llvm/clang/test/OpenMP/distribute_parallel_for_simd_num_threads_codegen.cpp

// 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 -no-enable-noundef-analysis -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1

// RUN: %clang_cc1 -no-enable-noundef-analysis -verify -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3

// RUN: %clang_cc1 -no-enable-noundef-analysis -verify -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK5
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5

// RUN: %clang_cc1 -no-enable-noundef-analysis -verify -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3

// RUN: %clang_cc1 -no-enable-noundef-analysis -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK9
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9

// RUN: %clang_cc1 -no-enable-noundef-analysis -verify -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11

// RUN: %clang_cc1 -no-enable-noundef-analysis -verify -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK13
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK13

// RUN: %clang_cc1 -no-enable-noundef-analysis -verify -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK11
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
// RUN: %clang_cc1 -no-enable-noundef-analysis -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11

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

typedef __INTPTR_TYPE__ intptr_t;


void foo();

struct S {
  intptr_t a, b, c;
  S(intptr_t a) : a(a) {}
  operator char() {  extern void mayThrow(); mayThrow(); return a; }
  ~S() {}
};

template <typename T, int C>
int tmain() {
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for simd num_threads(C)
  for (int i = 0; i < 100; i++)
    foo();
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for simd num_threads(T(23))
  for (int i = 0; i < 100; i++)
    foo();
  return 0;
}

int main() {
  S s(0);
  char a = s;
#pragma omp target
#pragma omp teams
#pragma omp distribute parallel for simd num_threads(2)
  for (int i = 0; i < 100; i++) {
    foo();
  }
#pragma omp target
#pragma omp teams

#pragma omp distribute parallel for simd num_threads(a)
  for (int i = 0; i < 100; i++) {
    foo();
  }
  return a + tmain<char, 5>() + tmain<S, 1>();
}

// tmain 5

// tmain 1








#endif
// CHECK1-LABEL: define {{[^@]+}}@main
// CHECK1-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK1-NEXT:    [[A:%.*]] = alloca i8, align 1
// CHECK1-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[EHSELECTOR_SLOT:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK1-NEXT:    call void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[S]], i64 0)
// CHECK1-NEXT:    [[CALL:%.*]] = invoke signext i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[S]])
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK1:       invoke.cont:
// CHECK1-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = 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_l68.region_id, ptr [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], 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_l68() #[[ATTR3:[0-9]+]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       lpad:
// CHECK1-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            cleanup
// CHECK1-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK1-NEXT:    store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
// CHECK1-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
// CHECK1-NEXT:    store i32 [[TMP17]], ptr [[EHSELECTOR_SLOT]], align 4
// CHECK1-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[EH_RESUME:%.*]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[A]], align 1
// CHECK1-NEXT:    store i8 [[TMP18]], ptr [[A_CASTED]], align 1
// CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[A_CASTED]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    store i64 [[TMP19]], ptr [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    store i64 [[TMP19]], ptr [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK1-NEXT:    store i32 3, ptr [[TMP25]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK1-NEXT:    store i32 1, ptr [[TMP26]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK1-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
// CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK1-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
// CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP29]], align 8
// CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP30]], align 8
// CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK1-NEXT:    store ptr null, ptr [[TMP31]], align 8
// CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK1-NEXT:    store ptr null, ptr [[TMP32]], align 8
// CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK1-NEXT:    store i64 100, ptr [[TMP33]], align 8
// CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK1-NEXT:    store i64 0, ptr [[TMP34]], align 8
// CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
// CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
// CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK1-NEXT:    store i32 0, ptr [[TMP37]], align 4
// CHECK1-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.region_id, ptr [[KERNEL_ARGS2]])
// CHECK1-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
// CHECK1-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK1:       omp_offload.failed3:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74(i64 [[TMP19]]) #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK1:       omp_offload.cont4:
// CHECK1-NEXT:    [[TMP40:%.*]] = load i8, ptr [[A]], align 1
// CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP40]] to i32
// CHECK1-NEXT:    [[CALL6:%.*]] = invoke i32 @_Z5tmainIcLi5EEiv()
// CHECK1-NEXT:            to label [[INVOKE_CONT5:%.*]] unwind label [[LPAD]]
// CHECK1:       invoke.cont5:
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CALL6]]
// CHECK1-NEXT:    [[CALL8:%.*]] = invoke i32 @_Z5tmainI1SLi1EEiv()
// CHECK1-NEXT:            to label [[INVOKE_CONT7:%.*]] unwind label [[LPAD]]
// CHECK1:       invoke.cont7:
// CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD]], [[CALL8]]
// CHECK1-NEXT:    store i32 [[ADD9]], ptr [[RETVAL]], align 4
// CHECK1-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK1-NEXT:    [[TMP41:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK1-NEXT:    ret i32 [[TMP41]]
// CHECK1:       eh.resume:
// CHECK1-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
// CHECK1-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
// CHECK1-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
// CHECK1-NEXT:    [[LPAD_VAL10:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
// CHECK1-NEXT:    resume { ptr, i32 } [[LPAD_VAL10]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SC1El
// CHECK1-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    call void @_ZN1SC2El(ptr nonnull align 8 dereferenceable(24) [[THIS1]], i64 [[TMP0]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1ScvcEv
// CHECK1-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[ATTR1]] comdat align 2 {
// 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 @_Z8mayThrowv()
// CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
// CHECK1-NEXT:    ret i8 [[CONV]]
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68
// CHECK1-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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:    [[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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 99, 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:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-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)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// 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:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 2), !llvm.access.group [[ACC_GRP9]]
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP9]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    store i32 [[ADD]], 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:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[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:    [[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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] 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:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK1-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)
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// 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:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13:![0-9]+]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK1-NEXT:    invoke void @_Z3foov()
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP13]]
// CHECK1:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
// CHECK1:       terminate.lpad:
// CHECK1-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            catch ptr null
// CHECK1-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7:[0-9]+]], !llvm.access.group [[ACC_GRP13]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK1-LABEL: define {{[^@]+}}@__clang_call_terminate
// CHECK1-SAME: (ptr [[TMP0:%.*]]) #[[ATTR5:[0-9]+]] comdat {
// CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR3]]
// CHECK1-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74
// CHECK1-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined, ptr [[A_ADDR]])
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 1 dereferenceable(1) [[A:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, 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:    [[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 ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 99, 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:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
// 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:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP0]], align 1, !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    [[TMP9:%.*]] = sext i8 [[TMP8]] to i32
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP2]], i32 [[TMP9]]), !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]]), !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
// CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK1-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[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:    [[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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] 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:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// 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:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21:![0-9]+]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK1-NEXT:    invoke void @_Z3foov()
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP21]]
// CHECK1:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
// CHECK1:       terminate.lpad:
// CHECK1-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            catch ptr null
// CHECK1-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP21]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
// CHECK1-SAME: () #[[ATTR6:[0-9]+]] comdat {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52() #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK1-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK1-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK1-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK1-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK1-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK1-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK1-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK1-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK1-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK1-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK1-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK1-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK1:       omp_offload.failed3:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57() #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK1:       omp_offload.cont4:
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
// CHECK1-SAME: () #[[ATTR6]] comdat {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK1-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK1-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK1:       omp_offload.failed:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52() #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK1:       omp_offload.cont:
// CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK1-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK1-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK1-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK1-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK1-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK1-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK1-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK1-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK1-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK1-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK1-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK1-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK1:       omp_offload.failed3:
// CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57() #[[ATTR3]]
// CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK1:       omp_offload.cont4:
// CHECK1-NEXT:    ret i32 0
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SD1Ev
// CHECK1-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] comdat align 2 {
// 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 @_ZN1SD2Ev(ptr nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR3]]
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SC2El
// CHECK1-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR6]] comdat align 2 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK1-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK1-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@_ZN1SD2Ev
// CHECK1-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] comdat align 2 {
// 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 {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52
// CHECK1-SAME: () #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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:    [[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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 99, 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:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// 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:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24:![0-9]+]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 5), !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[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:    [[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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] 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:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// 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:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27:![0-9]+]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK1-NEXT:    invoke void @_Z3foov()
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP27]]
// CHECK1:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
// CHECK1:       terminate.lpad:
// CHECK1-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            catch ptr null
// CHECK1-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP27]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57
// CHECK1-SAME: () #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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:    [[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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 99, 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:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// 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:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30:![0-9]+]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 23), !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[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:    [[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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] 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:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// 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:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33:![0-9]+]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK1-NEXT:    invoke void @_Z3foov()
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP33]]
// CHECK1:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
// CHECK1:       terminate.lpad:
// CHECK1-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            catch ptr null
// CHECK1-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP33]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52
// CHECK1-SAME: () #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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:    [[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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 99, 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:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// 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:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36:![0-9]+]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 1), !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[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:    [[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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] 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:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// 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:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39:![0-9]+]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK1-NEXT:    invoke void @_Z3foov()
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP39]]
// CHECK1:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
// CHECK1:       terminate.lpad:
// CHECK1-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            catch ptr null
// CHECK1-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP39]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57
// CHECK1-SAME: () #[[ATTR2]] {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined)
// CHECK1-NEXT:    ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK1-NEXT:  entry:
// CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, 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:    [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], 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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 99, 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:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// 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:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42:![0-9]+]]
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    invoke void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 23)
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP42]]
// CHECK1:       invoke.cont:
// CHECK1-NEXT:    [[CALL:%.*]] = invoke signext i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]])
// CHECK1-NEXT:            to label [[INVOKE_CONT2:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP42]]
// CHECK1:       invoke.cont2:
// CHECK1-NEXT:    [[TMP7:%.*]] = sext i8 [[CALL]] to i32
// CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP7]]), !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR3]], !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1:       omp.inner.for.inc:
// CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
// CHECK1:       terminate.lpad:
// CHECK1-NEXT:    [[TMP16:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            catch ptr null
// CHECK1-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP16]], 0
// CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP17]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP42]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined
// CHECK1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[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:    [[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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] 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:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// 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:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    br label [[COND_END]]
// CHECK1:       cond.end:
// CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1:       omp.inner.for.cond:
// CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45:![0-9]+]]
// CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1:       omp.inner.for.body:
// CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK1-NEXT:    invoke void @_Z3foov()
// CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP45]]
// CHECK1:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
// CHECK1:       omp.inner.for.end:
// CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1:       omp.loop.exit:
// CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1:       .omp.final.then:
// CHECK1-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK1:       .omp.final.done:
// CHECK1-NEXT:    ret void
// CHECK1:       terminate.lpad:
// CHECK1-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK1-NEXT:            catch ptr null
// CHECK1-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP45]]
// CHECK1-NEXT:    unreachable
//
//
// CHECK3-LABEL: define {{[^@]+}}@main
// CHECK3-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK3-NEXT:    [[A:%.*]] = alloca i8, align 1
// CHECK3-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
// CHECK3-NEXT:    [[EHSELECTOR_SLOT:%.*]] = 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_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK3-NEXT:    call void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[S]], i64 0)
// CHECK3-NEXT:    [[CALL:%.*]] = invoke signext i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[S]])
// CHECK3-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK3:       invoke.cont:
// CHECK3-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2:![0-9]+]]
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK3-NEXT:    invoke void @_Z3foov()
// CHECK3-NEXT:            to label [[INVOKE_CONT1:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP2]]
// CHECK3:       invoke.cont1:
// 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:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK3-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK3:       lpad:
// CHECK3-NEXT:    [[TMP5:%.*]] = landingpad { ptr, i32 }
// CHECK3-NEXT:            cleanup
// CHECK3-NEXT:    [[TMP6:%.*]] = extractvalue { ptr, i32 } [[TMP5]], 0
// CHECK3-NEXT:    store ptr [[TMP6]], ptr [[EXN_SLOT]], align 8
// CHECK3-NEXT:    [[TMP7:%.*]] = extractvalue { ptr, i32 } [[TMP5]], 1
// CHECK3-NEXT:    store i32 [[TMP7]], ptr [[EHSELECTOR_SLOT]], align 4
// CHECK3-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR5:[0-9]+]]
// CHECK3-NEXT:    br label [[EH_RESUME:%.*]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB4]], align 4
// CHECK3-NEXT:    store i32 99, ptr [[DOTOMP_UB5]], align 4
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_LB4]], align 4
// CHECK3-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_IV6]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK3:       omp.inner.for.cond8:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
// CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB5]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK3-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK3-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
// CHECK3:       omp.inner.for.body10:
// CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK3-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK3-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK3-NEXT:    store i32 [[ADD12]], ptr [[I7]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK3-NEXT:    invoke void @_Z3foov()
// CHECK3-NEXT:            to label [[INVOKE_CONT13:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP6]]
// CHECK3:       invoke.cont13:
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE14:%.*]]
// CHECK3:       omp.body.continue14:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC15:%.*]]
// CHECK3:       omp.inner.for.inc15:
// CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK3-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK3-NEXT:    store i32 [[ADD16]], ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK3:       omp.inner.for.end17:
// CHECK3-NEXT:    store i32 100, ptr [[I7]], align 4
// CHECK3-NEXT:    [[TMP13:%.*]] = load i8, ptr [[A]], align 1
// CHECK3-NEXT:    [[CONV:%.*]] = sext i8 [[TMP13]] to i32
// CHECK3-NEXT:    [[CALL19:%.*]] = invoke i32 @_Z5tmainIcLi5EEiv()
// CHECK3-NEXT:            to label [[INVOKE_CONT18:%.*]] unwind label [[LPAD]]
// CHECK3:       invoke.cont18:
// CHECK3-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV]], [[CALL19]]
// CHECK3-NEXT:    [[CALL22:%.*]] = invoke i32 @_Z5tmainI1SLi1EEiv()
// CHECK3-NEXT:            to label [[INVOKE_CONT21:%.*]] unwind label [[LPAD]]
// CHECK3:       invoke.cont21:
// CHECK3-NEXT:    [[ADD23:%.*]] = add nsw i32 [[ADD20]], [[CALL22]]
// CHECK3-NEXT:    store i32 [[ADD23]], ptr [[RETVAL]], align 4
// CHECK3-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR5]]
// CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK3-NEXT:    ret i32 [[TMP14]]
// CHECK3:       eh.resume:
// CHECK3-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
// CHECK3-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
// CHECK3-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
// CHECK3-NEXT:    [[LPAD_VAL24:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
// CHECK3-NEXT:    resume { ptr, i32 } [[LPAD_VAL24]]
// CHECK3:       terminate.lpad:
// CHECK3-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
// CHECK3-NEXT:            catch ptr null
// CHECK3-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK3-NEXT:    call void @__clang_call_terminate(ptr [[TMP16]]) #[[ATTR6:[0-9]+]], !llvm.access.group [[ACC_GRP2]]
// CHECK3-NEXT:    unreachable
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN1SC1El
// CHECK3-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK3-NEXT:    call void @_ZN1SC2El(ptr nonnull align 8 dereferenceable(24) [[THIS1]], i64 [[TMP0]])
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN1ScvcEv
// CHECK3-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[ATTR1]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    call void @_Z8mayThrowv()
// CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
// CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
// CHECK3-NEXT:    ret i8 [[CONV]]
//
//
// CHECK3-LABEL: define {{[^@]+}}@__clang_call_terminate
// CHECK3-SAME: (ptr [[TMP0:%.*]]) #[[ATTR3:[0-9]+]] comdat {
// CHECK3-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR5]]
// CHECK3-NEXT:    call void @_ZSt9terminatev() #[[ATTR6]]
// CHECK3-NEXT:    unreachable
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
// CHECK3-SAME: () #[[ATTR4:[0-9]+]] comdat personality ptr @__gxx_personality_v0 {
// CHECK3-NEXT:  entry:
// 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_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK3-NEXT:    invoke void @_Z3foov()
// CHECK3-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP9]]
// CHECK3:       invoke.cont:
// 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:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK3-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK3:       omp.inner.for.end:
// CHECK3-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB3]], align 4
// CHECK3-NEXT:    store i32 99, ptr [[DOTOMP_UB4]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB3]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV5]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK3:       omp.inner.for.cond7:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12:![0-9]+]]
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB4]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
// CHECK3:       omp.inner.for.body9:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK3-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK3-NEXT:    store i32 [[ADD11]], ptr [[I6]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK3-NEXT:    invoke void @_Z3foov()
// CHECK3-NEXT:            to label [[INVOKE_CONT12:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP12]]
// CHECK3:       invoke.cont12:
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE13:%.*]]
// CHECK3:       omp.body.continue13:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC14:%.*]]
// CHECK3:       omp.inner.for.inc14:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK3-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK3-NEXT:    store i32 [[ADD15]], ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK3:       omp.inner.for.end16:
// CHECK3-NEXT:    store i32 100, ptr [[I6]], align 4
// CHECK3-NEXT:    ret i32 0
// CHECK3:       terminate.lpad:
// CHECK3-NEXT:    [[TMP10:%.*]] = landingpad { ptr, i32 }
// CHECK3-NEXT:            catch ptr null
// CHECK3-NEXT:    [[TMP11:%.*]] = extractvalue { ptr, i32 } [[TMP10]], 0
// CHECK3-NEXT:    call void @__clang_call_terminate(ptr [[TMP11]]) #[[ATTR6]], !llvm.access.group [[ACC_GRP9]]
// CHECK3-NEXT:    unreachable
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
// CHECK3-SAME: () #[[ATTR4]] comdat personality ptr @__gxx_personality_v0 {
// CHECK3-NEXT:  entry:
// 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_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK3-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3:       omp.inner.for.cond:
// CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
// CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3:       omp.inner.for.body:
// CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK3-NEXT:    invoke void @_Z3foov()
// CHECK3-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP15]]
// CHECK3:       invoke.cont:
// 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:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK3-NEXT:    store i32 [[ADD1]], 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:    store i32 100, ptr [[I]], align 4
// CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB3]], align 4
// CHECK3-NEXT:    store i32 99, ptr [[DOTOMP_UB4]], align 4
// CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB3]], align 4
// CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV5]], align 4
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK3:       omp.inner.for.cond7:
// CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
// CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB4]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
// CHECK3:       omp.inner.for.body9:
// CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK3-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK3-NEXT:    store i32 [[ADD11]], ptr [[I6]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK3-NEXT:    invoke void @_Z3foov()
// CHECK3-NEXT:            to label [[INVOKE_CONT12:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP18]]
// CHECK3:       invoke.cont12:
// CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE13:%.*]]
// CHECK3:       omp.body.continue13:
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC14:%.*]]
// CHECK3:       omp.inner.for.inc14:
// CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK3-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK3-NEXT:    store i32 [[ADD15]], ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK3:       omp.inner.for.end16:
// CHECK3-NEXT:    store i32 100, ptr [[I6]], align 4
// CHECK3-NEXT:    ret i32 0
// CHECK3:       terminate.lpad:
// CHECK3-NEXT:    [[TMP10:%.*]] = landingpad { ptr, i32 }
// CHECK3-NEXT:            catch ptr null
// CHECK3-NEXT:    [[TMP11:%.*]] = extractvalue { ptr, i32 } [[TMP10]], 0
// CHECK3-NEXT:    call void @__clang_call_terminate(ptr [[TMP11]]) #[[ATTR6]], !llvm.access.group [[ACC_GRP15]]
// CHECK3-NEXT:    unreachable
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN1SD1Ev
// CHECK3-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR4]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    call void @_ZN1SD2Ev(ptr nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR5]]
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN1SC2El
// CHECK3-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR4]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK3-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK3-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
// CHECK3-NEXT:    ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_ZN1SD2Ev
// CHECK3-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR4]] comdat align 2 {
// CHECK3-NEXT:  entry:
// CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK3-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@main
// CHECK5-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK5-NEXT:    [[A:%.*]] = alloca i8, align 1
// CHECK5-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[EHSELECTOR_SLOT:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
// CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK5-NEXT:    call void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[S]], i64 0)
// CHECK5-NEXT:    [[CALL:%.*]] = invoke signext i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[S]])
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK5:       invoke.cont:
// CHECK5-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
// CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK5-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK5-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK5-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK5-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK5-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK5-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK5-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK5-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK5-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK5-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK5-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = 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_l68.region_id, ptr [[KERNEL_ARGS]])
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK5:       omp_offload.failed:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR3:[0-9]+]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK5:       lpad:
// CHECK5-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            cleanup
// CHECK5-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK5-NEXT:    store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
// CHECK5-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
// CHECK5-NEXT:    store i32 [[TMP17]], ptr [[EHSELECTOR_SLOT]], align 4
// CHECK5-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK5-NEXT:    br label [[EH_RESUME:%.*]]
// CHECK5:       omp_offload.cont:
// CHECK5-NEXT:    [[TMP18:%.*]] = load i8, ptr [[A]], align 1
// CHECK5-NEXT:    store i8 [[TMP18]], ptr [[A_CASTED]], align 1
// CHECK5-NEXT:    [[TMP19:%.*]] = load i64, ptr [[A_CASTED]], align 8
// CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    store i64 [[TMP19]], ptr [[TMP20]], align 8
// CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    store i64 [[TMP19]], ptr [[TMP21]], align 8
// CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK5-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK5-NEXT:    store i32 3, ptr [[TMP25]], align 4
// CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK5-NEXT:    store i32 1, ptr [[TMP26]], align 4
// CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK5-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
// CHECK5-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK5-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
// CHECK5-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK5-NEXT:    store ptr @.offload_sizes, ptr [[TMP29]], align 8
// CHECK5-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK5-NEXT:    store ptr @.offload_maptypes, ptr [[TMP30]], align 8
// CHECK5-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK5-NEXT:    store ptr null, ptr [[TMP31]], align 8
// CHECK5-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK5-NEXT:    store ptr null, ptr [[TMP32]], align 8
// CHECK5-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK5-NEXT:    store i64 100, ptr [[TMP33]], align 8
// CHECK5-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK5-NEXT:    store i64 0, ptr [[TMP34]], align 8
// CHECK5-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
// CHECK5-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
// CHECK5-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK5-NEXT:    store i32 0, ptr [[TMP37]], align 4
// CHECK5-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.region_id, ptr [[KERNEL_ARGS2]])
// CHECK5-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
// CHECK5-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK5:       omp_offload.failed3:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74(i64 [[TMP19]]) #[[ATTR3]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK5:       omp_offload.cont4:
// CHECK5-NEXT:    [[TMP40:%.*]] = load i8, ptr [[A]], align 1
// CHECK5-NEXT:    [[CONV:%.*]] = sext i8 [[TMP40]] to i32
// CHECK5-NEXT:    [[CALL6:%.*]] = invoke i32 @_Z5tmainIcLi5EEiv()
// CHECK5-NEXT:            to label [[INVOKE_CONT5:%.*]] unwind label [[LPAD]]
// CHECK5:       invoke.cont5:
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CALL6]]
// CHECK5-NEXT:    [[CALL8:%.*]] = invoke i32 @_Z5tmainI1SLi1EEiv()
// CHECK5-NEXT:            to label [[INVOKE_CONT7:%.*]] unwind label [[LPAD]]
// CHECK5:       invoke.cont7:
// CHECK5-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD]], [[CALL8]]
// CHECK5-NEXT:    store i32 [[ADD9]], ptr [[RETVAL]], align 4
// CHECK5-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK5-NEXT:    [[TMP41:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK5-NEXT:    ret i32 [[TMP41]]
// CHECK5:       eh.resume:
// CHECK5-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
// CHECK5-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
// CHECK5-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
// CHECK5-NEXT:    [[LPAD_VAL10:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
// CHECK5-NEXT:    resume { ptr, i32 } [[LPAD_VAL10]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN1SC1El
// CHECK5-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK5-NEXT:    call void @_ZN1SC2El(ptr nonnull align 8 dereferenceable(24) [[THIS1]], i64 [[TMP0]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN1ScvcEv
// CHECK5-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[ATTR1]] comdat align 2 {
// 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 @_Z8mayThrowv()
// CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
// CHECK5-NEXT:    ret i8 [[CONV]]
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68
// CHECK5-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined)
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK5-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)
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_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_GRP9:![0-9]+]]
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 2), !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = 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_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK5-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)
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13:![0-9]+]]
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK5-NEXT:    invoke void @_Z3foov()
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP13]]
// CHECK5:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK5-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
// CHECK5:       terminate.lpad:
// CHECK5-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            catch ptr null
// CHECK5-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7:[0-9]+]], !llvm.access.group [[ACC_GRP13]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@__clang_call_terminate
// CHECK5-SAME: (ptr [[TMP0:%.*]]) #[[ATTR5:[0-9]+]] comdat {
// CHECK5-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR3]]
// CHECK5-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74
// CHECK5-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined, ptr [[A_ADDR]])
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 1 dereferenceable(1) [[A:%.*]]) #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP0]], align 1, !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    [[TMP9:%.*]] = sext i8 [[TMP8]] to i32
// CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP2]], i32 [[TMP9]]), !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]]), !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
// CHECK5-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK5-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = 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_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21:![0-9]+]]
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK5-NEXT:    invoke void @_Z3foov()
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP21]]
// CHECK5:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK5-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
// CHECK5:       terminate.lpad:
// CHECK5-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            catch ptr null
// CHECK5-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP21]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
// CHECK5-SAME: () #[[ATTR6:[0-9]+]] comdat {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK5-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK5-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK5-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK5-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK5-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK5-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK5-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK5-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK5-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK5-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK5-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK5:       omp_offload.failed:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52() #[[ATTR3]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK5:       omp_offload.cont:
// CHECK5-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK5-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK5-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK5-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK5-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK5-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK5-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK5-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK5-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK5-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK5-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK5-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK5-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK5-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK5-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK5-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK5-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK5-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK5-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK5:       omp_offload.failed3:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57() #[[ATTR3]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK5:       omp_offload.cont4:
// CHECK5-NEXT:    ret i32 0
//
//
// CHECK5-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
// CHECK5-SAME: () #[[ATTR6]] comdat {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK5-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK5-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK5-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK5-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK5-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK5-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK5-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK5-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK5-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK5-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK5-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK5-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK5:       omp_offload.failed:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52() #[[ATTR3]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK5:       omp_offload.cont:
// CHECK5-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK5-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK5-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK5-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK5-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK5-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK5-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK5-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK5-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK5-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK5-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK5-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK5-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK5-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK5-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK5-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK5-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK5-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK5-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK5:       omp_offload.failed3:
// CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57() #[[ATTR3]]
// CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK5:       omp_offload.cont4:
// CHECK5-NEXT:    ret i32 0
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN1SD1Ev
// CHECK5-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] comdat align 2 {
// 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 @_ZN1SD2Ev(ptr nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR3]]
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN1SC2El
// CHECK5-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR6]] comdat align 2 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK5-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK5-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52
// CHECK5-SAME: () #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined)
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_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_GRP24:![0-9]+]]
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 5), !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = 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_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27:![0-9]+]]
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK5-NEXT:    invoke void @_Z3foov()
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP27]]
// CHECK5:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK5-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
// CHECK5:       terminate.lpad:
// CHECK5-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            catch ptr null
// CHECK5-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP27]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57
// CHECK5-SAME: () #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined)
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_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_GRP30:![0-9]+]]
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 23), !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = 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_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33:![0-9]+]]
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK5-NEXT:    invoke void @_Z3foov()
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP33]]
// CHECK5:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK5-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
// CHECK5:       terminate.lpad:
// CHECK5-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            catch ptr null
// CHECK5-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP33]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52
// CHECK5-SAME: () #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined)
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_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_GRP36:![0-9]+]]
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 1), !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK5:       omp.inner.for.inc:
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = 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_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39:![0-9]+]]
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK5-NEXT:    invoke void @_Z3foov()
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP39]]
// CHECK5:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK5-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
// CHECK5:       terminate.lpad:
// CHECK5-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            catch ptr null
// CHECK5-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP39]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57
// CHECK5-SAME: () #[[ATTR2]] {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined)
// CHECK5-NEXT:    ret void
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_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_GRP42:![0-9]+]]
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    invoke void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 23)
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP42]]
// CHECK5:       invoke.cont:
// CHECK5-NEXT:    [[CALL:%.*]] = invoke signext i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]])
// CHECK5-NEXT:            to label [[INVOKE_CONT2:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP42]]
// CHECK5:       invoke.cont2:
// CHECK5-NEXT:    [[TMP7:%.*]] = sext i8 [[CALL]] to i32
// CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP7]]), !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR3]], !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group [[ACC_GRP42]]
// 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_GRP42]]
// CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK5-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK5-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
// CHECK5:       terminate.lpad:
// CHECK5-NEXT:    [[TMP16:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            catch ptr null
// CHECK5-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP16]], 0
// CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP17]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP42]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined
// CHECK5-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK5-NEXT:  entry:
// CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = 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_STRIDE:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK5-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK5:       cond.true:
// CHECK5-NEXT:    br label [[COND_END:%.*]]
// CHECK5:       cond.false:
// CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    br label [[COND_END]]
// CHECK5:       cond.end:
// CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK5:       omp.inner.for.cond:
// CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45:![0-9]+]]
// CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK5:       omp.inner.for.body:
// CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK5-NEXT:    invoke void @_Z3foov()
// CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP45]]
// CHECK5:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
// CHECK5:       omp.inner.for.end:
// CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK5:       omp.loop.exit:
// CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK5-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK5:       .omp.final.then:
// CHECK5-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK5:       .omp.final.done:
// CHECK5-NEXT:    ret void
// CHECK5:       terminate.lpad:
// CHECK5-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK5-NEXT:            catch ptr null
// CHECK5-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP45]]
// CHECK5-NEXT:    unreachable
//
//
// CHECK5-LABEL: define {{[^@]+}}@_ZN1SD2Ev
// CHECK5-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] comdat align 2 {
// 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
//
//
// CHECK9-LABEL: define {{[^@]+}}@main
// CHECK9-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK9-NEXT:    [[A:%.*]] = alloca i8, align 1
// CHECK9-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[EHSELECTOR_SLOT:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    call void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[S]], i64 0)
// CHECK9-NEXT:    [[CALL:%.*]] = invoke i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[S]])
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK9:       invoke.cont:
// CHECK9-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
// CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = 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_l68.region_id, ptr [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK9:       omp_offload.failed:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR3:[0-9]+]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       lpad:
// CHECK9-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            cleanup
// CHECK9-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK9-NEXT:    store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
// CHECK9-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
// CHECK9-NEXT:    store i32 [[TMP17]], ptr [[EHSELECTOR_SLOT]], align 4
// CHECK9-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK9-NEXT:    br label [[EH_RESUME:%.*]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[TMP18:%.*]] = load i8, ptr [[A]], align 1
// CHECK9-NEXT:    store i8 [[TMP18]], ptr [[A_CASTED]], align 1
// CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[A_CASTED]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP19]], ptr [[TMP20]], align 8
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    store i64 [[TMP19]], ptr [[TMP21]], align 8
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK9-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP25]], align 4
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK9-NEXT:    store i32 1, ptr [[TMP26]], align 4
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK9-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
// CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK9-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
// CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK9-NEXT:    store ptr @.offload_sizes, ptr [[TMP29]], align 8
// CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP30]], align 8
// CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP31]], align 8
// CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP32]], align 8
// CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK9-NEXT:    store i64 100, ptr [[TMP33]], align 8
// CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP34]], align 8
// CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
// CHECK9-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
// CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP37]], align 4
// CHECK9-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.region_id, ptr [[KERNEL_ARGS2]])
// CHECK9-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
// CHECK9-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK9:       omp_offload.failed3:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74(i64 [[TMP19]]) #[[ATTR3]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK9:       omp_offload.cont4:
// CHECK9-NEXT:    [[TMP40:%.*]] = load i8, ptr [[A]], align 1
// CHECK9-NEXT:    [[CONV:%.*]] = sext i8 [[TMP40]] to i32
// CHECK9-NEXT:    [[CALL6:%.*]] = invoke i32 @_Z5tmainIcLi5EEiv()
// CHECK9-NEXT:            to label [[INVOKE_CONT5:%.*]] unwind label [[LPAD]]
// CHECK9:       invoke.cont5:
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CALL6]]
// CHECK9-NEXT:    [[CALL8:%.*]] = invoke i32 @_Z5tmainI1SLi1EEiv()
// CHECK9-NEXT:            to label [[INVOKE_CONT7:%.*]] unwind label [[LPAD]]
// CHECK9:       invoke.cont7:
// CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD]], [[CALL8]]
// CHECK9-NEXT:    store i32 [[ADD9]], ptr [[RETVAL]], align 4
// CHECK9-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK9-NEXT:    [[TMP41:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK9-NEXT:    ret i32 [[TMP41]]
// CHECK9:       eh.resume:
// CHECK9-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
// CHECK9-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
// CHECK9-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
// CHECK9-NEXT:    [[LPAD_VAL10:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
// CHECK9-NEXT:    resume { ptr, i32 } [[LPAD_VAL10]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SC1El
// CHECK9-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK9-NEXT:    call void @_ZN1SC2El(ptr nonnull align 8 dereferenceable(24) [[THIS1]], i64 [[TMP0]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1ScvcEv
// CHECK9-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[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 @_Z8mayThrowv()
// CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
// CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
// CHECK9-NEXT:    ret i8 [[CONV]]
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68
// CHECK9-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined)
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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]], 99
// 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 [ 99, [[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_GRP9:![0-9]+]]
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 2), !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![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:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, 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_l68.omp_outlined.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV1]], 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]], 99
// 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 [ 99, [[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_GRP13:![0-9]+]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP2]], 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_GRP13]]
// 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_GRP13]]
// CHECK9-NEXT:    invoke void @_Z3foov()
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP13]]
// CHECK9:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![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:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
// CHECK9:       terminate.lpad:
// CHECK9-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            catch ptr null
// CHECK9-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK9-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7:[0-9]+]], !llvm.access.group [[ACC_GRP13]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK9-LABEL: define {{[^@]+}}@__clang_call_terminate
// CHECK9-SAME: (ptr [[TMP0:%.*]]) #[[ATTR5:[0-9]+]] comdat {
// CHECK9-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR3]]
// CHECK9-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74
// CHECK9-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined, ptr [[A_ADDR]])
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 1 dereferenceable(1) [[A:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
// 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:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    br label [[COND_END]]
// CHECK9:       cond.end:
// CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK9:       omp.inner.for.cond:
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP0]], align 1, !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    [[TMP9:%.*]] = sext i8 [[TMP8]] to i32
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP2]], i32 [[TMP9]]), !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]]), !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![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 [[TMP2]])
// 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 100, 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_l74.omp_outlined.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV1]], 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]], 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]], 99
// 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 [ 99, [[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_GRP21:![0-9]+]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP2]], 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_GRP21]]
// 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_GRP21]]
// CHECK9-NEXT:    invoke void @_Z3foov()
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP21]]
// CHECK9:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![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:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
// CHECK9:       terminate.lpad:
// CHECK9-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            catch ptr null
// CHECK9-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK9-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP21]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
// CHECK9-SAME: () #[[ATTR6:[0-9]+]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK9:       omp_offload.failed:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52() #[[ATTR3]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK9-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK9-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK9-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK9-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK9-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK9:       omp_offload.failed3:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57() #[[ATTR3]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK9:       omp_offload.cont4:
// CHECK9-NEXT:    ret i32 0
//
//
// CHECK9-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
// CHECK9-SAME: () #[[ATTR6]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK9-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK9-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK9-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK9:       omp_offload.failed:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52() #[[ATTR3]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK9:       omp_offload.cont:
// CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK9-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK9-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK9-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK9-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK9-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK9-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK9-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK9-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK9-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK9-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK9-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK9-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK9-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK9-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK9:       omp_offload.failed3:
// CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57() #[[ATTR3]]
// CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK9:       omp_offload.cont4:
// CHECK9-NEXT:    ret i32 0
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SD1Ev
// CHECK9-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] 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 @_ZN1SD2Ev(ptr nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR3]]
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SC2El
// CHECK9-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR6]] comdat {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK9-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK9-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK9-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK9-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@_ZN1SD2Ev
// CHECK9-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] 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 {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52
// CHECK9-SAME: () #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined)
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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]], 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]], 99
// 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 [ 99, [[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_GRP24:![0-9]+]]
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 5), !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![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:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, 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]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV1]], 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]], 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]], 99
// 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 [ 99, [[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_GRP27:![0-9]+]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP2]], 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_GRP27]]
// 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_GRP27]]
// CHECK9-NEXT:    invoke void @_Z3foov()
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP27]]
// CHECK9:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![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:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
// CHECK9:       terminate.lpad:
// CHECK9-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            catch ptr null
// CHECK9-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK9-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP27]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57
// CHECK9-SAME: () #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined)
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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]], 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]], 99
// 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 [ 99, [[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_GRP30:![0-9]+]]
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 23), !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![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:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, 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]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV1]], 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]], 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]], 99
// 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 [ 99, [[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_GRP33:![0-9]+]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP2]], 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_GRP33]]
// 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_GRP33]]
// CHECK9-NEXT:    invoke void @_Z3foov()
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP33]]
// CHECK9:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![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:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
// CHECK9:       terminate.lpad:
// CHECK9-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            catch ptr null
// CHECK9-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK9-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP33]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52
// CHECK9-SAME: () #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined)
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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]], 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]], 99
// 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 [ 99, [[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_GRP36:![0-9]+]]
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 1), !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![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:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, 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]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV1]], 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]], 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]], 99
// 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 [ 99, [[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_GRP39:![0-9]+]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP2]], 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_GRP39]]
// 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_GRP39]]
// CHECK9-NEXT:    invoke void @_Z3foov()
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP39]]
// CHECK9:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![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:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
// CHECK9:       terminate.lpad:
// CHECK9-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            catch ptr null
// CHECK9-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK9-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP39]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57
// CHECK9-SAME: () #[[ATTR2]] {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined)
// CHECK9-NEXT:    ret void
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK9-NEXT:  entry:
// CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], 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 i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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]], 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]], 99
// 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 [ 99, [[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_GRP42:![0-9]+]]
// CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK9:       omp.inner.for.body:
// CHECK9-NEXT:    invoke void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 23)
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP42]]
// CHECK9:       invoke.cont:
// CHECK9-NEXT:    [[CALL:%.*]] = invoke i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]])
// CHECK9-NEXT:            to label [[INVOKE_CONT2:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP42]]
// CHECK9:       invoke.cont2:
// CHECK9-NEXT:    [[TMP7:%.*]] = sext i8 [[CALL]] to i32
// CHECK9-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP7]]), !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR3]], !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK9:       omp.inner.for.inc:
// CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![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:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
// CHECK9:       terminate.lpad:
// CHECK9-NEXT:    [[TMP16:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            catch ptr null
// CHECK9-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP16]], 0
// CHECK9-NEXT:    call void @__clang_call_terminate(ptr [[TMP17]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP42]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined
// CHECK9-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// 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:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK9-NEXT:    [[TMP:%.*]] = 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:    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 i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 99, 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:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK9-NEXT:    store i32 [[CONV1]], 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]], 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]], 99
// 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 [ 99, [[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_GRP45:![0-9]+]]
// CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK9-NEXT:    br i1 [[CMP2]], 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_GRP45]]
// 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_GRP45]]
// CHECK9-NEXT:    invoke void @_Z3foov()
// CHECK9-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP45]]
// CHECK9:       invoke.cont:
// 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:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![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:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK9:       .omp.final.then:
// CHECK9-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK9:       .omp.final.done:
// CHECK9-NEXT:    ret void
// CHECK9:       terminate.lpad:
// CHECK9-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK9-NEXT:            catch ptr null
// CHECK9-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK9-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP45]]
// CHECK9-NEXT:    unreachable
//
//
// CHECK11-LABEL: define {{[^@]+}}@main
// CHECK11-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK11-NEXT:    [[A:%.*]] = alloca i8, align 1
// CHECK11-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[EHSELECTOR_SLOT:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I7:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    call void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[S]], i64 0)
// CHECK11-NEXT:    [[CALL:%.*]] = invoke i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[S]])
// CHECK11-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK11:       invoke.cont:
// CHECK11-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2:![0-9]+]]
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK11-NEXT:    invoke void @_Z3foov()
// CHECK11-NEXT:            to label [[INVOKE_CONT1:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP2]]
// CHECK11:       invoke.cont1:
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK11-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP2]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
// CHECK11:       lpad:
// CHECK11-NEXT:    [[TMP5:%.*]] = landingpad { ptr, i32 }
// CHECK11-NEXT:            cleanup
// CHECK11-NEXT:    [[TMP6:%.*]] = extractvalue { ptr, i32 } [[TMP5]], 0
// CHECK11-NEXT:    store ptr [[TMP6]], ptr [[EXN_SLOT]], align 8
// CHECK11-NEXT:    [[TMP7:%.*]] = extractvalue { ptr, i32 } [[TMP5]], 1
// CHECK11-NEXT:    store i32 [[TMP7]], ptr [[EHSELECTOR_SLOT]], align 4
// CHECK11-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR5:[0-9]+]]
// CHECK11-NEXT:    br label [[EH_RESUME:%.*]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB4]], align 4
// CHECK11-NEXT:    store i32 99, ptr [[DOTOMP_UB5]], align 4
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_LB4]], align 4
// CHECK11-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_IV6]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
// CHECK11:       omp.inner.for.cond8:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6:![0-9]+]]
// CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB5]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK11-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
// CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
// CHECK11:       omp.inner.for.body10:
// CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK11-NEXT:    [[MUL11:%.*]] = mul nsw i32 [[TMP11]], 1
// CHECK11-NEXT:    [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
// CHECK11-NEXT:    store i32 [[ADD12]], ptr [[I7]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK11-NEXT:    invoke void @_Z3foov()
// CHECK11-NEXT:            to label [[INVOKE_CONT13:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP6]]
// CHECK11:       invoke.cont13:
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE14:%.*]]
// CHECK11:       omp.body.continue14:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC15:%.*]]
// CHECK11:       omp.inner.for.inc15:
// CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK11-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP12]], 1
// CHECK11-NEXT:    store i32 [[ADD16]], ptr [[DOTOMP_IV6]], align 4, !llvm.access.group [[ACC_GRP6]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
// CHECK11:       omp.inner.for.end17:
// CHECK11-NEXT:    store i32 100, ptr [[I7]], align 4
// CHECK11-NEXT:    [[TMP13:%.*]] = load i8, ptr [[A]], align 1
// CHECK11-NEXT:    [[CONV:%.*]] = sext i8 [[TMP13]] to i32
// CHECK11-NEXT:    [[CALL19:%.*]] = invoke i32 @_Z5tmainIcLi5EEiv()
// CHECK11-NEXT:            to label [[INVOKE_CONT18:%.*]] unwind label [[LPAD]]
// CHECK11:       invoke.cont18:
// CHECK11-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV]], [[CALL19]]
// CHECK11-NEXT:    [[CALL22:%.*]] = invoke i32 @_Z5tmainI1SLi1EEiv()
// CHECK11-NEXT:            to label [[INVOKE_CONT21:%.*]] unwind label [[LPAD]]
// CHECK11:       invoke.cont21:
// CHECK11-NEXT:    [[ADD23:%.*]] = add nsw i32 [[ADD20]], [[CALL22]]
// CHECK11-NEXT:    store i32 [[ADD23]], ptr [[RETVAL]], align 4
// CHECK11-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR5]]
// CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK11-NEXT:    ret i32 [[TMP14]]
// CHECK11:       eh.resume:
// CHECK11-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
// CHECK11-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
// CHECK11-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
// CHECK11-NEXT:    [[LPAD_VAL24:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
// CHECK11-NEXT:    resume { ptr, i32 } [[LPAD_VAL24]]
// CHECK11:       terminate.lpad:
// CHECK11-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
// CHECK11-NEXT:            catch ptr null
// CHECK11-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK11-NEXT:    call void @__clang_call_terminate(ptr [[TMP16]]) #[[ATTR6:[0-9]+]], !llvm.access.group [[ACC_GRP2]]
// CHECK11-NEXT:    unreachable
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SC1El
// CHECK11-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK11-NEXT:    call void @_ZN1SC2El(ptr nonnull align 8 dereferenceable(24) [[THIS1]], i64 [[TMP0]])
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1ScvcEv
// CHECK11-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[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 @_Z8mayThrowv()
// CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
// CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
// CHECK11-NEXT:    ret i8 [[CONV]]
//
//
// CHECK11-LABEL: define {{[^@]+}}@__clang_call_terminate
// CHECK11-SAME: (ptr [[TMP0:%.*]]) #[[ATTR3:[0-9]+]] comdat {
// CHECK11-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR5]]
// CHECK11-NEXT:    call void @_ZSt9terminatev() #[[ATTR6]]
// CHECK11-NEXT:    unreachable
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
// CHECK11-SAME: () #[[ATTR4:[0-9]+]] comdat personality ptr @__gxx_personality_v0 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK11-NEXT:    invoke void @_Z3foov()
// CHECK11-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP9]]
// CHECK11:       invoke.cont:
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK11-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK11-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB3]], align 4
// CHECK11-NEXT:    store i32 99, ptr [[DOTOMP_UB4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB3]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV5]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK11:       omp.inner.for.cond7:
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12:![0-9]+]]
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB4]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK11-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
// CHECK11:       omp.inner.for.body9:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK11-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK11-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK11-NEXT:    store i32 [[ADD11]], ptr [[I6]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK11-NEXT:    invoke void @_Z3foov()
// CHECK11-NEXT:            to label [[INVOKE_CONT12:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP12]]
// CHECK11:       invoke.cont12:
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE13:%.*]]
// CHECK11:       omp.body.continue13:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC14:%.*]]
// CHECK11:       omp.inner.for.inc14:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK11-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    store i32 [[ADD15]], ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP12]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP13:![0-9]+]]
// CHECK11:       omp.inner.for.end16:
// CHECK11-NEXT:    store i32 100, ptr [[I6]], align 4
// CHECK11-NEXT:    ret i32 0
// CHECK11:       terminate.lpad:
// CHECK11-NEXT:    [[TMP10:%.*]] = landingpad { ptr, i32 }
// CHECK11-NEXT:            catch ptr null
// CHECK11-NEXT:    [[TMP11:%.*]] = extractvalue { ptr, i32 } [[TMP10]], 0
// CHECK11-NEXT:    call void @__clang_call_terminate(ptr [[TMP11]]) #[[ATTR6]], !llvm.access.group [[ACC_GRP9]]
// CHECK11-NEXT:    unreachable
//
//
// CHECK11-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
// CHECK11-SAME: () #[[ATTR4]] comdat personality ptr @__gxx_personality_v0 {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    [[I6:%.*]] = alloca i32, align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK11-NEXT:    store i32 [[TMP0]], ptr [[DOTOMP_IV]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK11:       omp.inner.for.cond:
// CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
// CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK11-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
// CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK11:       omp.inner.for.body:
// CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
// CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK11-NEXT:    invoke void @_Z3foov()
// CHECK11-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP15]]
// CHECK11:       invoke.cont:
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK11:       omp.body.continue:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK11:       omp.inner.for.inc:
// CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK11-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
// CHECK11-NEXT:    store i32 [[ADD1]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
// CHECK11:       omp.inner.for.end:
// CHECK11-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB3]], align 4
// CHECK11-NEXT:    store i32 99, ptr [[DOTOMP_UB4]], align 4
// CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB3]], align 4
// CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV5]], align 4
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
// CHECK11:       omp.inner.for.cond7:
// CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
// CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB4]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK11-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK11-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
// CHECK11:       omp.inner.for.body9:
// CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK11-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
// CHECK11-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
// CHECK11-NEXT:    store i32 [[ADD11]], ptr [[I6]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK11-NEXT:    invoke void @_Z3foov()
// CHECK11-NEXT:            to label [[INVOKE_CONT12:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP18]]
// CHECK11:       invoke.cont12:
// CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE13:%.*]]
// CHECK11:       omp.body.continue13:
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC14:%.*]]
// CHECK11:       omp.inner.for.inc14:
// CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK11-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP9]], 1
// CHECK11-NEXT:    store i32 [[ADD15]], ptr [[DOTOMP_IV5]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK11:       omp.inner.for.end16:
// CHECK11-NEXT:    store i32 100, ptr [[I6]], align 4
// CHECK11-NEXT:    ret i32 0
// CHECK11:       terminate.lpad:
// CHECK11-NEXT:    [[TMP10:%.*]] = landingpad { ptr, i32 }
// CHECK11-NEXT:            catch ptr null
// CHECK11-NEXT:    [[TMP11:%.*]] = extractvalue { ptr, i32 } [[TMP10]], 0
// CHECK11-NEXT:    call void @__clang_call_terminate(ptr [[TMP11]]) #[[ATTR6]], !llvm.access.group [[ACC_GRP15]]
// CHECK11-NEXT:    unreachable
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SD1Ev
// CHECK11-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR4]] 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 @_ZN1SD2Ev(ptr nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR5]]
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SC2El
// CHECK11-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR4]] comdat {
// CHECK11-NEXT:  entry:
// CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK11-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK11-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK11-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK11-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK11-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
// CHECK11-NEXT:    ret void
//
//
// CHECK11-LABEL: define {{[^@]+}}@_ZN1SD2Ev
// CHECK11-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR4]] 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
//
//
// CHECK13-LABEL: define {{[^@]+}}@main
// CHECK13-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK13-NEXT:    [[A:%.*]] = alloca i8, align 1
// CHECK13-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[EHSELECTOR_SLOT:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK13-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
// CHECK13-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
// CHECK13-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[RETVAL]], align 4
// CHECK13-NEXT:    call void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[S]], i64 0)
// CHECK13-NEXT:    [[CALL:%.*]] = invoke i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[S]])
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
// CHECK13-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK13-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK13-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK13-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK13-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK13-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK13-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK13-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK13-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK13-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK13-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK13-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK13-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK13-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK13-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = 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_l68.region_id, ptr [[KERNEL_ARGS]])
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK13:       omp_offload.failed:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68() #[[ATTR3:[0-9]+]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK13:       lpad:
// CHECK13-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            cleanup
// CHECK13-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
// CHECK13-NEXT:    store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
// CHECK13-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
// CHECK13-NEXT:    store i32 [[TMP17]], ptr [[EHSELECTOR_SLOT]], align 4
// CHECK13-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[EH_RESUME:%.*]]
// CHECK13:       omp_offload.cont:
// CHECK13-NEXT:    [[TMP18:%.*]] = load i8, ptr [[A]], align 1
// CHECK13-NEXT:    store i8 [[TMP18]], ptr [[A_CASTED]], align 1
// CHECK13-NEXT:    [[TMP19:%.*]] = load i64, ptr [[A_CASTED]], align 8
// CHECK13-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    store i64 [[TMP19]], ptr [[TMP20]], align 8
// CHECK13-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    store i64 [[TMP19]], ptr [[TMP21]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
// CHECK13-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK13-NEXT:    store i32 3, ptr [[TMP25]], align 4
// CHECK13-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK13-NEXT:    store i32 1, ptr [[TMP26]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK13-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
// CHECK13-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK13-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
// CHECK13-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK13-NEXT:    store ptr @.offload_sizes, ptr [[TMP29]], align 8
// CHECK13-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK13-NEXT:    store ptr @.offload_maptypes, ptr [[TMP30]], align 8
// CHECK13-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK13-NEXT:    store ptr null, ptr [[TMP31]], align 8
// CHECK13-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK13-NEXT:    store ptr null, ptr [[TMP32]], align 8
// CHECK13-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK13-NEXT:    store i64 100, ptr [[TMP33]], align 8
// CHECK13-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK13-NEXT:    store i64 0, ptr [[TMP34]], align 8
// CHECK13-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
// CHECK13-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
// CHECK13-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK13-NEXT:    store i32 0, ptr [[TMP37]], align 4
// CHECK13-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.region_id, ptr [[KERNEL_ARGS2]])
// CHECK13-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
// CHECK13-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK13:       omp_offload.failed3:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74(i64 [[TMP19]]) #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK13:       omp_offload.cont4:
// CHECK13-NEXT:    [[TMP40:%.*]] = load i8, ptr [[A]], align 1
// CHECK13-NEXT:    [[CONV:%.*]] = sext i8 [[TMP40]] to i32
// CHECK13-NEXT:    [[CALL6:%.*]] = invoke i32 @_Z5tmainIcLi5EEiv()
// CHECK13-NEXT:            to label [[INVOKE_CONT5:%.*]] unwind label [[LPAD]]
// CHECK13:       invoke.cont5:
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CALL6]]
// CHECK13-NEXT:    [[CALL8:%.*]] = invoke i32 @_Z5tmainI1SLi1EEiv()
// CHECK13-NEXT:            to label [[INVOKE_CONT7:%.*]] unwind label [[LPAD]]
// CHECK13:       invoke.cont7:
// CHECK13-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD]], [[CALL8]]
// CHECK13-NEXT:    store i32 [[ADD9]], ptr [[RETVAL]], align 4
// CHECK13-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR3]]
// CHECK13-NEXT:    [[TMP41:%.*]] = load i32, ptr [[RETVAL]], align 4
// CHECK13-NEXT:    ret i32 [[TMP41]]
// CHECK13:       eh.resume:
// CHECK13-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
// CHECK13-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
// CHECK13-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
// CHECK13-NEXT:    [[LPAD_VAL10:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
// CHECK13-NEXT:    resume { ptr, i32 } [[LPAD_VAL10]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SC1El
// CHECK13-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK13-NEXT:    call void @_ZN1SC2El(ptr nonnull align 8 dereferenceable(24) [[THIS1]], i64 [[TMP0]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1ScvcEv
// CHECK13-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[ATTR1]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    call void @_Z8mayThrowv()
// CHECK13-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
// CHECK13-NEXT:    ret i8 [[CONV]]
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68
// CHECK13-SAME: () #[[ATTR2:[0-9]+]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined)
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK13-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)
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 2), !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.omp_outlined.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK13-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)
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13:![0-9]+]]
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK13-NEXT:    invoke void @_Z3foov()
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP13]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP13]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK13-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
// CHECK13:       terminate.lpad:
// CHECK13-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            catch ptr null
// CHECK13-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK13-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7:[0-9]+]], !llvm.access.group [[ACC_GRP13]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@__clang_call_terminate
// CHECK13-SAME: (ptr [[TMP0:%.*]]) #[[ATTR5:[0-9]+]] comdat {
// CHECK13-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR3]]
// CHECK13-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74
// CHECK13-SAME: (i64 [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined, ptr [[A_ADDR]])
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 1 dereferenceable(1) [[A:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18:![0-9]+]]
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP0]], align 1, !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    [[TMP9:%.*]] = sext i8 [[TMP8]] to i32
// CHECK13-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP2]], i32 [[TMP9]]), !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]]), !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP18]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
// CHECK13-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
// CHECK13-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l74.omp_outlined.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21:![0-9]+]]
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK13-NEXT:    invoke void @_Z3foov()
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP21]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP21]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK13-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
// CHECK13:       terminate.lpad:
// CHECK13-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            catch ptr null
// CHECK13-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK13-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP21]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
// CHECK13-SAME: () #[[ATTR6:[0-9]+]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK13-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK13-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK13-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK13-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK13-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK13-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK13-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK13-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK13-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK13-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK13-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK13-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK13-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK13-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK13-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK13:       omp_offload.failed:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52() #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK13:       omp_offload.cont:
// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK13-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK13-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK13-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK13-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK13-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK13-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK13-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK13-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK13-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK13-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK13-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK13-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK13-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK13-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK13-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK13-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK13-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK13-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK13-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK13-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK13-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK13:       omp_offload.failed3:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57() #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK13:       omp_offload.cont4:
// CHECK13-NEXT:    ret i32 0
//
//
// CHECK13-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
// CHECK13-SAME: () #[[ATTR6]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
// CHECK13-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
// CHECK13-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
// CHECK13-NEXT:    store i32 3, ptr [[TMP0]], align 4
// CHECK13-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
// CHECK13-NEXT:    store i32 0, ptr [[TMP1]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
// CHECK13-NEXT:    store ptr null, ptr [[TMP2]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
// CHECK13-NEXT:    store ptr null, ptr [[TMP3]], align 8
// CHECK13-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
// CHECK13-NEXT:    store ptr null, ptr [[TMP4]], align 8
// CHECK13-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
// CHECK13-NEXT:    store ptr null, ptr [[TMP5]], align 8
// CHECK13-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
// CHECK13-NEXT:    store ptr null, ptr [[TMP6]], align 8
// CHECK13-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
// CHECK13-NEXT:    store ptr null, ptr [[TMP7]], align 8
// CHECK13-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
// CHECK13-NEXT:    store i64 100, ptr [[TMP8]], align 8
// CHECK13-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
// CHECK13-NEXT:    store i64 0, ptr [[TMP9]], align 8
// CHECK13-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
// CHECK13-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
// CHECK13-NEXT:    store i32 0, ptr [[TMP12]], align 4
// CHECK13-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.region_id, ptr [[KERNEL_ARGS]])
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
// CHECK13:       omp_offload.failed:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52() #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT]]
// CHECK13:       omp_offload.cont:
// CHECK13-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
// CHECK13-NEXT:    store i32 3, ptr [[TMP15]], align 4
// CHECK13-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
// CHECK13-NEXT:    store i32 0, ptr [[TMP16]], align 4
// CHECK13-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
// CHECK13-NEXT:    store ptr null, ptr [[TMP17]], align 8
// CHECK13-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
// CHECK13-NEXT:    store ptr null, ptr [[TMP18]], align 8
// CHECK13-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
// CHECK13-NEXT:    store ptr null, ptr [[TMP19]], align 8
// CHECK13-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
// CHECK13-NEXT:    store ptr null, ptr [[TMP20]], align 8
// CHECK13-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
// CHECK13-NEXT:    store ptr null, ptr [[TMP21]], align 8
// CHECK13-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
// CHECK13-NEXT:    store ptr null, ptr [[TMP22]], align 8
// CHECK13-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
// CHECK13-NEXT:    store i64 100, ptr [[TMP23]], align 8
// CHECK13-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
// CHECK13-NEXT:    store i64 0, ptr [[TMP24]], align 8
// CHECK13-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
// CHECK13-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
// CHECK13-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
// CHECK13-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
// CHECK13-NEXT:    store i32 0, ptr [[TMP27]], align 4
// CHECK13-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.region_id, ptr [[KERNEL_ARGS2]])
// CHECK13-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
// CHECK13-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
// CHECK13:       omp_offload.failed3:
// CHECK13-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57() #[[ATTR3]]
// CHECK13-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
// CHECK13:       omp_offload.cont4:
// CHECK13-NEXT:    ret i32 0
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SD1Ev
// CHECK13-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    call void @_ZN1SD2Ev(ptr nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR3]]
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SC2El
// CHECK13-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 [[A:%.*]]) unnamed_addr #[[ATTR6]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
// CHECK13-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52
// CHECK13-SAME: () #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined)
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24:![0-9]+]]
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 5), !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP24]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l52.omp_outlined.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27:![0-9]+]]
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK13-NEXT:    invoke void @_Z3foov()
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP27]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP27]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK13-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
// CHECK13:       terminate.lpad:
// CHECK13-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            catch ptr null
// CHECK13-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK13-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP27]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57
// CHECK13-SAME: () #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined)
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30:![0-9]+]]
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 23), !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP30]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l57.omp_outlined.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33:![0-9]+]]
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK13-NEXT:    invoke void @_Z3foov()
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP33]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP33]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK13-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
// CHECK13:       terminate.lpad:
// CHECK13-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            catch ptr null
// CHECK13-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK13-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP33]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52
// CHECK13-SAME: () #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined)
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36:![0-9]+]]
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 1), !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP36]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
// CHECK13-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l52.omp_outlined.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39:![0-9]+]]
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK13-NEXT:    invoke void @_Z3foov()
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP39]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP39]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK13-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
// CHECK13:       terminate.lpad:
// CHECK13-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            catch ptr null
// CHECK13-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK13-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP39]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57
// CHECK13-SAME: () #[[ATTR2]] {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined)
// CHECK13-NEXT:    ret void
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 8
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42:![0-9]+]]
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
// CHECK13-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    invoke void @_ZN1SC1El(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 23)
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP42]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    [[CALL:%.*]] = invoke i8 @_ZN1ScvcEv(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]])
// CHECK13-NEXT:            to label [[INVOKE_CONT2:%.*]] unwind label [[TERMINATE_LPAD]], !llvm.access.group [[ACC_GRP42]]
// CHECK13:       invoke.cont2:
// CHECK13-NEXT:    [[TMP7:%.*]] = sext i8 [[CALL]] to i32
// CHECK13-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP7]]), !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    call void @_ZN1SD1Ev(ptr nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR3]], !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
// CHECK13-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
// CHECK13-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
// CHECK13-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
// CHECK13:       terminate.lpad:
// CHECK13-NEXT:    [[TMP16:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            catch ptr null
// CHECK13-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP16]], 0
// CHECK13-NEXT:    call void @__clang_call_terminate(ptr [[TMP17]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP42]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l57.omp_outlined.omp_outlined
// CHECK13-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR2]] personality ptr @__gxx_personality_v0 {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
// CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
// CHECK13-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
// CHECK13-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
// CHECK13-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
// CHECK13-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
// CHECK13-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK13-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
// CHECK13-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK13-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
// CHECK13-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK13:       cond.true:
// CHECK13-NEXT:    br label [[COND_END:%.*]]
// CHECK13:       cond.false:
// CHECK13-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    br label [[COND_END]]
// CHECK13:       cond.end:
// CHECK13-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK13-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
// CHECK13-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
// CHECK13-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK13:       omp.inner.for.cond:
// CHECK13-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45:![0-9]+]]
// CHECK13-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK13-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK13:       omp.inner.for.body:
// CHECK13-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK13-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK13-NEXT:    invoke void @_Z3foov()
// CHECK13-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !llvm.access.group [[ACC_GRP45]]
// CHECK13:       invoke.cont:
// CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK13:       omp.body.continue:
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK13:       omp.inner.for.inc:
// CHECK13-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK13-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK13-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP45]]
// CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
// CHECK13:       omp.inner.for.end:
// CHECK13-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
// CHECK13:       omp.loop.exit:
// CHECK13-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
// CHECK13-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
// CHECK13-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK13-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK13:       .omp.final.then:
// CHECK13-NEXT:    store i32 100, ptr [[I]], align 4
// CHECK13-NEXT:    br label [[DOTOMP_FINAL_DONE]]
// CHECK13:       .omp.final.done:
// CHECK13-NEXT:    ret void
// CHECK13:       terminate.lpad:
// CHECK13-NEXT:    [[TMP13:%.*]] = landingpad { ptr, i32 }
// CHECK13-NEXT:            catch ptr null
// CHECK13-NEXT:    [[TMP14:%.*]] = extractvalue { ptr, i32 } [[TMP13]], 0
// CHECK13-NEXT:    call void @__clang_call_terminate(ptr [[TMP14]]) #[[ATTR7]], !llvm.access.group [[ACC_GRP45]]
// CHECK13-NEXT:    unreachable
//
//
// CHECK13-LABEL: define {{[^@]+}}@_ZN1SD2Ev
// CHECK13-SAME: (ptr nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR6]] comdat {
// CHECK13-NEXT:  entry:
// CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
// CHECK13-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
// CHECK13-NEXT:    ret void
//