clang/test/OpenMP/teams_distribute_parallel_for_num_threads_codegen.cpp

   1 // 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 _
   2 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
   3 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -emit-pch -o %t %s
   4 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
   5
   6 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
   7 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -emit-pch -o %t %s
   8 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
   9
  10 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK5
  11 // RUN: %clang_cc1 -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -emit-pch -o %t %s
  12 // RUN: %clang_cc1 -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5
  13
  14 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  15 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -emit-pch -o %t %s
  16 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple powerpc64le-ibm-linux-gnu -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  17
  18 // expected-no-diagnostics
  19 #ifndef HEADER
  20 #define HEADER
  21
  22 typedef __INTPTR_TYPE__ intptr_t;
  23
  24
  25 void foo();
  26
  27 struct S {
  28   intptr_t a, b, c;
  29   S(intptr_t a) : a(a) {}
  30   operator char() { return a; }
  31   ~S() {}
  32 };
  33
  34 template <typename T, int C>
  35 int tmain() {
  36 #pragma omp target
  37 #pragma omp teams distribute parallel for num_threads(C)
  38   for (int i = 0; i < 100; i++)
  39     foo();
  40 #pragma omp target
  41 #pragma omp teams distribute parallel for num_threads(T(23))
  42   for (int i = 0; i < 100; i++)
  43     foo();
  44   return 0;
  45 }
  46
  47 int main() {
  48   S s(0);
  49   char a = s;
  50 #pragma omp target
  51 #pragma omp teams distribute parallel for num_threads(2)
  52   for (int i = 0; i < 100; i++) {
  53     foo();
  54   }
  55 #pragma omp target
  56
  57 #pragma omp teams distribute parallel for num_threads(a)
  58   for (int i = 0; i < 100; i++) {
  59     foo();
  60   }
  61   return a + tmain<char, 5>() + tmain<S, 1>();
  62 }
  63
  64 // tmain 5
  65
  66 // tmain 1
  67
  68
  69
  70
  71
  72
  73
  74
  75 #endif
  76 // CHECK1-LABEL: define {{[^@]+}}@main
  77 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
  78 // CHECK1-NEXT:  entry:
  79 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
  80 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
  81 // CHECK1-NEXT:    [[A:%.*]] = alloca i8, align 1
  82 // CHECK1-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
  83 // CHECK1-NEXT:    [[EHSELECTOR_SLOT:%.*]] = alloca i32, align 4
  84 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
  85 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
  86 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
  87 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
  88 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
  89 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
  90 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
  91 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
  92 // CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
  93 // CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
  94 // CHECK1-NEXT:    call void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[S]], i64 noundef 0)
  95 // CHECK1-NEXT:    [[CALL:%.*]] = invoke noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[S]])
  96 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
  97 // CHECK1:       invoke.cont:
  98 // CHECK1-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
  99 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 100 // CHECK1-NEXT:    store i32 3, ptr [[TMP0]], align 4
 101 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 102 // CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
 103 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 104 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 105 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 106 // CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
 107 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 108 // CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
 109 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 110 // CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
 111 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 112 // CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
 113 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 114 // CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
 115 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 116 // CHECK1-NEXT:    store i64 100, ptr [[TMP8]], align 8
 117 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 118 // CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
 119 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 120 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
 121 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 122 // CHECK1-NEXT:    store [3 x i32] [i32 2, i32 0, i32 0], ptr [[TMP11]], align 4
 123 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 124 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
 125 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 2, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.region_id, ptr [[KERNEL_ARGS]])
 126 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 127 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 128 // CHECK1:       omp_offload.failed:
 129 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50() #[[ATTR4:[0-9]+]]
 130 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 131 // CHECK1:       lpad:
 132 // CHECK1-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
 133 // CHECK1-NEXT:            cleanup
 134 // CHECK1-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
 135 // CHECK1-NEXT:    store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
 136 // CHECK1-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
 137 // CHECK1-NEXT:    store i32 [[TMP17]], ptr [[EHSELECTOR_SLOT]], align 4
 138 // CHECK1-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR4]]
 139 // CHECK1-NEXT:    br label [[EH_RESUME:%.*]]
 140 // CHECK1:       omp_offload.cont:
 141 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[A]], align 1
 142 // CHECK1-NEXT:    store i8 [[TMP18]], ptr [[A_CASTED]], align 1
 143 // CHECK1-NEXT:    [[TMP19:%.*]] = load i64, ptr [[A_CASTED]], align 8
 144 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 145 // CHECK1-NEXT:    store i64 [[TMP19]], ptr [[TMP20]], align 8
 146 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 147 // CHECK1-NEXT:    store i64 [[TMP19]], ptr [[TMP21]], align 8
 148 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 149 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 150 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 151 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 152 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8, ptr [[A]], align 1
 153 // CHECK1-NEXT:    store i8 [[TMP25]], ptr [[DOTCAPTURE_EXPR_]], align 1
 154 // CHECK1-NEXT:    [[TMP26:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
 155 // CHECK1-NEXT:    [[TMP27:%.*]] = zext i8 [[TMP26]] to i32
 156 // CHECK1-NEXT:    [[TMP28:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP27]], 0
 157 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
 158 // CHECK1-NEXT:    store i32 3, ptr [[TMP29]], align 4
 159 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
 160 // CHECK1-NEXT:    store i32 1, ptr [[TMP30]], align 4
 161 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
 162 // CHECK1-NEXT:    store ptr [[TMP23]], ptr [[TMP31]], align 8
 163 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
 164 // CHECK1-NEXT:    store ptr [[TMP24]], ptr [[TMP32]], align 8
 165 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
 166 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP33]], align 8
 167 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
 168 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP34]], align 8
 169 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
 170 // CHECK1-NEXT:    store ptr null, ptr [[TMP35]], align 8
 171 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
 172 // CHECK1-NEXT:    store ptr null, ptr [[TMP36]], align 8
 173 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
 174 // CHECK1-NEXT:    store i64 100, ptr [[TMP37]], align 8
 175 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
 176 // CHECK1-NEXT:    store i64 0, ptr [[TMP38]], align 8
 177 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
 178 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP39]], align 4
 179 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
 180 // CHECK1-NEXT:    store [3 x i32] [[TMP28]], ptr [[TMP40]], align 4
 181 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
 182 // CHECK1-NEXT:    store i32 0, ptr [[TMP41]], align 4
 183 // CHECK1-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 [[TMP27]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.region_id, ptr [[KERNEL_ARGS2]])
 184 // CHECK1-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
 185 // CHECK1-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
 186 // CHECK1:       omp_offload.failed3:
 187 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55(i64 [[TMP19]]) #[[ATTR4]]
 188 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
 189 // CHECK1:       omp_offload.cont4:
 190 // CHECK1-NEXT:    [[TMP44:%.*]] = load i8, ptr [[A]], align 1
 191 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP44]] to i32
 192 // CHECK1-NEXT:    [[CALL6:%.*]] = invoke noundef signext i32 @_Z5tmainIcLi5EEiv()
 193 // CHECK1-NEXT:            to label [[INVOKE_CONT5:%.*]] unwind label [[LPAD]]
 194 // CHECK1:       invoke.cont5:
 195 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CALL6]]
 196 // CHECK1-NEXT:    [[CALL8:%.*]] = invoke noundef signext i32 @_Z5tmainI1SLi1EEiv()
 197 // CHECK1-NEXT:            to label [[INVOKE_CONT7:%.*]] unwind label [[LPAD]]
 198 // CHECK1:       invoke.cont7:
 199 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD]], [[CALL8]]
 200 // CHECK1-NEXT:    store i32 [[ADD9]], ptr [[RETVAL]], align 4
 201 // CHECK1-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR4]]
 202 // CHECK1-NEXT:    [[TMP45:%.*]] = load i32, ptr [[RETVAL]], align 4
 203 // CHECK1-NEXT:    ret i32 [[TMP45]]
 204 // CHECK1:       eh.resume:
 205 // CHECK1-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
 206 // CHECK1-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
 207 // CHECK1-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
 208 // CHECK1-NEXT:    [[LPAD_VAL10:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
 209 // CHECK1-NEXT:    resume { ptr, i32 } [[LPAD_VAL10]]
 210 //
 211 //
 212 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SC1El
 213 // CHECK1-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 noundef [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
 214 // CHECK1-NEXT:  entry:
 215 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 216 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 217 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 218 // CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 219 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 220 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
 221 // CHECK1-NEXT:    call void @_ZN1SC2El(ptr noundef nonnull align 8 dereferenceable(24) [[THIS1]], i64 noundef [[TMP0]])
 222 // CHECK1-NEXT:    ret void
 223 //
 224 //
 225 // CHECK1-LABEL: define {{[^@]+}}@_ZN1ScvcEv
 226 // CHECK1-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[ATTR2:[0-9]+]] comdat {
 227 // CHECK1-NEXT:  entry:
 228 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 229 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 230 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 231 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 232 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
 233 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
 234 // CHECK1-NEXT:    ret i8 [[CONV]]
 235 //
 236 //
 237 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50
 238 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
 239 // CHECK1-NEXT:  entry:
 240 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined)
 241 // CHECK1-NEXT:    ret void
 242 //
 243 //
 244 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined
 245 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
 246 // CHECK1-NEXT:  entry:
 247 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 248 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 249 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 250 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 251 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 252 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 253 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 254 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 255 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 256 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 257 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 258 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 259 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
 260 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 261 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 262 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 263 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 264 // 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)
 265 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 266 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
 267 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 268 // CHECK1:       cond.true:
 269 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 270 // CHECK1:       cond.false:
 271 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 272 // CHECK1-NEXT:    br label [[COND_END]]
 273 // CHECK1:       cond.end:
 274 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 275 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 276 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 277 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 278 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 279 // CHECK1:       omp.inner.for.cond:
 280 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 281 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 282 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 283 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 284 // CHECK1:       omp.inner.for.body:
 285 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 2)
 286 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 287 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
 288 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 289 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 290 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
 291 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 292 // CHECK1:       omp.inner.for.inc:
 293 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 294 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 295 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
 296 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 297 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 298 // CHECK1:       omp.inner.for.end:
 299 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 300 // CHECK1:       omp.loop.exit:
 301 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 302 // CHECK1-NEXT:    ret void
 303 //
 304 //
 305 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined.omp_outlined
 306 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
 307 // CHECK1-NEXT:  entry:
 308 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 309 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 310 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 311 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 312 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 313 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 314 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 315 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 316 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 317 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 318 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 319 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 320 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 321 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 322 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 323 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 324 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
 325 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 326 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 327 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 328 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
 329 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 330 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 331 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 332 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 333 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 334 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 335 // 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)
 336 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 337 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
 338 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 339 // CHECK1:       cond.true:
 340 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 341 // CHECK1:       cond.false:
 342 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 343 // CHECK1-NEXT:    br label [[COND_END]]
 344 // CHECK1:       cond.end:
 345 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
 346 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 347 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 348 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
 349 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 350 // CHECK1:       omp.inner.for.cond:
 351 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 352 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 353 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
 354 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 355 // CHECK1:       omp.inner.for.body:
 356 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 357 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
 358 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 359 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 360 // CHECK1-NEXT:    invoke void @_Z3foov()
 361 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
 362 // CHECK1:       invoke.cont:
 363 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 364 // CHECK1:       omp.body.continue:
 365 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 366 // CHECK1:       omp.inner.for.inc:
 367 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 368 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
 369 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
 370 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 371 // CHECK1:       omp.inner.for.end:
 372 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 373 // CHECK1:       omp.loop.exit:
 374 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
 375 // CHECK1-NEXT:    ret void
 376 // CHECK1:       terminate.lpad:
 377 // CHECK1-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
 378 // CHECK1-NEXT:            catch ptr null
 379 // CHECK1-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
 380 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7:[0-9]+]]
 381 // CHECK1-NEXT:    unreachable
 382 //
 383 //
 384 // CHECK1-LABEL: define {{[^@]+}}@__clang_call_terminate
 385 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR6:[0-9]+]] comdat {
 386 // CHECK1-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR4]]
 387 // CHECK1-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
 388 // CHECK1-NEXT:    unreachable
 389 //
 390 //
 391 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55
 392 // CHECK1-SAME: (i64 noundef [[A:%.*]]) #[[ATTR3]] {
 393 // CHECK1-NEXT:  entry:
 394 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 395 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
 396 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
 397 // CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 398 // CHECK1-NEXT:    [[TMP0:%.*]] = load i8, ptr [[A_ADDR]], align 1
 399 // CHECK1-NEXT:    store i8 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 1
 400 // CHECK1-NEXT:    [[TMP1:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
 401 // CHECK1-NEXT:    store i8 [[TMP1]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
 402 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
 403 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined, i64 [[TMP2]])
 404 // CHECK1-NEXT:    ret void
 405 //
 406 //
 407 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined
 408 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
 409 // CHECK1-NEXT:  entry:
 410 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 411 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 412 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
 413 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 414 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 415 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 416 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 417 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 418 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 419 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 420 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 421 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 422 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
 423 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 424 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
 425 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 426 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 427 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 428 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 429 // 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)
 430 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 431 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
 432 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 433 // CHECK1:       cond.true:
 434 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 435 // CHECK1:       cond.false:
 436 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 437 // CHECK1-NEXT:    br label [[COND_END]]
 438 // CHECK1:       cond.end:
 439 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 440 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 441 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 442 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 443 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 444 // CHECK1:       omp.inner.for.cond:
 445 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 446 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 447 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 448 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 449 // CHECK1:       omp.inner.for.body:
 450 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
 451 // CHECK1-NEXT:    [[TMP8:%.*]] = sext i8 [[TMP7]] to i32
 452 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP8]])
 453 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 454 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 455 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 456 // CHECK1-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
 457 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined, i64 [[TMP10]], i64 [[TMP12]])
 458 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 459 // CHECK1:       omp.inner.for.inc:
 460 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 461 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 462 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
 463 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 464 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 465 // CHECK1:       omp.inner.for.end:
 466 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 467 // CHECK1:       omp.loop.exit:
 468 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 469 // CHECK1-NEXT:    ret void
 470 //
 471 //
 472 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined
 473 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
 474 // CHECK1-NEXT:  entry:
 475 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 476 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 477 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 478 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 479 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 480 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 481 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 482 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 483 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 484 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 485 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 486 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 487 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 488 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 489 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 490 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 491 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
 492 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 493 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 494 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 495 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
 496 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 497 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 498 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 499 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 500 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 501 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 502 // 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)
 503 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 504 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
 505 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 506 // CHECK1:       cond.true:
 507 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 508 // CHECK1:       cond.false:
 509 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 510 // CHECK1-NEXT:    br label [[COND_END]]
 511 // CHECK1:       cond.end:
 512 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
 513 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 514 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 515 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
 516 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 517 // CHECK1:       omp.inner.for.cond:
 518 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 519 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 520 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
 521 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 522 // CHECK1:       omp.inner.for.body:
 523 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 524 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
 525 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 526 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 527 // CHECK1-NEXT:    invoke void @_Z3foov()
 528 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
 529 // CHECK1:       invoke.cont:
 530 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 531 // CHECK1:       omp.body.continue:
 532 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 533 // CHECK1:       omp.inner.for.inc:
 534 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 535 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
 536 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
 537 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 538 // CHECK1:       omp.inner.for.end:
 539 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 540 // CHECK1:       omp.loop.exit:
 541 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
 542 // CHECK1-NEXT:    ret void
 543 // CHECK1:       terminate.lpad:
 544 // CHECK1-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
 545 // CHECK1-NEXT:            catch ptr null
 546 // CHECK1-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
 547 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
 548 // CHECK1-NEXT:    unreachable
 549 //
 550 //
 551 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
 552 // CHECK1-SAME: () #[[ATTR2]] comdat {
 553 // CHECK1-NEXT:  entry:
 554 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 555 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 556 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 557 // CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 558 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 559 // CHECK1-NEXT:    store i32 3, ptr [[TMP0]], align 4
 560 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 561 // CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
 562 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 563 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 564 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 565 // CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
 566 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 567 // CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
 568 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 569 // CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
 570 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 571 // CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
 572 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 573 // CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
 574 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 575 // CHECK1-NEXT:    store i64 100, ptr [[TMP8]], align 8
 576 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 577 // CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
 578 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 579 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
 580 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 581 // CHECK1-NEXT:    store [3 x i32] [i32 5, i32 0, i32 0], ptr [[TMP11]], align 4
 582 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 583 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
 584 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 5, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.region_id, ptr [[KERNEL_ARGS]])
 585 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 586 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 587 // CHECK1:       omp_offload.failed:
 588 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36() #[[ATTR4]]
 589 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 590 // CHECK1:       omp_offload.cont:
 591 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
 592 // CHECK1-NEXT:    store i32 3, ptr [[TMP15]], align 4
 593 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
 594 // CHECK1-NEXT:    store i32 0, ptr [[TMP16]], align 4
 595 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
 596 // CHECK1-NEXT:    store ptr null, ptr [[TMP17]], align 8
 597 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
 598 // CHECK1-NEXT:    store ptr null, ptr [[TMP18]], align 8
 599 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
 600 // CHECK1-NEXT:    store ptr null, ptr [[TMP19]], align 8
 601 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
 602 // CHECK1-NEXT:    store ptr null, ptr [[TMP20]], align 8
 603 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
 604 // CHECK1-NEXT:    store ptr null, ptr [[TMP21]], align 8
 605 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
 606 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 607 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
 608 // CHECK1-NEXT:    store i64 100, ptr [[TMP23]], align 8
 609 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
 610 // CHECK1-NEXT:    store i64 0, ptr [[TMP24]], align 8
 611 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
 612 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
 613 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
 614 // CHECK1-NEXT:    store [3 x i32] [i32 23, i32 0, i32 0], ptr [[TMP26]], align 4
 615 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
 616 // CHECK1-NEXT:    store i32 0, ptr [[TMP27]], align 4
 617 // CHECK1-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 23, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.region_id, ptr [[KERNEL_ARGS2]])
 618 // CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 619 // CHECK1-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
 620 // CHECK1:       omp_offload.failed3:
 621 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40() #[[ATTR4]]
 622 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
 623 // CHECK1:       omp_offload.cont4:
 624 // CHECK1-NEXT:    ret i32 0
 625 //
 626 //
 627 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
 628 // CHECK1-SAME: () #[[ATTR2]] comdat personality ptr @__gxx_personality_v0 {
 629 // CHECK1-NEXT:  entry:
 630 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 631 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 632 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
 633 // CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 8
 634 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 635 // CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 636 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 637 // CHECK1-NEXT:    store i32 3, ptr [[TMP0]], align 4
 638 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 639 // CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
 640 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 641 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 642 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 643 // CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
 644 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 645 // CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
 646 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 647 // CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
 648 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 649 // CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
 650 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 651 // CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
 652 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 653 // CHECK1-NEXT:    store i64 100, ptr [[TMP8]], align 8
 654 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 655 // CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
 656 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 657 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
 658 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 659 // CHECK1-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP11]], align 4
 660 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 661 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
 662 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.region_id, ptr [[KERNEL_ARGS]])
 663 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 664 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 665 // CHECK1:       omp_offload.failed:
 666 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36() #[[ATTR4]]
 667 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 668 // CHECK1:       omp_offload.cont:
 669 // CHECK1-NEXT:    invoke void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 noundef 23)
 670 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
 671 // CHECK1:       invoke.cont:
 672 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
 673 // CHECK1-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR4]]
 674 // CHECK1-NEXT:    store i8 [[CALL]], ptr [[DOTCAPTURE_EXPR_]], align 1
 675 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
 676 // CHECK1-NEXT:    [[TMP16:%.*]] = zext i8 [[TMP15]] to i32
 677 // CHECK1-NEXT:    [[TMP17:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP16]], 0
 678 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
 679 // CHECK1-NEXT:    store i32 3, ptr [[TMP18]], align 4
 680 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
 681 // CHECK1-NEXT:    store i32 0, ptr [[TMP19]], align 4
 682 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
 683 // CHECK1-NEXT:    store ptr null, ptr [[TMP20]], align 8
 684 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
 685 // CHECK1-NEXT:    store ptr null, ptr [[TMP21]], align 8
 686 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
 687 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 688 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
 689 // CHECK1-NEXT:    store ptr null, ptr [[TMP23]], align 8
 690 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
 691 // CHECK1-NEXT:    store ptr null, ptr [[TMP24]], align 8
 692 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
 693 // CHECK1-NEXT:    store ptr null, ptr [[TMP25]], align 8
 694 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
 695 // CHECK1-NEXT:    store i64 100, ptr [[TMP26]], align 8
 696 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
 697 // CHECK1-NEXT:    store i64 0, ptr [[TMP27]], align 8
 698 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
 699 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP28]], align 4
 700 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
 701 // CHECK1-NEXT:    store [3 x i32] [[TMP17]], ptr [[TMP29]], align 4
 702 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
 703 // CHECK1-NEXT:    store i32 0, ptr [[TMP30]], align 4
 704 // CHECK1-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.region_id, ptr [[KERNEL_ARGS2]])
 705 // CHECK1-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
 706 // CHECK1-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
 707 // CHECK1:       omp_offload.failed3:
 708 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40() #[[ATTR4]]
 709 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
 710 // CHECK1:       omp_offload.cont4:
 711 // CHECK1-NEXT:    ret i32 0
 712 // CHECK1:       terminate.lpad:
 713 // CHECK1-NEXT:    [[TMP33:%.*]] = landingpad { ptr, i32 }
 714 // CHECK1-NEXT:            catch ptr null
 715 // CHECK1-NEXT:    [[TMP34:%.*]] = extractvalue { ptr, i32 } [[TMP33]], 0
 716 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP34]]) #[[ATTR7]]
 717 // CHECK1-NEXT:    unreachable
 718 //
 719 //
 720 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SD1Ev
 721 // CHECK1-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR2]] comdat {
 722 // CHECK1-NEXT:  entry:
 723 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 724 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 725 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 726 // CHECK1-NEXT:    call void @_ZN1SD2Ev(ptr noundef nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR4]]
 727 // CHECK1-NEXT:    ret void
 728 //
 729 //
 730 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SC2El
 731 // CHECK1-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 noundef [[A:%.*]]) unnamed_addr #[[ATTR2]] comdat {
 732 // CHECK1-NEXT:  entry:
 733 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 734 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 735 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 736 // CHECK1-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 737 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 738 // CHECK1-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
 739 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
 740 // CHECK1-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
 741 // CHECK1-NEXT:    ret void
 742 //
 743 //
 744 // CHECK1-LABEL: define {{[^@]+}}@_ZN1SD2Ev
 745 // CHECK1-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR2]] comdat {
 746 // CHECK1-NEXT:  entry:
 747 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 748 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 749 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 750 // CHECK1-NEXT:    ret void
 751 //
 752 //
 753 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36
 754 // CHECK1-SAME: () #[[ATTR3]] {
 755 // CHECK1-NEXT:  entry:
 756 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined)
 757 // CHECK1-NEXT:    ret void
 758 //
 759 //
 760 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined
 761 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
 762 // CHECK1-NEXT:  entry:
 763 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 764 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 765 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 766 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 767 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 768 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 769 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 770 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 771 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 772 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 773 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 774 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 775 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
 776 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 777 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 778 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 779 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 780 // 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)
 781 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 782 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
 783 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 784 // CHECK1:       cond.true:
 785 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 786 // CHECK1:       cond.false:
 787 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 788 // CHECK1-NEXT:    br label [[COND_END]]
 789 // CHECK1:       cond.end:
 790 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 791 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 792 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 793 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 794 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 795 // CHECK1:       omp.inner.for.cond:
 796 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 797 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 798 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 799 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 800 // CHECK1:       omp.inner.for.body:
 801 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 5)
 802 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 803 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
 804 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 805 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 806 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
 807 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 808 // CHECK1:       omp.inner.for.inc:
 809 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 810 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 811 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
 812 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 813 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 814 // CHECK1:       omp.inner.for.end:
 815 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 816 // CHECK1:       omp.loop.exit:
 817 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 818 // CHECK1-NEXT:    ret void
 819 //
 820 //
 821 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined.omp_outlined
 822 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
 823 // CHECK1-NEXT:  entry:
 824 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 825 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 826 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 827 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 828 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 829 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 830 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 831 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 832 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 833 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 834 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 835 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 836 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 837 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 838 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 839 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 840 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
 841 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 842 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 843 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 844 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
 845 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 846 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 847 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 848 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 849 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 850 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 851 // 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)
 852 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 853 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
 854 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 855 // CHECK1:       cond.true:
 856 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 857 // CHECK1:       cond.false:
 858 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 859 // CHECK1-NEXT:    br label [[COND_END]]
 860 // CHECK1:       cond.end:
 861 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
 862 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 863 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 864 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
 865 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 866 // CHECK1:       omp.inner.for.cond:
 867 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 868 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 869 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
 870 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 871 // CHECK1:       omp.inner.for.body:
 872 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 873 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
 874 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 875 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 876 // CHECK1-NEXT:    invoke void @_Z3foov()
 877 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
 878 // CHECK1:       invoke.cont:
 879 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 880 // CHECK1:       omp.body.continue:
 881 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 882 // CHECK1:       omp.inner.for.inc:
 883 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 884 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
 885 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
 886 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 887 // CHECK1:       omp.inner.for.end:
 888 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 889 // CHECK1:       omp.loop.exit:
 890 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
 891 // CHECK1-NEXT:    ret void
 892 // CHECK1:       terminate.lpad:
 893 // CHECK1-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
 894 // CHECK1-NEXT:            catch ptr null
 895 // CHECK1-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
 896 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
 897 // CHECK1-NEXT:    unreachable
 898 //
 899 //
 900 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40
 901 // CHECK1-SAME: () #[[ATTR3]] {
 902 // CHECK1-NEXT:  entry:
 903 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined)
 904 // CHECK1-NEXT:    ret void
 905 //
 906 //
 907 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined
 908 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
 909 // CHECK1-NEXT:  entry:
 910 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 911 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 912 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 913 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 914 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 915 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 916 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 917 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 918 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 919 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 920 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 921 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 922 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
 923 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 924 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 925 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 926 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 927 // 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)
 928 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 929 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
 930 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 931 // CHECK1:       cond.true:
 932 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 933 // CHECK1:       cond.false:
 934 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 935 // CHECK1-NEXT:    br label [[COND_END]]
 936 // CHECK1:       cond.end:
 937 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 938 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 939 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 940 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 941 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 942 // CHECK1:       omp.inner.for.cond:
 943 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 944 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 945 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 946 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 947 // CHECK1:       omp.inner.for.body:
 948 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 23)
 949 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 950 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
 951 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 952 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 953 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
 954 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 955 // CHECK1:       omp.inner.for.inc:
 956 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 957 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 958 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
 959 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 960 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 961 // CHECK1:       omp.inner.for.end:
 962 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 963 // CHECK1:       omp.loop.exit:
 964 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 965 // CHECK1-NEXT:    ret void
 966 //
 967 //
 968 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined.omp_outlined
 969 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
 970 // CHECK1-NEXT:  entry:
 971 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 972 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 973 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 974 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 975 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 976 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 977 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 978 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 979 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 980 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 981 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 982 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 983 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 984 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 985 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 986 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 987 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
 988 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 989 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 990 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 991 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
 992 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 993 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 994 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 995 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 996 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 997 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 998 // 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)
 999 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1000 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1001 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1002 // CHECK1:       cond.true:
1003 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1004 // CHECK1:       cond.false:
1005 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1006 // CHECK1-NEXT:    br label [[COND_END]]
1007 // CHECK1:       cond.end:
1008 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1009 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1010 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1011 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
1012 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1013 // CHECK1:       omp.inner.for.cond:
1014 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1015 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1016 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1017 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1018 // CHECK1:       omp.inner.for.body:
1019 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1020 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1021 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1022 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1023 // CHECK1-NEXT:    invoke void @_Z3foov()
1024 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1025 // CHECK1:       invoke.cont:
1026 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1027 // CHECK1:       omp.body.continue:
1028 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1029 // CHECK1:       omp.inner.for.inc:
1030 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1031 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1032 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1033 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1034 // CHECK1:       omp.inner.for.end:
1035 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1036 // CHECK1:       omp.loop.exit:
1037 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
1038 // CHECK1-NEXT:    ret void
1039 // CHECK1:       terminate.lpad:
1040 // CHECK1-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
1041 // CHECK1-NEXT:            catch ptr null
1042 // CHECK1-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
1043 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
1044 // CHECK1-NEXT:    unreachable
1045 //
1046 //
1047 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36
1048 // CHECK1-SAME: () #[[ATTR3]] {
1049 // CHECK1-NEXT:  entry:
1050 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined)
1051 // CHECK1-NEXT:    ret void
1052 //
1053 //
1054 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined
1055 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
1056 // CHECK1-NEXT:  entry:
1057 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1058 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1059 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1060 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1061 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1062 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1063 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1064 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1065 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1066 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1067 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1068 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1069 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
1070 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1071 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1072 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1073 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
1074 // 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)
1075 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1076 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1077 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1078 // CHECK1:       cond.true:
1079 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1080 // CHECK1:       cond.false:
1081 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1082 // CHECK1-NEXT:    br label [[COND_END]]
1083 // CHECK1:       cond.end:
1084 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1085 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1086 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1087 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
1088 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1089 // CHECK1:       omp.inner.for.cond:
1090 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1091 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1092 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1093 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1094 // CHECK1:       omp.inner.for.body:
1095 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 1)
1096 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1097 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1098 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1099 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1100 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
1101 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1102 // CHECK1:       omp.inner.for.inc:
1103 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1104 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1105 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1106 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1107 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1108 // CHECK1:       omp.inner.for.end:
1109 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1110 // CHECK1:       omp.loop.exit:
1111 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
1112 // CHECK1-NEXT:    ret void
1113 //
1114 //
1115 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined.omp_outlined
1116 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
1117 // CHECK1-NEXT:  entry:
1118 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1119 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1120 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1121 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1122 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1123 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1124 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1125 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1126 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1127 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1128 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1129 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1130 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1131 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1132 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1133 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1134 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
1135 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1136 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1137 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1138 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1139 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1140 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
1141 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1142 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1143 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1144 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
1145 // 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)
1146 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1147 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1148 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1149 // CHECK1:       cond.true:
1150 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1151 // CHECK1:       cond.false:
1152 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1153 // CHECK1-NEXT:    br label [[COND_END]]
1154 // CHECK1:       cond.end:
1155 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1156 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1157 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1158 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
1159 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1160 // CHECK1:       omp.inner.for.cond:
1161 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1162 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1163 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1164 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1165 // CHECK1:       omp.inner.for.body:
1166 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1167 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1168 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1169 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1170 // CHECK1-NEXT:    invoke void @_Z3foov()
1171 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1172 // CHECK1:       invoke.cont:
1173 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1174 // CHECK1:       omp.body.continue:
1175 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1176 // CHECK1:       omp.inner.for.inc:
1177 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1178 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1179 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1180 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1181 // CHECK1:       omp.inner.for.end:
1182 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1183 // CHECK1:       omp.loop.exit:
1184 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
1185 // CHECK1-NEXT:    ret void
1186 // CHECK1:       terminate.lpad:
1187 // CHECK1-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
1188 // CHECK1-NEXT:            catch ptr null
1189 // CHECK1-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
1190 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
1191 // CHECK1-NEXT:    unreachable
1192 //
1193 //
1194 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40
1195 // CHECK1-SAME: () #[[ATTR3]] personality ptr @__gxx_personality_v0 {
1196 // CHECK1-NEXT:  entry:
1197 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
1198 // CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 8
1199 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1200 // CHECK1-NEXT:    invoke void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 noundef 23)
1201 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1202 // CHECK1:       invoke.cont:
1203 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
1204 // CHECK1-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR4]]
1205 // CHECK1-NEXT:    store i8 [[CALL]], ptr [[DOTCAPTURE_EXPR_]], align 1
1206 // CHECK1-NEXT:    [[TMP0:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
1207 // CHECK1-NEXT:    store i8 [[TMP0]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
1208 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
1209 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined, i64 [[TMP1]])
1210 // CHECK1-NEXT:    ret void
1211 // CHECK1:       terminate.lpad:
1212 // CHECK1-NEXT:    [[TMP2:%.*]] = landingpad { ptr, i32 }
1213 // CHECK1-NEXT:            catch ptr null
1214 // CHECK1-NEXT:    [[TMP3:%.*]] = extractvalue { ptr, i32 } [[TMP2]], 0
1215 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP3]]) #[[ATTR7]]
1216 // CHECK1-NEXT:    unreachable
1217 //
1218 //
1219 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined
1220 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
1221 // CHECK1-NEXT:  entry:
1222 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1223 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1224 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1225 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1226 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1227 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1228 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1229 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1230 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1231 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1232 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1233 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1234 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
1235 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1236 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
1237 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1238 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1239 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1240 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
1241 // 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)
1242 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1243 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1244 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1245 // CHECK1:       cond.true:
1246 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1247 // CHECK1:       cond.false:
1248 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1249 // CHECK1-NEXT:    br label [[COND_END]]
1250 // CHECK1:       cond.end:
1251 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1252 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1253 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1254 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
1255 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1256 // CHECK1:       omp.inner.for.cond:
1257 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1258 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1259 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1260 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1261 // CHECK1:       omp.inner.for.body:
1262 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
1263 // CHECK1-NEXT:    [[TMP8:%.*]] = sext i8 [[TMP7]] to i32
1264 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP8]])
1265 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1266 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1267 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1268 // CHECK1-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
1269 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined.omp_outlined, i64 [[TMP10]], i64 [[TMP12]])
1270 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1271 // CHECK1:       omp.inner.for.inc:
1272 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1273 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1274 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1275 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1276 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1277 // CHECK1:       omp.inner.for.end:
1278 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1279 // CHECK1:       omp.loop.exit:
1280 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
1281 // CHECK1-NEXT:    ret void
1282 //
1283 //
1284 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined.omp_outlined
1285 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
1286 // CHECK1-NEXT:  entry:
1287 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1288 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1289 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1290 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1291 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1292 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1293 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1294 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1295 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1296 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1297 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1298 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1299 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1300 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1301 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1302 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1303 // CHECK1-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
1304 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1305 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1306 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1307 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1308 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1309 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
1310 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1311 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1312 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1313 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
1314 // 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)
1315 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1316 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1317 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1318 // CHECK1:       cond.true:
1319 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1320 // CHECK1:       cond.false:
1321 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1322 // CHECK1-NEXT:    br label [[COND_END]]
1323 // CHECK1:       cond.end:
1324 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1325 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1326 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1327 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
1328 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1329 // CHECK1:       omp.inner.for.cond:
1330 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1331 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1332 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1333 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1334 // CHECK1:       omp.inner.for.body:
1335 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1336 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1337 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1338 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1339 // CHECK1-NEXT:    invoke void @_Z3foov()
1340 // CHECK1-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1341 // CHECK1:       invoke.cont:
1342 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1343 // CHECK1:       omp.body.continue:
1344 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1345 // CHECK1:       omp.inner.for.inc:
1346 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1347 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1348 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1349 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1350 // CHECK1:       omp.inner.for.end:
1351 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1352 // CHECK1:       omp.loop.exit:
1353 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
1354 // CHECK1-NEXT:    ret void
1355 // CHECK1:       terminate.lpad:
1356 // CHECK1-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
1357 // CHECK1-NEXT:            catch ptr null
1358 // CHECK1-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
1359 // CHECK1-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
1360 // CHECK1-NEXT:    unreachable
1361 //
1362 //
1363 // CHECK5-LABEL: define {{[^@]+}}@main
1364 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] personality ptr @__gxx_personality_v0 {
1365 // CHECK5-NEXT:  entry:
1366 // CHECK5-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1367 // CHECK5-NEXT:    [[S:%.*]] = alloca [[STRUCT_S:%.*]], align 8
1368 // CHECK5-NEXT:    [[A:%.*]] = alloca i8, align 1
1369 // CHECK5-NEXT:    [[EXN_SLOT:%.*]] = alloca ptr, align 8
1370 // CHECK5-NEXT:    [[EHSELECTOR_SLOT:%.*]] = alloca i32, align 4
1371 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1372 // CHECK5-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1373 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1374 // CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
1375 // CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
1376 // CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
1377 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
1378 // CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1379 // CHECK5-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1380 // CHECK5-NEXT:    store i32 0, ptr [[RETVAL]], align 4
1381 // CHECK5-NEXT:    call void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[S]], i64 noundef 0)
1382 // CHECK5-NEXT:    [[CALL:%.*]] = invoke noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[S]])
1383 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[LPAD:%.*]]
1384 // CHECK5:       invoke.cont:
1385 // CHECK5-NEXT:    store i8 [[CALL]], ptr [[A]], align 1
1386 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1387 // CHECK5-NEXT:    store i32 3, ptr [[TMP0]], align 4
1388 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1389 // CHECK5-NEXT:    store i32 0, ptr [[TMP1]], align 4
1390 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1391 // CHECK5-NEXT:    store ptr null, ptr [[TMP2]], align 8
1392 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1393 // CHECK5-NEXT:    store ptr null, ptr [[TMP3]], align 8
1394 // CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1395 // CHECK5-NEXT:    store ptr null, ptr [[TMP4]], align 8
1396 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1397 // CHECK5-NEXT:    store ptr null, ptr [[TMP5]], align 8
1398 // CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1399 // CHECK5-NEXT:    store ptr null, ptr [[TMP6]], align 8
1400 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1401 // CHECK5-NEXT:    store ptr null, ptr [[TMP7]], align 8
1402 // CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1403 // CHECK5-NEXT:    store i64 100, ptr [[TMP8]], align 8
1404 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1405 // CHECK5-NEXT:    store i64 0, ptr [[TMP9]], align 8
1406 // CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1407 // CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
1408 // CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1409 // CHECK5-NEXT:    store [3 x i32] [i32 2, i32 0, i32 0], ptr [[TMP11]], align 4
1410 // CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1411 // CHECK5-NEXT:    store i32 0, ptr [[TMP12]], align 4
1412 // CHECK5-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 2, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.region_id, ptr [[KERNEL_ARGS]])
1413 // CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1414 // CHECK5-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1415 // CHECK5:       omp_offload.failed:
1416 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50() #[[ATTR4:[0-9]+]]
1417 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1418 // CHECK5:       lpad:
1419 // CHECK5-NEXT:    [[TMP15:%.*]] = landingpad { ptr, i32 }
1420 // CHECK5-NEXT:            cleanup
1421 // CHECK5-NEXT:    [[TMP16:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 0
1422 // CHECK5-NEXT:    store ptr [[TMP16]], ptr [[EXN_SLOT]], align 8
1423 // CHECK5-NEXT:    [[TMP17:%.*]] = extractvalue { ptr, i32 } [[TMP15]], 1
1424 // CHECK5-NEXT:    store i32 [[TMP17]], ptr [[EHSELECTOR_SLOT]], align 4
1425 // CHECK5-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR4]]
1426 // CHECK5-NEXT:    br label [[EH_RESUME:%.*]]
1427 // CHECK5:       omp_offload.cont:
1428 // CHECK5-NEXT:    [[TMP18:%.*]] = load i8, ptr [[A]], align 1
1429 // CHECK5-NEXT:    store i8 [[TMP18]], ptr [[A_CASTED]], align 1
1430 // CHECK5-NEXT:    [[TMP19:%.*]] = load i64, ptr [[A_CASTED]], align 8
1431 // CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1432 // CHECK5-NEXT:    store i64 [[TMP19]], ptr [[TMP20]], align 8
1433 // CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1434 // CHECK5-NEXT:    store i64 [[TMP19]], ptr [[TMP21]], align 8
1435 // CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1436 // CHECK5-NEXT:    store ptr null, ptr [[TMP22]], align 8
1437 // CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1438 // CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1439 // CHECK5-NEXT:    [[TMP25:%.*]] = load i8, ptr [[A]], align 1
1440 // CHECK5-NEXT:    store i8 [[TMP25]], ptr [[DOTCAPTURE_EXPR_]], align 1
1441 // CHECK5-NEXT:    [[TMP26:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
1442 // CHECK5-NEXT:    [[TMP27:%.*]] = zext i8 [[TMP26]] to i32
1443 // CHECK5-NEXT:    [[TMP28:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP27]], 0
1444 // CHECK5-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
1445 // CHECK5-NEXT:    store i32 3, ptr [[TMP29]], align 4
1446 // CHECK5-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
1447 // CHECK5-NEXT:    store i32 1, ptr [[TMP30]], align 4
1448 // CHECK5-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
1449 // CHECK5-NEXT:    store ptr [[TMP23]], ptr [[TMP31]], align 8
1450 // CHECK5-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
1451 // CHECK5-NEXT:    store ptr [[TMP24]], ptr [[TMP32]], align 8
1452 // CHECK5-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
1453 // CHECK5-NEXT:    store ptr @.offload_sizes, ptr [[TMP33]], align 8
1454 // CHECK5-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
1455 // CHECK5-NEXT:    store ptr @.offload_maptypes, ptr [[TMP34]], align 8
1456 // CHECK5-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
1457 // CHECK5-NEXT:    store ptr null, ptr [[TMP35]], align 8
1458 // CHECK5-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
1459 // CHECK5-NEXT:    store ptr null, ptr [[TMP36]], align 8
1460 // CHECK5-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
1461 // CHECK5-NEXT:    store i64 100, ptr [[TMP37]], align 8
1462 // CHECK5-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
1463 // CHECK5-NEXT:    store i64 0, ptr [[TMP38]], align 8
1464 // CHECK5-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
1465 // CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP39]], align 4
1466 // CHECK5-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
1467 // CHECK5-NEXT:    store [3 x i32] [[TMP28]], ptr [[TMP40]], align 4
1468 // CHECK5-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
1469 // CHECK5-NEXT:    store i32 0, ptr [[TMP41]], align 4
1470 // CHECK5-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 [[TMP27]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.region_id, ptr [[KERNEL_ARGS2]])
1471 // CHECK5-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
1472 // CHECK5-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
1473 // CHECK5:       omp_offload.failed3:
1474 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55(i64 [[TMP19]]) #[[ATTR4]]
1475 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
1476 // CHECK5:       omp_offload.cont4:
1477 // CHECK5-NEXT:    [[TMP44:%.*]] = load i8, ptr [[A]], align 1
1478 // CHECK5-NEXT:    [[CONV:%.*]] = sext i8 [[TMP44]] to i32
1479 // CHECK5-NEXT:    [[CALL6:%.*]] = invoke noundef signext i32 @_Z5tmainIcLi5EEiv()
1480 // CHECK5-NEXT:            to label [[INVOKE_CONT5:%.*]] unwind label [[LPAD]]
1481 // CHECK5:       invoke.cont5:
1482 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CALL6]]
1483 // CHECK5-NEXT:    [[CALL8:%.*]] = invoke noundef signext i32 @_Z5tmainI1SLi1EEiv()
1484 // CHECK5-NEXT:            to label [[INVOKE_CONT7:%.*]] unwind label [[LPAD]]
1485 // CHECK5:       invoke.cont7:
1486 // CHECK5-NEXT:    [[ADD9:%.*]] = add nsw i32 [[ADD]], [[CALL8]]
1487 // CHECK5-NEXT:    store i32 [[ADD9]], ptr [[RETVAL]], align 4
1488 // CHECK5-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[S]]) #[[ATTR4]]
1489 // CHECK5-NEXT:    [[TMP45:%.*]] = load i32, ptr [[RETVAL]], align 4
1490 // CHECK5-NEXT:    ret i32 [[TMP45]]
1491 // CHECK5:       eh.resume:
1492 // CHECK5-NEXT:    [[EXN:%.*]] = load ptr, ptr [[EXN_SLOT]], align 8
1493 // CHECK5-NEXT:    [[SEL:%.*]] = load i32, ptr [[EHSELECTOR_SLOT]], align 4
1494 // CHECK5-NEXT:    [[LPAD_VAL:%.*]] = insertvalue { ptr, i32 } poison, ptr [[EXN]], 0
1495 // CHECK5-NEXT:    [[LPAD_VAL10:%.*]] = insertvalue { ptr, i32 } [[LPAD_VAL]], i32 [[SEL]], 1
1496 // CHECK5-NEXT:    resume { ptr, i32 } [[LPAD_VAL10]]
1497 //
1498 //
1499 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SC1El
1500 // CHECK5-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 noundef [[A:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
1501 // CHECK5-NEXT:  entry:
1502 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1503 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1504 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1505 // CHECK5-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
1506 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1507 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
1508 // CHECK5-NEXT:    call void @_ZN1SC2El(ptr noundef nonnull align 8 dereferenceable(24) [[THIS1]], i64 noundef [[TMP0]])
1509 // CHECK5-NEXT:    ret void
1510 //
1511 //
1512 // CHECK5-LABEL: define {{[^@]+}}@_ZN1ScvcEv
1513 // CHECK5-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]]) #[[ATTR2:[0-9]+]] comdat {
1514 // CHECK5-NEXT:  entry:
1515 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1516 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1517 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1518 // CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
1519 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A]], align 8
1520 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i8
1521 // CHECK5-NEXT:    ret i8 [[CONV]]
1522 //
1523 //
1524 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50
1525 // CHECK5-SAME: () #[[ATTR3:[0-9]+]] {
1526 // CHECK5-NEXT:  entry:
1527 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined)
1528 // CHECK5-NEXT:    ret void
1529 //
1530 //
1531 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined
1532 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
1533 // CHECK5-NEXT:  entry:
1534 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1535 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1536 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1537 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1538 // CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1539 // CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1540 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1541 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1542 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
1543 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1544 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1545 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1546 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
1547 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1548 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1549 // CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1550 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
1551 // 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)
1552 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1553 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1554 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1555 // CHECK5:       cond.true:
1556 // CHECK5-NEXT:    br label [[COND_END:%.*]]
1557 // CHECK5:       cond.false:
1558 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1559 // CHECK5-NEXT:    br label [[COND_END]]
1560 // CHECK5:       cond.end:
1561 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1562 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1563 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1564 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
1565 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1566 // CHECK5:       omp.inner.for.cond:
1567 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1568 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1569 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1570 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1571 // CHECK5:       omp.inner.for.body:
1572 // CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 2)
1573 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1574 // CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1575 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1576 // CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1577 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
1578 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1579 // CHECK5:       omp.inner.for.inc:
1580 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1581 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1582 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1583 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1584 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
1585 // CHECK5:       omp.inner.for.end:
1586 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1587 // CHECK5:       omp.loop.exit:
1588 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
1589 // CHECK5-NEXT:    ret void
1590 //
1591 //
1592 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50.omp_outlined.omp_outlined
1593 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
1594 // CHECK5-NEXT:  entry:
1595 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1596 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1597 // CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1598 // CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1599 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1600 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1601 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1602 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1603 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1604 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1605 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
1606 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1607 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1608 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1609 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1610 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1611 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
1612 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1613 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1614 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1615 // CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1616 // CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1617 // CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
1618 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1619 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1620 // CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1621 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
1622 // 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)
1623 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1624 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1625 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1626 // CHECK5:       cond.true:
1627 // CHECK5-NEXT:    br label [[COND_END:%.*]]
1628 // CHECK5:       cond.false:
1629 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1630 // CHECK5-NEXT:    br label [[COND_END]]
1631 // CHECK5:       cond.end:
1632 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1633 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1634 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1635 // CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
1636 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1637 // CHECK5:       omp.inner.for.cond:
1638 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1639 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1640 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1641 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1642 // CHECK5:       omp.inner.for.body:
1643 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1644 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1645 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1646 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1647 // CHECK5-NEXT:    invoke void @_Z3foov()
1648 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1649 // CHECK5:       invoke.cont:
1650 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1651 // CHECK5:       omp.body.continue:
1652 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1653 // CHECK5:       omp.inner.for.inc:
1654 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1655 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1656 // CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1657 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
1658 // CHECK5:       omp.inner.for.end:
1659 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1660 // CHECK5:       omp.loop.exit:
1661 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
1662 // CHECK5-NEXT:    ret void
1663 // CHECK5:       terminate.lpad:
1664 // CHECK5-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
1665 // CHECK5-NEXT:            catch ptr null
1666 // CHECK5-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
1667 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7:[0-9]+]]
1668 // CHECK5-NEXT:    unreachable
1669 //
1670 //
1671 // CHECK5-LABEL: define {{[^@]+}}@__clang_call_terminate
1672 // CHECK5-SAME: (ptr noundef [[TMP0:%.*]]) #[[ATTR6:[0-9]+]] comdat {
1673 // CHECK5-NEXT:    [[TMP2:%.*]] = call ptr @__cxa_begin_catch(ptr [[TMP0]]) #[[ATTR4]]
1674 // CHECK5-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
1675 // CHECK5-NEXT:    unreachable
1676 //
1677 //
1678 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55
1679 // CHECK5-SAME: (i64 noundef [[A:%.*]]) #[[ATTR3]] {
1680 // CHECK5-NEXT:  entry:
1681 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1682 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
1683 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1684 // CHECK5-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
1685 // CHECK5-NEXT:    [[TMP0:%.*]] = load i8, ptr [[A_ADDR]], align 1
1686 // CHECK5-NEXT:    store i8 [[TMP0]], ptr [[DOTCAPTURE_EXPR_]], align 1
1687 // CHECK5-NEXT:    [[TMP1:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
1688 // CHECK5-NEXT:    store i8 [[TMP1]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
1689 // CHECK5-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
1690 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined, i64 [[TMP2]])
1691 // CHECK5-NEXT:    ret void
1692 //
1693 //
1694 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined
1695 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
1696 // CHECK5-NEXT:  entry:
1697 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1698 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1699 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1700 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1701 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1702 // CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1703 // CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1704 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1705 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1706 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
1707 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1708 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1709 // CHECK5-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
1710 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1711 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
1712 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1713 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1714 // CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1715 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
1716 // 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)
1717 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1718 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1719 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1720 // CHECK5:       cond.true:
1721 // CHECK5-NEXT:    br label [[COND_END:%.*]]
1722 // CHECK5:       cond.false:
1723 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1724 // CHECK5-NEXT:    br label [[COND_END]]
1725 // CHECK5:       cond.end:
1726 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1727 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1728 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1729 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
1730 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1731 // CHECK5:       omp.inner.for.cond:
1732 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1733 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1734 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1735 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1736 // CHECK5:       omp.inner.for.body:
1737 // CHECK5-NEXT:    [[TMP7:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
1738 // CHECK5-NEXT:    [[TMP8:%.*]] = sext i8 [[TMP7]] to i32
1739 // CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP8]])
1740 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1741 // CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1742 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1743 // CHECK5-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
1744 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined, i64 [[TMP10]], i64 [[TMP12]])
1745 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1746 // CHECK5:       omp.inner.for.inc:
1747 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1748 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1749 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1750 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1751 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
1752 // CHECK5:       omp.inner.for.end:
1753 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1754 // CHECK5:       omp.loop.exit:
1755 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
1756 // CHECK5-NEXT:    ret void
1757 //
1758 //
1759 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l55.omp_outlined.omp_outlined
1760 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
1761 // CHECK5-NEXT:  entry:
1762 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1763 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1764 // CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1765 // CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1766 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1767 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1768 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1769 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1770 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1771 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1772 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
1773 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1774 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1775 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1776 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1777 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1778 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
1779 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1780 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1781 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1782 // CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1783 // CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1784 // CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
1785 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1786 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1787 // CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1788 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
1789 // 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)
1790 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1791 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1792 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1793 // CHECK5:       cond.true:
1794 // CHECK5-NEXT:    br label [[COND_END:%.*]]
1795 // CHECK5:       cond.false:
1796 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1797 // CHECK5-NEXT:    br label [[COND_END]]
1798 // CHECK5:       cond.end:
1799 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1800 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1801 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1802 // CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
1803 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1804 // CHECK5:       omp.inner.for.cond:
1805 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1806 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1807 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1808 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1809 // CHECK5:       omp.inner.for.body:
1810 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1811 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1812 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1813 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1814 // CHECK5-NEXT:    invoke void @_Z3foov()
1815 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1816 // CHECK5:       invoke.cont:
1817 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1818 // CHECK5:       omp.body.continue:
1819 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1820 // CHECK5:       omp.inner.for.inc:
1821 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1822 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1823 // CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1824 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
1825 // CHECK5:       omp.inner.for.end:
1826 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1827 // CHECK5:       omp.loop.exit:
1828 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
1829 // CHECK5-NEXT:    ret void
1830 // CHECK5:       terminate.lpad:
1831 // CHECK5-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
1832 // CHECK5-NEXT:            catch ptr null
1833 // CHECK5-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
1834 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
1835 // CHECK5-NEXT:    unreachable
1836 //
1837 //
1838 // CHECK5-LABEL: define {{[^@]+}}@_Z5tmainIcLi5EEiv
1839 // CHECK5-SAME: () #[[ATTR2]] comdat {
1840 // CHECK5-NEXT:  entry:
1841 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1842 // CHECK5-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1843 // CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1844 // CHECK5-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1845 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1846 // CHECK5-NEXT:    store i32 3, ptr [[TMP0]], align 4
1847 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1848 // CHECK5-NEXT:    store i32 0, ptr [[TMP1]], align 4
1849 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1850 // CHECK5-NEXT:    store ptr null, ptr [[TMP2]], align 8
1851 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1852 // CHECK5-NEXT:    store ptr null, ptr [[TMP3]], align 8
1853 // CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1854 // CHECK5-NEXT:    store ptr null, ptr [[TMP4]], align 8
1855 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1856 // CHECK5-NEXT:    store ptr null, ptr [[TMP5]], align 8
1857 // CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1858 // CHECK5-NEXT:    store ptr null, ptr [[TMP6]], align 8
1859 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1860 // CHECK5-NEXT:    store ptr null, ptr [[TMP7]], align 8
1861 // CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1862 // CHECK5-NEXT:    store i64 100, ptr [[TMP8]], align 8
1863 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1864 // CHECK5-NEXT:    store i64 0, ptr [[TMP9]], align 8
1865 // CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1866 // CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
1867 // CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1868 // CHECK5-NEXT:    store [3 x i32] [i32 5, i32 0, i32 0], ptr [[TMP11]], align 4
1869 // CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1870 // CHECK5-NEXT:    store i32 0, ptr [[TMP12]], align 4
1871 // CHECK5-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 5, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.region_id, ptr [[KERNEL_ARGS]])
1872 // CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1873 // CHECK5-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1874 // CHECK5:       omp_offload.failed:
1875 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36() #[[ATTR4]]
1876 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1877 // CHECK5:       omp_offload.cont:
1878 // CHECK5-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
1879 // CHECK5-NEXT:    store i32 3, ptr [[TMP15]], align 4
1880 // CHECK5-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
1881 // CHECK5-NEXT:    store i32 0, ptr [[TMP16]], align 4
1882 // CHECK5-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
1883 // CHECK5-NEXT:    store ptr null, ptr [[TMP17]], align 8
1884 // CHECK5-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
1885 // CHECK5-NEXT:    store ptr null, ptr [[TMP18]], align 8
1886 // CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
1887 // CHECK5-NEXT:    store ptr null, ptr [[TMP19]], align 8
1888 // CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
1889 // CHECK5-NEXT:    store ptr null, ptr [[TMP20]], align 8
1890 // CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
1891 // CHECK5-NEXT:    store ptr null, ptr [[TMP21]], align 8
1892 // CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
1893 // CHECK5-NEXT:    store ptr null, ptr [[TMP22]], align 8
1894 // CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
1895 // CHECK5-NEXT:    store i64 100, ptr [[TMP23]], align 8
1896 // CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
1897 // CHECK5-NEXT:    store i64 0, ptr [[TMP24]], align 8
1898 // CHECK5-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
1899 // CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
1900 // CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
1901 // CHECK5-NEXT:    store [3 x i32] [i32 23, i32 0, i32 0], ptr [[TMP26]], align 4
1902 // CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
1903 // CHECK5-NEXT:    store i32 0, ptr [[TMP27]], align 4
1904 // CHECK5-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 23, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.region_id, ptr [[KERNEL_ARGS2]])
1905 // CHECK5-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
1906 // CHECK5-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
1907 // CHECK5:       omp_offload.failed3:
1908 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40() #[[ATTR4]]
1909 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
1910 // CHECK5:       omp_offload.cont4:
1911 // CHECK5-NEXT:    ret i32 0
1912 //
1913 //
1914 // CHECK5-LABEL: define {{[^@]+}}@_Z5tmainI1SLi1EEiv
1915 // CHECK5-SAME: () #[[ATTR2]] comdat personality ptr @__gxx_personality_v0 {
1916 // CHECK5-NEXT:  entry:
1917 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1918 // CHECK5-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1919 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
1920 // CHECK5-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 8
1921 // CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1922 // CHECK5-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1923 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1924 // CHECK5-NEXT:    store i32 3, ptr [[TMP0]], align 4
1925 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1926 // CHECK5-NEXT:    store i32 0, ptr [[TMP1]], align 4
1927 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1928 // CHECK5-NEXT:    store ptr null, ptr [[TMP2]], align 8
1929 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1930 // CHECK5-NEXT:    store ptr null, ptr [[TMP3]], align 8
1931 // CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1932 // CHECK5-NEXT:    store ptr null, ptr [[TMP4]], align 8
1933 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1934 // CHECK5-NEXT:    store ptr null, ptr [[TMP5]], align 8
1935 // CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1936 // CHECK5-NEXT:    store ptr null, ptr [[TMP6]], align 8
1937 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1938 // CHECK5-NEXT:    store ptr null, ptr [[TMP7]], align 8
1939 // CHECK5-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1940 // CHECK5-NEXT:    store i64 100, ptr [[TMP8]], align 8
1941 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1942 // CHECK5-NEXT:    store i64 0, ptr [[TMP9]], align 8
1943 // CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1944 // CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
1945 // CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1946 // CHECK5-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP11]], align 4
1947 // CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1948 // CHECK5-NEXT:    store i32 0, ptr [[TMP12]], align 4
1949 // CHECK5-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 1, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.region_id, ptr [[KERNEL_ARGS]])
1950 // CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1951 // CHECK5-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1952 // CHECK5:       omp_offload.failed:
1953 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36() #[[ATTR4]]
1954 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1955 // CHECK5:       omp_offload.cont:
1956 // CHECK5-NEXT:    invoke void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 noundef 23)
1957 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1958 // CHECK5:       invoke.cont:
1959 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
1960 // CHECK5-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR4]]
1961 // CHECK5-NEXT:    store i8 [[CALL]], ptr [[DOTCAPTURE_EXPR_]], align 1
1962 // CHECK5-NEXT:    [[TMP15:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
1963 // CHECK5-NEXT:    [[TMP16:%.*]] = zext i8 [[TMP15]] to i32
1964 // CHECK5-NEXT:    [[TMP17:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP16]], 0
1965 // CHECK5-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
1966 // CHECK5-NEXT:    store i32 3, ptr [[TMP18]], align 4
1967 // CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
1968 // CHECK5-NEXT:    store i32 0, ptr [[TMP19]], align 4
1969 // CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
1970 // CHECK5-NEXT:    store ptr null, ptr [[TMP20]], align 8
1971 // CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
1972 // CHECK5-NEXT:    store ptr null, ptr [[TMP21]], align 8
1973 // CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
1974 // CHECK5-NEXT:    store ptr null, ptr [[TMP22]], align 8
1975 // CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
1976 // CHECK5-NEXT:    store ptr null, ptr [[TMP23]], align 8
1977 // CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
1978 // CHECK5-NEXT:    store ptr null, ptr [[TMP24]], align 8
1979 // CHECK5-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
1980 // CHECK5-NEXT:    store ptr null, ptr [[TMP25]], align 8
1981 // CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
1982 // CHECK5-NEXT:    store i64 100, ptr [[TMP26]], align 8
1983 // CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
1984 // CHECK5-NEXT:    store i64 0, ptr [[TMP27]], align 8
1985 // CHECK5-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
1986 // CHECK5-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP28]], align 4
1987 // CHECK5-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
1988 // CHECK5-NEXT:    store [3 x i32] [[TMP17]], ptr [[TMP29]], align 4
1989 // CHECK5-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
1990 // CHECK5-NEXT:    store i32 0, ptr [[TMP30]], align 4
1991 // CHECK5-NEXT:    [[TMP31:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.region_id, ptr [[KERNEL_ARGS2]])
1992 // CHECK5-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
1993 // CHECK5-NEXT:    br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
1994 // CHECK5:       omp_offload.failed3:
1995 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40() #[[ATTR4]]
1996 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
1997 // CHECK5:       omp_offload.cont4:
1998 // CHECK5-NEXT:    ret i32 0
1999 // CHECK5:       terminate.lpad:
2000 // CHECK5-NEXT:    [[TMP33:%.*]] = landingpad { ptr, i32 }
2001 // CHECK5-NEXT:            catch ptr null
2002 // CHECK5-NEXT:    [[TMP34:%.*]] = extractvalue { ptr, i32 } [[TMP33]], 0
2003 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP34]]) #[[ATTR7]]
2004 // CHECK5-NEXT:    unreachable
2005 //
2006 //
2007 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SD1Ev
2008 // CHECK5-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR2]] comdat {
2009 // CHECK5-NEXT:  entry:
2010 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
2011 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
2012 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
2013 // CHECK5-NEXT:    call void @_ZN1SD2Ev(ptr noundef nonnull align 8 dereferenceable(24) [[THIS1]]) #[[ATTR4]]
2014 // CHECK5-NEXT:    ret void
2015 //
2016 //
2017 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SC2El
2018 // CHECK5-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]], i64 noundef [[A:%.*]]) unnamed_addr #[[ATTR2]] comdat {
2019 // CHECK5-NEXT:  entry:
2020 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
2021 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2022 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
2023 // CHECK5-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
2024 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
2025 // CHECK5-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
2026 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_ADDR]], align 8
2027 // CHECK5-NEXT:    store i64 [[TMP0]], ptr [[A2]], align 8
2028 // CHECK5-NEXT:    ret void
2029 //
2030 //
2031 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36
2032 // CHECK5-SAME: () #[[ATTR3]] {
2033 // CHECK5-NEXT:  entry:
2034 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined)
2035 // CHECK5-NEXT:    ret void
2036 //
2037 //
2038 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined
2039 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
2040 // CHECK5-NEXT:  entry:
2041 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2042 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2043 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2044 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2045 // CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2046 // CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2047 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2048 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2049 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2050 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2051 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2052 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2053 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
2054 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2055 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2056 // CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2057 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
2058 // 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)
2059 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2060 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2061 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2062 // CHECK5:       cond.true:
2063 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2064 // CHECK5:       cond.false:
2065 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2066 // CHECK5-NEXT:    br label [[COND_END]]
2067 // CHECK5:       cond.end:
2068 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2069 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2070 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2071 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
2072 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2073 // CHECK5:       omp.inner.for.cond:
2074 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2075 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2076 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2077 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2078 // CHECK5:       omp.inner.for.body:
2079 // CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 5)
2080 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2081 // CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2082 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2083 // CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2084 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
2085 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2086 // CHECK5:       omp.inner.for.inc:
2087 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2088 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2089 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2090 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2091 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2092 // CHECK5:       omp.inner.for.end:
2093 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2094 // CHECK5:       omp.loop.exit:
2095 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
2096 // CHECK5-NEXT:    ret void
2097 //
2098 //
2099 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l36.omp_outlined.omp_outlined
2100 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
2101 // CHECK5-NEXT:  entry:
2102 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2103 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2104 // CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2105 // CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2106 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2107 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2108 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2109 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2110 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2111 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2112 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2113 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2114 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2115 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2116 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2117 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2118 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
2119 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2120 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2121 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2122 // CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2123 // CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
2124 // CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
2125 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2126 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2127 // CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2128 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
2129 // 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)
2130 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2131 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2132 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2133 // CHECK5:       cond.true:
2134 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2135 // CHECK5:       cond.false:
2136 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2137 // CHECK5-NEXT:    br label [[COND_END]]
2138 // CHECK5:       cond.end:
2139 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2140 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2141 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2142 // CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
2143 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2144 // CHECK5:       omp.inner.for.cond:
2145 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2146 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2147 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2148 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2149 // CHECK5:       omp.inner.for.body:
2150 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2151 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2152 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2153 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2154 // CHECK5-NEXT:    invoke void @_Z3foov()
2155 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
2156 // CHECK5:       invoke.cont:
2157 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2158 // CHECK5:       omp.body.continue:
2159 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2160 // CHECK5:       omp.inner.for.inc:
2161 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2162 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2163 // CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
2164 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2165 // CHECK5:       omp.inner.for.end:
2166 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2167 // CHECK5:       omp.loop.exit:
2168 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
2169 // CHECK5-NEXT:    ret void
2170 // CHECK5:       terminate.lpad:
2171 // CHECK5-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
2172 // CHECK5-NEXT:            catch ptr null
2173 // CHECK5-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
2174 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
2175 // CHECK5-NEXT:    unreachable
2176 //
2177 //
2178 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40
2179 // CHECK5-SAME: () #[[ATTR3]] {
2180 // CHECK5-NEXT:  entry:
2181 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined)
2182 // CHECK5-NEXT:    ret void
2183 //
2184 //
2185 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined
2186 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
2187 // CHECK5-NEXT:  entry:
2188 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2189 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2190 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2191 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2192 // CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2193 // CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2194 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2195 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2196 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2197 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2198 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2199 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2200 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
2201 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2202 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2203 // CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2204 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
2205 // 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)
2206 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2207 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2208 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2209 // CHECK5:       cond.true:
2210 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2211 // CHECK5:       cond.false:
2212 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2213 // CHECK5-NEXT:    br label [[COND_END]]
2214 // CHECK5:       cond.end:
2215 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2216 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2217 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2218 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
2219 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2220 // CHECK5:       omp.inner.for.cond:
2221 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2222 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2223 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2224 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2225 // CHECK5:       omp.inner.for.body:
2226 // CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 23)
2227 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2228 // CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2229 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2230 // CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2231 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
2232 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2233 // CHECK5:       omp.inner.for.inc:
2234 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2235 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2236 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2237 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2238 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2239 // CHECK5:       omp.inner.for.end:
2240 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2241 // CHECK5:       omp.loop.exit:
2242 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
2243 // CHECK5-NEXT:    ret void
2244 //
2245 //
2246 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIcLi5EEiv_l40.omp_outlined.omp_outlined
2247 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
2248 // CHECK5-NEXT:  entry:
2249 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2250 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2251 // CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2252 // CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2253 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2254 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2255 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2256 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2257 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2258 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2259 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2260 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2261 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2262 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2263 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2264 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2265 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
2266 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2267 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2268 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2269 // CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2270 // CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
2271 // CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
2272 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2273 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2274 // CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2275 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
2276 // 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)
2277 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2278 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2279 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2280 // CHECK5:       cond.true:
2281 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2282 // CHECK5:       cond.false:
2283 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2284 // CHECK5-NEXT:    br label [[COND_END]]
2285 // CHECK5:       cond.end:
2286 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2287 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2288 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2289 // CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
2290 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2291 // CHECK5:       omp.inner.for.cond:
2292 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2293 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2294 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2295 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2296 // CHECK5:       omp.inner.for.body:
2297 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2298 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2299 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2300 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2301 // CHECK5-NEXT:    invoke void @_Z3foov()
2302 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
2303 // CHECK5:       invoke.cont:
2304 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2305 // CHECK5:       omp.body.continue:
2306 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2307 // CHECK5:       omp.inner.for.inc:
2308 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2309 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2310 // CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
2311 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2312 // CHECK5:       omp.inner.for.end:
2313 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2314 // CHECK5:       omp.loop.exit:
2315 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
2316 // CHECK5-NEXT:    ret void
2317 // CHECK5:       terminate.lpad:
2318 // CHECK5-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
2319 // CHECK5-NEXT:            catch ptr null
2320 // CHECK5-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
2321 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
2322 // CHECK5-NEXT:    unreachable
2323 //
2324 //
2325 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36
2326 // CHECK5-SAME: () #[[ATTR3]] {
2327 // CHECK5-NEXT:  entry:
2328 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined)
2329 // CHECK5-NEXT:    ret void
2330 //
2331 //
2332 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined
2333 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3]] {
2334 // CHECK5-NEXT:  entry:
2335 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2336 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2337 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2338 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2339 // CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2340 // CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2341 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2342 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2343 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2344 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2345 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2346 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2347 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
2348 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2349 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2350 // CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2351 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
2352 // 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)
2353 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2354 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2355 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2356 // CHECK5:       cond.true:
2357 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2358 // CHECK5:       cond.false:
2359 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2360 // CHECK5-NEXT:    br label [[COND_END]]
2361 // CHECK5:       cond.end:
2362 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2363 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2364 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2365 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
2366 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2367 // CHECK5:       omp.inner.for.cond:
2368 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2369 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2370 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2371 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2372 // CHECK5:       omp.inner.for.body:
2373 // CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 1)
2374 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2375 // CHECK5-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2376 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2377 // CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2378 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
2379 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2380 // CHECK5:       omp.inner.for.inc:
2381 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2382 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2383 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2384 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2385 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2386 // CHECK5:       omp.inner.for.end:
2387 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2388 // CHECK5:       omp.loop.exit:
2389 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
2390 // CHECK5-NEXT:    ret void
2391 //
2392 //
2393 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l36.omp_outlined.omp_outlined
2394 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
2395 // CHECK5-NEXT:  entry:
2396 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2397 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2398 // CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2399 // CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2400 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2401 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2402 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2403 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2404 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2405 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2406 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2407 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2408 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2409 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2410 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2411 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2412 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
2413 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2414 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2415 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2416 // CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2417 // CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
2418 // CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
2419 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2420 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2421 // CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2422 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
2423 // 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)
2424 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2425 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2426 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2427 // CHECK5:       cond.true:
2428 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2429 // CHECK5:       cond.false:
2430 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2431 // CHECK5-NEXT:    br label [[COND_END]]
2432 // CHECK5:       cond.end:
2433 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2434 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2435 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2436 // CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
2437 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2438 // CHECK5:       omp.inner.for.cond:
2439 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2440 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2441 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2442 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2443 // CHECK5:       omp.inner.for.body:
2444 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2445 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2446 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2447 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2448 // CHECK5-NEXT:    invoke void @_Z3foov()
2449 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
2450 // CHECK5:       invoke.cont:
2451 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2452 // CHECK5:       omp.body.continue:
2453 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2454 // CHECK5:       omp.inner.for.inc:
2455 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2456 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2457 // CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
2458 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2459 // CHECK5:       omp.inner.for.end:
2460 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2461 // CHECK5:       omp.loop.exit:
2462 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
2463 // CHECK5-NEXT:    ret void
2464 // CHECK5:       terminate.lpad:
2465 // CHECK5-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
2466 // CHECK5-NEXT:            catch ptr null
2467 // CHECK5-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
2468 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
2469 // CHECK5-NEXT:    unreachable
2470 //
2471 //
2472 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40
2473 // CHECK5-SAME: () #[[ATTR3]] personality ptr @__gxx_personality_v0 {
2474 // CHECK5-NEXT:  entry:
2475 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2476 // CHECK5-NEXT:    [[REF_TMP:%.*]] = alloca [[STRUCT_S:%.*]], align 8
2477 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2478 // CHECK5-NEXT:    invoke void @_ZN1SC1El(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]], i64 noundef 23)
2479 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
2480 // CHECK5:       invoke.cont:
2481 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef signext i8 @_ZN1ScvcEv(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
2482 // CHECK5-NEXT:    call void @_ZN1SD1Ev(ptr noundef nonnull align 8 dereferenceable(24) [[REF_TMP]]) #[[ATTR4]]
2483 // CHECK5-NEXT:    store i8 [[CALL]], ptr [[DOTCAPTURE_EXPR_]], align 1
2484 // CHECK5-NEXT:    [[TMP0:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR_]], align 1
2485 // CHECK5-NEXT:    store i8 [[TMP0]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 1
2486 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
2487 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined, i64 [[TMP1]])
2488 // CHECK5-NEXT:    ret void
2489 // CHECK5:       terminate.lpad:
2490 // CHECK5-NEXT:    [[TMP2:%.*]] = landingpad { ptr, i32 }
2491 // CHECK5-NEXT:            catch ptr null
2492 // CHECK5-NEXT:    [[TMP3:%.*]] = extractvalue { ptr, i32 } [[TMP2]], 0
2493 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP3]]) #[[ATTR7]]
2494 // CHECK5-NEXT:    unreachable
2495 //
2496 //
2497 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined
2498 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
2499 // CHECK5-NEXT:  entry:
2500 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2501 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2502 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2503 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2504 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2505 // CHECK5-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2506 // CHECK5-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2507 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2508 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2509 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2510 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2511 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2512 // CHECK5-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
2513 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2514 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
2515 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2516 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2517 // CHECK5-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2518 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
2519 // 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)
2520 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2521 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2522 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2523 // CHECK5:       cond.true:
2524 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2525 // CHECK5:       cond.false:
2526 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2527 // CHECK5-NEXT:    br label [[COND_END]]
2528 // CHECK5:       cond.end:
2529 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2530 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2531 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2532 // CHECK5-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
2533 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2534 // CHECK5:       omp.inner.for.cond:
2535 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2536 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2537 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2538 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2539 // CHECK5:       omp.inner.for.body:
2540 // CHECK5-NEXT:    [[TMP7:%.*]] = load i8, ptr [[DOTCAPTURE_EXPR__ADDR]], align 1
2541 // CHECK5-NEXT:    [[TMP8:%.*]] = sext i8 [[TMP7]] to i32
2542 // CHECK5-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB3]], i32 [[TMP1]], i32 [[TMP8]])
2543 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2544 // CHECK5-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2545 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2546 // CHECK5-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
2547 // CHECK5-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined.omp_outlined, i64 [[TMP10]], i64 [[TMP12]])
2548 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2549 // CHECK5:       omp.inner.for.inc:
2550 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2551 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2552 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
2553 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2554 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2555 // CHECK5:       omp.inner.for.end:
2556 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2557 // CHECK5:       omp.loop.exit:
2558 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
2559 // CHECK5-NEXT:    ret void
2560 //
2561 //
2562 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainI1SLi1EEiv_l40.omp_outlined.omp_outlined
2563 // CHECK5-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR3]] personality ptr @__gxx_personality_v0 {
2564 // CHECK5-NEXT:  entry:
2565 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2566 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2567 // CHECK5-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2568 // CHECK5-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2569 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2570 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2571 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2572 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2573 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2574 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2575 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
2576 // CHECK5-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2577 // CHECK5-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2578 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2579 // CHECK5-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2580 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2581 // CHECK5-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
2582 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2583 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2584 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2585 // CHECK5-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2586 // CHECK5-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
2587 // CHECK5-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
2588 // CHECK5-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2589 // CHECK5-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2590 // CHECK5-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2591 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
2592 // 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)
2593 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2594 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2595 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2596 // CHECK5:       cond.true:
2597 // CHECK5-NEXT:    br label [[COND_END:%.*]]
2598 // CHECK5:       cond.false:
2599 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2600 // CHECK5-NEXT:    br label [[COND_END]]
2601 // CHECK5:       cond.end:
2602 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2603 // CHECK5-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2604 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2605 // CHECK5-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
2606 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2607 // CHECK5:       omp.inner.for.cond:
2608 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2609 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2610 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2611 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2612 // CHECK5:       omp.inner.for.body:
2613 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2614 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2615 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2616 // CHECK5-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2617 // CHECK5-NEXT:    invoke void @_Z3foov()
2618 // CHECK5-NEXT:            to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
2619 // CHECK5:       invoke.cont:
2620 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2621 // CHECK5:       omp.body.continue:
2622 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2623 // CHECK5:       omp.inner.for.inc:
2624 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2625 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2626 // CHECK5-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
2627 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]]
2628 // CHECK5:       omp.inner.for.end:
2629 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2630 // CHECK5:       omp.loop.exit:
2631 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
2632 // CHECK5-NEXT:    ret void
2633 // CHECK5:       terminate.lpad:
2634 // CHECK5-NEXT:    [[TMP11:%.*]] = landingpad { ptr, i32 }
2635 // CHECK5-NEXT:            catch ptr null
2636 // CHECK5-NEXT:    [[TMP12:%.*]] = extractvalue { ptr, i32 } [[TMP11]], 0
2637 // CHECK5-NEXT:    call void @__clang_call_terminate(ptr [[TMP12]]) #[[ATTR7]]
2638 // CHECK5-NEXT:    unreachable
2639 //
2640 //
2641 // CHECK5-LABEL: define {{[^@]+}}@_ZN1SD2Ev
2642 // CHECK5-SAME: (ptr noundef nonnull align 8 dereferenceable(24) [[THIS:%.*]]) unnamed_addr #[[ATTR2]] comdat {
2643 // CHECK5-NEXT:  entry:
2644 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
2645 // CHECK5-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
2646 // CHECK5-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
2647 // CHECK5-NEXT:    ret void
2648 //