[TMP11]], i32 [[TMP12]], ptr [[VEC1]], i32 [[TMP14]], ptr [[S_ARR2]], ptr [[VAR4]], i32 [[TMP16]]), !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK3-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN7]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP23]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN7]]
 // CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE8:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK3:       arraydestroy.done8:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN2StC1Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN2StC2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfEC1ERKS0_2St
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIfEC2ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0]], ptr noundef [[T]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN2StD1Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN2StD2Ev(ptr noundef nonnull align 4 dereferenceable(8) [[THIS1]]) #[[ATTR2]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l101.omp_outlined.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 noundef [[SIVAR:%.*]]) #[[ATTR4]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC1:%.*]] = alloca [2 x i32], align 4
 // CHECK3-NEXT:    [[S_ARR2:%.*]] = alloca [2 x %struct.S], align 4
 // CHECK3-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK3-NEXT:    [[VAR4:%.*]] = alloca [[STRUCT_S:%.*]], align 4
 // CHECK3-NEXT:    [[AGG_TMP5:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC1]], ptr align 4 [[TMP0]], i32 8, i1 false)
 // CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE3:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK3:       omp.arraycpy.body:
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE3]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK3:       omp.arraycpy.done3:
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP5]])
 // CHECK3-NEXT:    call void @_ZN1SIfEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR4]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP2]], ptr noundef [[AGG_TMP5]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP5]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP7]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 1
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP10]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10:![0-9]+]]
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP13]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC1]], i32 0, i32 [[TMP15]]
 // CHECK3-NEXT:    store i32 [[TMP14]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 [[TMP16]]
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX7]], ptr align 4 [[VAR4]], i32 4, i1 false), !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP17]]
 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK3-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP10]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK3-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR4]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR2]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN10]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP24]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
 // CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK3:       arraydestroy.done11:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
 // CHECK3-SAME: () #[[ATTR1]] comdat {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK3-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
 // CHECK3-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK3-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 4
 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 4
 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 // CHECK3-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
 // CHECK3-NEXT:    store i32 0, ptr [[T_VAR]], align 4
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i32 8, i1 false)
 // CHECK3-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
 // CHECK3-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef 1)
 // CHECK3-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i32 1
 // CHECK3-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef 2)
 // CHECK3-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 4
 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[TMP6]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[TMP7]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 // CHECK3-NEXT:    store ptr null, ptr [[TMP8]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP9]], align 4
 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP10]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[TMP12]], align 4
 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[TMP13]], align 4
 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr null, ptr [[TMP14]], align 4
 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP15]], align 4
 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr [[TMP5]], ptr [[TMP16]], align 4
 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr null, ptr [[TMP17]], align 4
 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 3, ptr [[TMP20]], align 4
 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 4, ptr [[TMP21]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 // CHECK3-NEXT:    store ptr [[TMP18]], ptr [[TMP22]], align 4
 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 // CHECK3-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 4
 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 // CHECK3-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP24]], align 4
 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 // CHECK3-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP25]], align 4
 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 // CHECK3-NEXT:    store ptr null, ptr [[TMP26]], align 4
 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 // CHECK3-NEXT:    store ptr null, ptr [[TMP27]], align 4
 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 // CHECK3-NEXT:    store i64 2, ptr [[TMP28]], align 8
 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 // CHECK3-NEXT:    store i64 0, ptr [[TMP29]], align 8
 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP30]], align 4
 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 // CHECK3-NEXT:    store i32 0, ptr [[TMP32]], align 4
 // CHECK3-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.region_id, ptr [[KERNEL_ARGS]])
 // CHECK3-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
 // CHECK3-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 // CHECK3:       omp_offload.failed:
 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56(i32 [[TMP2]], ptr [[VEC]], ptr [[S_ARR]], ptr [[TMP3]]) #[[ATTR2]]
 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 // CHECK3:       omp_offload.cont:
 // CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP35]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
 // CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK3:       arraydestroy.done2:
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[TMP36:%.*]] = load i32, ptr [[RETVAL]], align 4
 // CHECK3-NEXT:    ret i32 [[TMP36]]
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN2StC2Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK3-NEXT:    store i32 0, ptr [[A]], align 4
 // CHECK3-NEXT:    [[B:%.*]] = getelementptr inbounds [[STRUCT_ST]], ptr [[THIS1]], i32 0, i32 1
 // CHECK3-NEXT:    store i32 0, ptr [[B]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIfEC2ERKS0_2St
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    [[F2:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP1:%.*]] = load float, ptr [[F2]], align 4
 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP2]] to float
 // CHECK3-NEXT:    [[ADD:%.*]] = fadd float [[TMP1]], [[CONV]]
 // CHECK3-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN2StD2Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef [[TMP0]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56
 // CHECK3-SAME: (i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined, ptr [[TMP0]], i32 [[TMP4]], ptr [[TMP1]], ptr [[TMP5]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
 // CHECK3-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK3-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK3-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK3-NEXT:    [[AGG_TMP6:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK3-NEXT:    [[_TMP7:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC2]], ptr align 4 [[TMP0]], i32 8, i1 false)
 // CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP3]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK3:       omp.arraycpy.body:
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP3]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK3:       omp.arraycpy.done4:
 // CHECK3-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]])
 // CHECK3-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP4]], ptr noundef [[AGG_TMP6]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]]) #[[ATTR2]]
 // CHECK3-NEXT:    store ptr [[VAR5]], ptr [[_TMP7]], align 4
 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP6]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 // CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    store i32 [[TMP14]], ptr [[T_VAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[_TMP7]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined, i32 [[TMP12]], i32 [[TMP13]], ptr [[VEC2]], i32 [[TMP15]], ptr [[S_ARR3]], ptr [[TMP16]]), !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
 // CHECK3-NEXT:    br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN9:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN9]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP23]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN9]]
 // CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE10:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK3:       arraydestroy.done10:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiEC1ERKS0_2St
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIiEC2ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP0]], ptr noundef [[T]])
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l56.omp_outlined.omp_outlined
 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR4]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
 // CHECK3-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK3-NEXT:    [[AGG_TMP:%.*]] = alloca [[STRUCT_ST:%.*]], align 4
 // CHECK3-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK3-NEXT:    [[AGG_TMP6:%.*]] = alloca [[STRUCT_ST]], align 4
 // CHECK3-NEXT:    [[_TMP7:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC2]], ptr align 4 [[TMP0]], i32 8, i1 false)
 // CHECK3-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 // CHECK3:       omp.arraycpy.body:
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr noundef nonnull align 4 dereferenceable(4) [[OMP_ARRAYCPY_SRCELEMENTPAST]], ptr noundef [[AGG_TMP]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 // CHECK3-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 // CHECK3-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
 // CHECK3:       omp.arraycpy.done4:
 // CHECK3-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK3-NEXT:    call void @_ZN2StC1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]])
 // CHECK3-NEXT:    call void @_ZN1SIiEC1ERKS0_2St(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]], ptr noundef nonnull align 4 dereferenceable(4) [[TMP6]], ptr noundef [[AGG_TMP6]])
 // CHECK3-NEXT:    call void @_ZN2StD1Ev(ptr noundef nonnull align 4 dereferenceable(8) [[AGG_TMP6]]) #[[ATTR2]]
 // CHECK3-NEXT:    store ptr [[VAR5]], ptr [[_TMP7]], align 4
 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1
 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK3:       cond.true:
 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 // CHECK3:       cond.false:
 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    br label [[COND_END]]
 // CHECK3:       cond.end:
 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK3-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK3:       omp.inner.for.cond:
 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
 // CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 // CHECK3:       omp.inner.for.cond.cleanup:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 // CHECK3:       omp.inner.for.body:
 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC2]], i32 0, i32 [[TMP16]]
 // CHECK3-NEXT:    store i32 [[TMP15]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[_TMP7]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 [[TMP18]]
 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX9]], ptr align 4 [[TMP17]], i32 4, i1 false), !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK3:       omp.body.continue:
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK3:       omp.inner.for.inc:
 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP19]], 1
 // CHECK3-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
 // CHECK3:       omp.inner.for.end:
 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK3:       omp.loop.exit:
 // CHECK3-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK3-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
 // CHECK3-NEXT:    br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK3:       .omp.final.then:
 // CHECK3-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK3:       .omp.final.done:
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN11]], i32 2
 // CHECK3-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK3:       arraydestroy.body:
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP24]], [[DOTOMP_FINAL_DONE]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK3-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK3-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK3-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN11]]
 // CHECK3-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE12:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK3:       arraydestroy.done12:
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK3-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK3-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP1]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiEC2ERKS0_2St
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[S:%.*]], ptr noundef [[T:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[S_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    [[T_INDIRECT_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[S]], ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    store ptr [[T]], ptr [[T_INDIRECT_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[S_ADDR]], align 4
 // CHECK3-NEXT:    [[F2:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[TMP0]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[F2]], align 4
 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_ST:%.*]], ptr [[T]], i32 0, i32 0
 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[A]], align 4
 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[TMP2]]
 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK3-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_teams_distribute_parallel_for_simd_firstprivate_codegen.cpp
 // CHECK3-SAME: () #[[ATTR0]] {
 // CHECK3-NEXT:  entry:
 // CHECK3-NEXT:    call void @__cxx_global_var_init()
 // CHECK3-NEXT:    call void @__cxx_global_var_init.1()
 // CHECK3-NEXT:    call void @__cxx_global_var_init.2()
 // CHECK3-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__cxx_global_var_init
 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) @test)
 // CHECK5-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @test, ptr @__dso_handle) #[[ATTR2:[0-9]+]]
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__cxx_global_var_init.1
 // CHECK5-SAME: () #[[ATTR0]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @s_arr, float noundef 1.000000e+00)
 // CHECK5-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 1), float noundef 2.000000e+00)
 // CHECK5-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2]]
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
 // CHECK5-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR0]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK5-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK5:       arraydestroy.body:
 // CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK5-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK5-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @s_arr
 // CHECK5-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK5:       arraydestroy.done1:
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@__cxx_global_var_init.2
 // CHECK5-SAME: () #[[ATTR0]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @var, float noundef 3.000000e+00)
 // CHECK5-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @var, ptr @__dso_handle) #[[ATTR2]]
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@main
 // CHECK5-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2:![0-9]+]]
 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr @t_var, align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr @vec, i64 0, i64 [[IDXPROM]]
 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[IDXPROM1:%.*]] = sext i32 [[TMP6]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x %struct.S], ptr @s_arr, i64 0, i64 [[IDXPROM1]]
 // CHECK5-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX2]], ptr align 4 @var, i64 4, i1 false), !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr @_ZZ4mainE5sivar, align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], [[TMP7]]
 // CHECK5-NEXT:    store i32 [[ADD3]], ptr @_ZZ4mainE5sivar, align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK5-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
 // CHECK5-NEXT:    ret i32 [[CALL]]
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
 // CHECK5-SAME: () #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK5-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
 // CHECK5-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK5-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
 // CHECK5-NEXT:    store i32 0, ptr [[T_VAR]], align 4
 // CHECK5-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i64 8, i1 false)
 // CHECK5-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 0
 // CHECK5-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef signext 1)
 // CHECK5-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i64 1
 // CHECK5-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef signext 2)
 // CHECK5-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
 // CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 8
 // CHECK5-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 8
 // CHECK5-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VAR]], align 8
 // CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK5-NEXT:    store ptr [[TMP2]], ptr [[_TMP1]], align 8
 // CHECK5-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK5:       omp.inner.for.cond:
 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK5:       omp.inner.for.body:
 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i64 0, i64 [[IDXPROM]]
 // CHECK5-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP1]], align 8, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[IDXPROM3:%.*]] = sext i32 [[TMP11]] to i64
 // CHECK5-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i64 0, i64 [[IDXPROM3]]
 // CHECK5-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX4]], ptr align 4 [[TMP10]], i64 4, i1 false), !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK5:       omp.body.continue:
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK5:       omp.inner.for.inc:
 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK5-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP6]]
 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
 // CHECK5:       omp.inner.for.end:
 // CHECK5-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK5-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK5-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
 // CHECK5-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK5:       arraydestroy.body:
 // CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_INNER_FOR_END]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK5-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK5-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK5-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
 // CHECK5-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE6:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK5:       arraydestroy.done6:
 // CHECK5-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR2]]
 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, ptr [[RETVAL]], align 4
 // CHECK5-NEXT:    ret i32 [[TMP14]]
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK5-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
 // CHECK5-NEXT:    store float [[CONV]], ptr [[F]], align 4
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK5-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 // CHECK5-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
 // CHECK5-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef signext [[TMP0]])
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK5-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK5-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP1]]
 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
 // CHECK5-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK5-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_teams_distribute_parallel_for_simd_firstprivate_codegen.cpp
 // CHECK5-SAME: () #[[ATTR0]] {
 // CHECK5-NEXT:  entry:
 // CHECK5-NEXT:    call void @__cxx_global_var_init()
 // CHECK5-NEXT:    call void @__cxx_global_var_init.1()
 // CHECK5-NEXT:    call void @__cxx_global_var_init.2()
 // CHECK5-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@__cxx_global_var_init
 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) @test)
 // CHECK7-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @test, ptr @__dso_handle) #[[ATTR2:[0-9]+]]
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@__cxx_global_var_init.1
 // CHECK7-SAME: () #[[ATTR0]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @s_arr, float noundef 1.000000e+00)
 // CHECK7-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i32 1), float noundef 2.000000e+00)
 // CHECK7-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2]]
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
 // CHECK7-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR0]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
 // CHECK7-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK7:       arraydestroy.body:
 // CHECK7-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i32 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK7-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK7-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK7-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @s_arr
 // CHECK7-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK7:       arraydestroy.done1:
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@__cxx_global_var_init.2
 // CHECK7-SAME: () #[[ATTR0]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @var, float noundef 3.000000e+00)
 // CHECK7-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @var, ptr @__dso_handle) #[[ATTR2]]
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@main
 // CHECK7-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK7:       omp.inner.for.cond:
 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3:![0-9]+]]
 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK7:       omp.inner.for.body:
 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK7-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, ptr @t_var, align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr @vec, i32 0, i32 [[TMP5]]
 // CHECK7-NEXT:    store i32 [[TMP4]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds [2 x %struct.S], ptr @s_arr, i32 0, i32 [[TMP6]]
 // CHECK7-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX1]], ptr align 4 @var, i32 4, i1 false), !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, ptr @_ZZ4mainE5sivar, align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], [[TMP7]]
 // CHECK7-NEXT:    store i32 [[ADD2]], ptr @_ZZ4mainE5sivar, align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK7:       omp.body.continue:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK7:       omp.inner.for.inc:
 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
 // CHECK7-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
 // CHECK7:       omp.inner.for.end:
 // CHECK7-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK7-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v()
 // CHECK7-NEXT:    ret i32 [[CALL]]
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
 // CHECK7-SAME: () #[[ATTR1]] comdat {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
 // CHECK7-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
 // CHECK7-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
 // CHECK7-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
 // CHECK7-NEXT:    store i32 0, ptr [[T_VAR]], align 4
 // CHECK7-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i32 8, i1 false)
 // CHECK7-NEXT:    [[ARRAYINIT_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
 // CHECK7-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_BEGIN]], i32 noundef 1)
 // CHECK7-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYINIT_BEGIN]], i32 1
 // CHECK7-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef 2)
 // CHECK7-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 4
 // CHECK7-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 4
 // CHECK7-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VAR]], align 4
 // CHECK7-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK7-NEXT:    store ptr [[TMP2]], ptr [[_TMP1]], align 4
 // CHECK7-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK7-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK7:       omp.inner.for.cond:
 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7:![0-9]+]]
 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK7:       omp.inner.for.body:
 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK7-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, ptr [[T_VAR]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC]], i32 0, i32 [[TMP9]]
 // CHECK7-NEXT:    store i32 [[TMP8]], ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[_TMP1]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 [[TMP11]]
 // CHECK7-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX3]], ptr align 4 [[TMP10]], i32 4, i1 false), !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK7:       omp.body.continue:
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK7:       omp.inner.for.inc:
 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
 // CHECK7-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP7]]
 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
 // CHECK7:       omp.inner.for.end:
 // CHECK7-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK7-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK7-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
 // CHECK7-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
 // CHECK7-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK7:       arraydestroy.body:
 // CHECK7-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP13]], [[OMP_INNER_FOR_END]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK7-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
 // CHECK7-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK7-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
 // CHECK7-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE5:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK7:       arraydestroy.done5:
 // CHECK7-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR2]]
 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, ptr [[RETVAL]], align 4
 // CHECK7-NEXT:    ret i32 [[TMP14]]
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK7-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK7-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
 // CHECK7-NEXT:    store float [[CONV]], ptr [[F]], align 4
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK7-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK7-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 // CHECK7-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
 // CHECK7-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef [[TMP0]])
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK7-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK7-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 // CHECK7-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], [[TMP1]]
 // CHECK7-NEXT:    store i32 [[ADD]], ptr [[F]], align 4
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
 // CHECK7-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 // CHECK7-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK7-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_teams_distribute_parallel_for_simd_firstprivate_codegen.cpp
 // CHECK7-SAME: () #[[ATTR0]] {
 // CHECK7-NEXT:  entry:
 // CHECK7-NEXT:    call void @__cxx_global_var_init()
 // CHECK7-NEXT:    call void @__cxx_global_var_init.1()
 // CHECK7-NEXT:    call void @__cxx_global_var_init.2()
 // CHECK7-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@__cxx_global_var_init
 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) @test)
 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @test, ptr @__dso_handle) #[[ATTR2:[0-9]+]]
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK9-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
 // CHECK9-NEXT:    store float [[CONV]], ptr [[F]], align 4
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@__cxx_global_var_init.1
 // CHECK9-SAME: () #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @s_arr, float noundef 1.000000e+00)
 // CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 1), float noundef 2.000000e+00)
 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2]]
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK9-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK9-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK9:       arraydestroy.body:
 // CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK9-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK9-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK9-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @s_arr
 // CHECK9-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK9:       arraydestroy.done1:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK9-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK9-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK9-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 // CHECK9-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
 // CHECK9-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@__cxx_global_var_init.2
 // CHECK9-SAME: () #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @var, float noundef 3.000000e+00)
 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @var, ptr @__dso_handle) #[[ATTR2]]
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@main
 // CHECK9-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
 // CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
 // CHECK9-NEXT:    ret i32 0
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75
 // CHECK9-SAME: (i64 noundef [[G:%.*]], i64 noundef [[SIVAR:%.*]], i64 noundef [[G1:%.*]]) #[[ATTR4:[0-9]+]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[G_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[G1_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    store i64 [[G]], ptr [[G_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[G1]], ptr [[G1_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[G1_ADDR]], ptr [[TMP]], align 8
 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[G_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP0]], ptr [[G_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[G_CASTED]], align 8
 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[TMP]], align 8
 // CHECK9-NEXT:    [[TMP3:%.*]] = load volatile i32, ptr [[TMP2]], align 4
 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[G1_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[G1_CASTED]], align 8
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[SIVAR_CASTED]], align 4
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined, i64 [[TMP1]], i64 [[TMP4]], i64 [[TMP6]])
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[G:%.*]], i64 noundef [[G1:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR4]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[G_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[G1_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[G]], ptr [[G_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[G1]], ptr [[G1_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[G1_ADDR]], ptr [[TMP]], align 8
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 1
 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP4:![0-9]+]]
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[G_ADDR]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    store i32 [[TMP11]], ptr [[G_CASTED]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP12:%.*]] = load i64, ptr [[G_CASTED]], align 8, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP]], align 8, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP14:%.*]] = load volatile i32, ptr [[TMP13]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    store i32 [[TMP14]], ptr [[G1_CASTED]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, ptr [[G1_CASTED]], align 8, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    store i32 [[TMP16]], ptr [[SIVAR_CASTED]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP17:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]], i64 [[TMP12]], i64 [[TMP15]], i64 [[TMP17]]), !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP4]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 // CHECK9-NEXT:    br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l75.omp_outlined.omp_outlined
 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[G:%.*]], i64 noundef [[G1:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR4]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
 // CHECK9-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[G]], ptr [[G_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[G1]], ptr [[G1_ADDR]], align 8
 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 // CHECK9-NEXT:    store ptr [[G1_ADDR]], ptr [[TMP]], align 8
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
 // CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 // CHECK9:       cond.true:
 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 // CHECK9:       cond.false:
 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    br label [[COND_END]]
 // CHECK9:       cond.end:
 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 // CHECK9:       omp.inner.for.cond:
 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8:![0-9]+]]
 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
 // CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 // CHECK9:       omp.inner.for.body:
 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    store i32 1, ptr [[G_ADDR]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[TMP]], align 8, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    store volatile i32 1, ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    store i32 2, ptr [[SIVAR_ADDR]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
 // CHECK9-NEXT:    store ptr [[G_ADDR]], ptr [[TMP11]], align 8, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
 // CHECK9-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP]], align 8, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
 // CHECK9-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP14]], align 8, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]]), !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 // CHECK9:       omp.body.continue:
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 // CHECK9:       omp.inner.for.inc:
 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP15]], 1
 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
 // CHECK9:       omp.inner.for.end:
 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 // CHECK9:       omp.loop.exit:
 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 // CHECK9-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
 // CHECK9-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
 // CHECK9:       .omp.final.then:
 // CHECK9-NEXT:    store i32 2, ptr [[I]], align 4
 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
 // CHECK9:       .omp.final.done:
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK9-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_teams_distribute_parallel_for_simd_firstprivate_codegen.cpp
 // CHECK9-SAME: () #[[ATTR0]] {
 // CHECK9-NEXT:  entry:
 // CHECK9-NEXT:    call void @__cxx_global_var_init()
 // CHECK9-NEXT:    call void @__cxx_global_var_init.1()
 // CHECK9-NEXT:    call void @__cxx_global_var_init.2()
 // CHECK9-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@__cxx_global_var_init
 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) @test)
 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @test, ptr @__dso_handle) #[[ATTR2:[0-9]+]]
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR2]]
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@__cxx_global_var_init.1
 // CHECK11-SAME: () #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @s_arr, float noundef 1.000000e+00)
 // CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 1), float noundef 2.000000e+00)
 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR2]]
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@__cxx_global_array_dtor
 // CHECK11-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 // CHECK11-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 // CHECK11-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
 // CHECK11:       arraydestroy.body:
 // CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ getelementptr inbounds ([[STRUCT_S:%.*]], ptr @s_arr, i64 2), [[ENTRY:%.*]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
 // CHECK11-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
 // CHECK11-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR2]]
 // CHECK11-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], @s_arr
 // CHECK11-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE1:%.*]], label [[ARRAYDESTROY_BODY]]
 // CHECK11:       arraydestroy.done1:
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@__cxx_global_var_init.2
 // CHECK11-SAME: () #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) @var, float noundef 3.000000e+00)
 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @_ZN1SIfED1Ev, ptr @var, ptr @__dso_handle) #[[ATTR2]]
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@main
 // CHECK11-SAME: () #[[ATTR3:[0-9]+]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 // CHECK11-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
 // CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 // CHECK11-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
 // CHECK11-NEXT:    ret i32 0
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK11-NEXT:    [[TMP0:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
 // CHECK11-NEXT:    store float [[CONV]], ptr [[F]], align 4
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
 // CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 // CHECK11-NEXT:    [[F:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 // CHECK11-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
 // CHECK11-NEXT:    [[TMP1:%.*]] = load volatile i32, ptr @g, align 4
 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP1]] to float
 // CHECK11-NEXT:    [[ADD:%.*]] = fadd float [[TMP0]], [[CONV]]
 // CHECK11-NEXT:    store float [[ADD]], ptr [[F]], align 4
 // CHECK11-NEXT:    ret void
 //
 //
 // CHECK11-LABEL: define {{[^@]+}}@_GLOBAL__sub_I_teams_distribute_parallel_for_simd_firstprivate_codegen.cpp
 // CHECK11-SAME: () #[[ATTR0]] {
 // CHECK11-NEXT:  entry:
 // CHECK11-NEXT:    call void @__cxx_global_var_init()
 // CHECK11-NEXT:    call void @__cxx_global_var_init.1()
 // CHECK11-NEXT:    call void @__cxx_global_var_init.2()
 // CHECK11-NEXT:    ret void
 //