clang/test/OpenMP/target_teams_distribute_dist_schedule_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 // expected-no-diagnostics
   3 #ifndef HEADER
   4 #define HEADER
   5
   6 // Test host codegen.
   7 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
   8 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
   9 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
  10 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
  11 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  12 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
  13
  14 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  15 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  16 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  17 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  18 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  19 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  20 #ifdef CK1
  21
  22 template <typename T, int X, long long Y>
  23 struct SS{
  24   T a[X];
  25   float b;
  26   int foo(void) {
  27
  28     #pragma omp target teams distribute
  29     for(int i = 0; i < X; i++) {
  30       a[i] = (T)0;
  31     }
  32     #pragma omp target teams distribute dist_schedule(static)
  33     for(int i = 0; i < X; i++) {
  34       a[i] = (T)0;
  35     }
  36     #pragma omp target teams distribute dist_schedule(static, X/2)
  37     for(int i = 0; i < X; i++) {
  38       a[i] = (T)0;
  39     }
  40
  41
  42
  43
  44
  45
  46     return a[0];
  47   }
  48 };
  49
  50 int teams_template_struct(void) {
  51   SS<int, 123, 456> V;
  52   return V.foo();
  53
  54 }
  55 #endif // CK1
  56
  57 // Test host codegen.
  58 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
  59 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  60 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
  61 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
  62 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  63 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
  64
  65 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  66 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  67 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  68 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  69 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  70 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  71 #ifdef CK2
  72
  73 template <typename T, int n>
  74 int tmain(T argc) {
  75   T a[n];
  76 #pragma omp target teams distribute
  77   for(int i = 0; i < n; i++) {
  78     a[i] = (T)0;
  79   }
  80 #pragma omp target teams distribute dist_schedule(static)
  81   for(int i = 0; i < n; i++) {
  82     a[i] = (T)0;
  83   }
  84 #pragma omp target teams distribute dist_schedule(static, n)
  85   for(int i = 0; i < n; i++) {
  86     a[i] = (T)0;
  87   }
  88   return 0;
  89 }
  90
  91 int main (int argc, char **argv) {
  92   int n = 100;
  93   int a[n];
  94 #pragma omp target teams distribute
  95   for(int i = 0; i < n; i++) {
  96     a[i] = 0;
  97   }
  98 #pragma omp target teams distribute dist_schedule(static)
  99   for(int i = 0; i < n; i++) {
 100     a[i] = 0;
 101   }
 102 #pragma omp target teams distribute dist_schedule(static, n)
 103   for(int i = 0; i < n; i++) {
 104     a[i] = 0;
 105   }
 106   return tmain<int, 10>(argc);
 107 }
 108
 109
 110
 111
 112
 113
 114
 115
 116
 117
 118
 119
 120
 121
 122
 123 #endif // CK2
 124 #endif // #ifndef HEADER
 125 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 126 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 127 // CHECK1-NEXT:  entry:
 128 // CHECK1-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 129 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(496) [[V]])
 130 // CHECK1-NEXT:    ret i32 [[CALL]]
 131 //
 132 //
 133 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 134 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat {
 135 // CHECK1-NEXT:  entry:
 136 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 137 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 138 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 139 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 140 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 141 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 142 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 8
 143 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 8
 144 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 8
 145 // CHECK1-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
 146 // CHECK1-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 147 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x ptr], align 8
 148 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x ptr], align 8
 149 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x ptr], align 8
 150 // CHECK1-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
 151 // CHECK1-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 152 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 153 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 154 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 155 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 156 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 8
 157 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 158 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
 159 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 160 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 161 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 162 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 163 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 164 // CHECK1-NEXT:    store i32 3, ptr [[TMP5]], align 4
 165 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 166 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
 167 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 168 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 169 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 170 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 171 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 172 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 8
 173 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 174 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 8
 175 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 176 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 177 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 178 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 179 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 180 // CHECK1-NEXT:    store i64 123, ptr [[TMP13]], align 8
 181 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 182 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
 183 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 184 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 185 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 186 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 187 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 188 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
 189 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 190 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 191 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 192 // CHECK1:       omp_offload.failed:
 193 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 194 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 195 // CHECK1:       omp_offload.cont:
 196 // CHECK1-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 197 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 198 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP20]], align 8
 199 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 200 // CHECK1-NEXT:    store ptr [[A2]], ptr [[TMP21]], align 8
 201 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
 202 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 203 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 204 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 205 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 0
 206 // CHECK1-NEXT:    store i32 3, ptr [[TMP25]], align 4
 207 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 1
 208 // CHECK1-NEXT:    store i32 1, ptr [[TMP26]], align 4
 209 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 2
 210 // CHECK1-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
 211 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 3
 212 // CHECK1-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
 213 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 4
 214 // CHECK1-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP29]], align 8
 215 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 5
 216 // CHECK1-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP30]], align 8
 217 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 6
 218 // CHECK1-NEXT:    store ptr null, ptr [[TMP31]], align 8
 219 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 7
 220 // CHECK1-NEXT:    store ptr null, ptr [[TMP32]], align 8
 221 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 8
 222 // CHECK1-NEXT:    store i64 123, ptr [[TMP33]], align 8
 223 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 9
 224 // CHECK1-NEXT:    store i64 0, ptr [[TMP34]], align 8
 225 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 10
 226 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
 227 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 11
 228 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
 229 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 12
 230 // CHECK1-NEXT:    store i32 0, ptr [[TMP37]], align 4
 231 // CHECK1-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32.region_id, ptr [[KERNEL_ARGS7]])
 232 // CHECK1-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 233 // CHECK1-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
 234 // CHECK1:       omp_offload.failed8:
 235 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32(ptr [[THIS1]]) #[[ATTR2]]
 236 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
 237 // CHECK1:       omp_offload.cont9:
 238 // CHECK1-NEXT:    [[A10:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 239 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 240 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP40]], align 8
 241 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 242 // CHECK1-NEXT:    store ptr [[A10]], ptr [[TMP41]], align 8
 243 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
 244 // CHECK1-NEXT:    store ptr null, ptr [[TMP42]], align 8
 245 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 246 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 247 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
 248 // CHECK1-NEXT:    store i32 3, ptr [[TMP45]], align 4
 249 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
 250 // CHECK1-NEXT:    store i32 1, ptr [[TMP46]], align 4
 251 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
 252 // CHECK1-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 8
 253 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
 254 // CHECK1-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 8
 255 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
 256 // CHECK1-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP49]], align 8
 257 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
 258 // CHECK1-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP50]], align 8
 259 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
 260 // CHECK1-NEXT:    store ptr null, ptr [[TMP51]], align 8
 261 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
 262 // CHECK1-NEXT:    store ptr null, ptr [[TMP52]], align 8
 263 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
 264 // CHECK1-NEXT:    store i64 123, ptr [[TMP53]], align 8
 265 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
 266 // CHECK1-NEXT:    store i64 0, ptr [[TMP54]], align 8
 267 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
 268 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
 269 // CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
 270 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
 271 // CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 272 // CHECK1-NEXT:    store i32 0, ptr [[TMP57]], align 4
 273 // CHECK1-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, ptr [[KERNEL_ARGS15]])
 274 // CHECK1-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
 275 // CHECK1-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 276 // CHECK1:       omp_offload.failed16:
 277 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(ptr [[THIS1]]) #[[ATTR2]]
 278 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
 279 // CHECK1:       omp_offload.cont17:
 280 // CHECK1-NEXT:    [[A18:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 281 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A18]], i64 0, i64 0
 282 // CHECK1-NEXT:    [[TMP60:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 283 // CHECK1-NEXT:    ret i32 [[TMP60]]
 284 //
 285 //
 286 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 287 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 288 // CHECK1-NEXT:  entry:
 289 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 290 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 291 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 292 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 293 // CHECK1-NEXT:    ret void
 294 //
 295 //
 296 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 297 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 298 // CHECK1-NEXT:  entry:
 299 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 300 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 301 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 302 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 303 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 304 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 305 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 306 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 307 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 308 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 309 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 310 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 311 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 312 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 313 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 314 // CHECK1-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 315 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 316 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 317 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 318 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 319 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 320 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 321 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 322 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 323 // CHECK1:       cond.true:
 324 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 325 // CHECK1:       cond.false:
 326 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 327 // CHECK1-NEXT:    br label [[COND_END]]
 328 // CHECK1:       cond.end:
 329 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 330 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 331 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 332 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 333 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 334 // CHECK1:       omp.inner.for.cond:
 335 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 336 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 337 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 338 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 339 // CHECK1:       omp.inner.for.body:
 340 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 341 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 342 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 343 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 344 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 345 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 346 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 347 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 348 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 349 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 350 // CHECK1:       omp.body.continue:
 351 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 352 // CHECK1:       omp.inner.for.inc:
 353 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 354 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 355 // CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 356 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 357 // CHECK1:       omp.inner.for.end:
 358 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 359 // CHECK1:       omp.loop.exit:
 360 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 361 // CHECK1-NEXT:    ret void
 362 //
 363 //
 364 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32
 365 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 366 // CHECK1-NEXT:  entry:
 367 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 368 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 369 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 370 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32.omp_outlined, ptr [[TMP0]])
 371 // CHECK1-NEXT:    ret void
 372 //
 373 //
 374 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32.omp_outlined
 375 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 376 // CHECK1-NEXT:  entry:
 377 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 378 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 379 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 380 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 381 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 382 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 383 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 384 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 385 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 386 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 387 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 388 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 389 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 390 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 391 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 392 // CHECK1-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 393 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 394 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 395 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 396 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 397 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 398 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 399 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 400 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 401 // CHECK1:       cond.true:
 402 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 403 // CHECK1:       cond.false:
 404 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 405 // CHECK1-NEXT:    br label [[COND_END]]
 406 // CHECK1:       cond.end:
 407 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 408 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 409 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 410 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 411 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 412 // CHECK1:       omp.inner.for.cond:
 413 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 414 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 415 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 416 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 417 // CHECK1:       omp.inner.for.body:
 418 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 419 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 420 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 421 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 422 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 423 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 424 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 425 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 426 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 427 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 428 // CHECK1:       omp.body.continue:
 429 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 430 // CHECK1:       omp.inner.for.inc:
 431 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 432 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 433 // CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 434 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 435 // CHECK1:       omp.inner.for.end:
 436 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 437 // CHECK1:       omp.loop.exit:
 438 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 439 // CHECK1-NEXT:    ret void
 440 //
 441 //
 442 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
 443 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 444 // CHECK1-NEXT:  entry:
 445 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 446 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 447 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 448 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.omp_outlined, ptr [[TMP0]])
 449 // CHECK1-NEXT:    ret void
 450 //
 451 //
 452 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.omp_outlined
 453 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 454 // CHECK1-NEXT:  entry:
 455 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 456 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 457 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 458 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 459 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 460 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 461 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 462 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 463 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 464 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 465 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 466 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 467 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 468 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 469 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 470 // CHECK1-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 471 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 472 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 473 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 474 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 475 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 61)
 476 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 477 // CHECK1:       omp.dispatch.cond:
 478 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 479 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 480 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 481 // CHECK1:       cond.true:
 482 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 483 // CHECK1:       cond.false:
 484 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 485 // CHECK1-NEXT:    br label [[COND_END]]
 486 // CHECK1:       cond.end:
 487 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 488 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 489 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 490 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 491 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 492 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 493 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 494 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 495 // CHECK1:       omp.dispatch.body:
 496 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 497 // CHECK1:       omp.inner.for.cond:
 498 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8:![0-9]+]]
 499 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP8]]
 500 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 501 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 502 // CHECK1:       omp.inner.for.body:
 503 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 504 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 505 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 506 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP8]]
 507 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 508 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP8]]
 509 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
 510 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 511 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP8]]
 512 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 513 // CHECK1:       omp.body.continue:
 514 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 515 // CHECK1:       omp.inner.for.inc:
 516 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 517 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
 518 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 519 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
 520 // CHECK1:       omp.inner.for.end:
 521 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 522 // CHECK1:       omp.dispatch.inc:
 523 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 524 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 525 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
 526 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
 527 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 528 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 529 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
 530 // CHECK1-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
 531 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 532 // CHECK1:       omp.dispatch.end:
 533 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 534 // CHECK1-NEXT:    ret void
 535 //
 536 //
 537 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 538 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 539 // CHECK3-NEXT:  entry:
 540 // CHECK3-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 541 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(496) [[V]])
 542 // CHECK3-NEXT:    ret i32 [[CALL]]
 543 //
 544 //
 545 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 546 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
 547 // CHECK3-NEXT:  entry:
 548 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 549 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
 550 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
 551 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
 552 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 553 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 554 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 4
 555 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 4
 556 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 4
 557 // CHECK3-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
 558 // CHECK3-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 559 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x ptr], align 4
 560 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x ptr], align 4
 561 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x ptr], align 4
 562 // CHECK3-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
 563 // CHECK3-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 564 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 565 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 566 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 567 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 568 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 4
 569 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 570 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
 571 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 572 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
 573 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 574 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 575 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 576 // CHECK3-NEXT:    store i32 3, ptr [[TMP5]], align 4
 577 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 578 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
 579 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 580 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
 581 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 582 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
 583 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 584 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 4
 585 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 586 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 4
 587 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 588 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
 589 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 590 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 591 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 592 // CHECK3-NEXT:    store i64 123, ptr [[TMP13]], align 8
 593 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 594 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
 595 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 596 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 597 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 598 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 599 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 600 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
 601 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 602 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 603 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 604 // CHECK3:       omp_offload.failed:
 605 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 606 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 607 // CHECK3:       omp_offload.cont:
 608 // CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 609 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 610 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP20]], align 4
 611 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 612 // CHECK3-NEXT:    store ptr [[A2]], ptr [[TMP21]], align 4
 613 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
 614 // CHECK3-NEXT:    store ptr null, ptr [[TMP22]], align 4
 615 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 616 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 617 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 0
 618 // CHECK3-NEXT:    store i32 3, ptr [[TMP25]], align 4
 619 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 1
 620 // CHECK3-NEXT:    store i32 1, ptr [[TMP26]], align 4
 621 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 2
 622 // CHECK3-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 4
 623 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 3
 624 // CHECK3-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 4
 625 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 4
 626 // CHECK3-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP29]], align 4
 627 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 5
 628 // CHECK3-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP30]], align 4
 629 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 6
 630 // CHECK3-NEXT:    store ptr null, ptr [[TMP31]], align 4
 631 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 7
 632 // CHECK3-NEXT:    store ptr null, ptr [[TMP32]], align 4
 633 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 8
 634 // CHECK3-NEXT:    store i64 123, ptr [[TMP33]], align 8
 635 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 9
 636 // CHECK3-NEXT:    store i64 0, ptr [[TMP34]], align 8
 637 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 10
 638 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
 639 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 11
 640 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
 641 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 12
 642 // CHECK3-NEXT:    store i32 0, ptr [[TMP37]], align 4
 643 // CHECK3-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32.region_id, ptr [[KERNEL_ARGS7]])
 644 // CHECK3-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 645 // CHECK3-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
 646 // CHECK3:       omp_offload.failed8:
 647 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32(ptr [[THIS1]]) #[[ATTR2]]
 648 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
 649 // CHECK3:       omp_offload.cont9:
 650 // CHECK3-NEXT:    [[A10:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 651 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 652 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP40]], align 4
 653 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 654 // CHECK3-NEXT:    store ptr [[A10]], ptr [[TMP41]], align 4
 655 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
 656 // CHECK3-NEXT:    store ptr null, ptr [[TMP42]], align 4
 657 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 658 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 659 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
 660 // CHECK3-NEXT:    store i32 3, ptr [[TMP45]], align 4
 661 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
 662 // CHECK3-NEXT:    store i32 1, ptr [[TMP46]], align 4
 663 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
 664 // CHECK3-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 4
 665 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
 666 // CHECK3-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 4
 667 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
 668 // CHECK3-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP49]], align 4
 669 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
 670 // CHECK3-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP50]], align 4
 671 // CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
 672 // CHECK3-NEXT:    store ptr null, ptr [[TMP51]], align 4
 673 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
 674 // CHECK3-NEXT:    store ptr null, ptr [[TMP52]], align 4
 675 // CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
 676 // CHECK3-NEXT:    store i64 123, ptr [[TMP53]], align 8
 677 // CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
 678 // CHECK3-NEXT:    store i64 0, ptr [[TMP54]], align 8
 679 // CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
 680 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
 681 // CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
 682 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
 683 // CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 684 // CHECK3-NEXT:    store i32 0, ptr [[TMP57]], align 4
 685 // CHECK3-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.region_id, ptr [[KERNEL_ARGS15]])
 686 // CHECK3-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
 687 // CHECK3-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 688 // CHECK3:       omp_offload.failed16:
 689 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36(ptr [[THIS1]]) #[[ATTR2]]
 690 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
 691 // CHECK3:       omp_offload.cont17:
 692 // CHECK3-NEXT:    [[A18:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 693 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A18]], i32 0, i32 0
 694 // CHECK3-NEXT:    [[TMP60:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 695 // CHECK3-NEXT:    ret i32 [[TMP60]]
 696 //
 697 //
 698 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 699 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 700 // CHECK3-NEXT:  entry:
 701 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 702 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 703 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 704 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 705 // CHECK3-NEXT:    ret void
 706 //
 707 //
 708 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 709 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 710 // CHECK3-NEXT:  entry:
 711 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 712 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 713 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 714 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 715 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 716 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 717 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 718 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 719 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 720 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 721 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 722 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 723 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 724 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 725 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 726 // CHECK3-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 727 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 728 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 729 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 730 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 731 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 732 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 733 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 734 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 735 // CHECK3:       cond.true:
 736 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 737 // CHECK3:       cond.false:
 738 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 739 // CHECK3-NEXT:    br label [[COND_END]]
 740 // CHECK3:       cond.end:
 741 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 742 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 743 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 744 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 745 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 746 // CHECK3:       omp.inner.for.cond:
 747 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 748 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 749 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 750 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 751 // CHECK3:       omp.inner.for.body:
 752 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 753 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 754 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 755 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 756 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 757 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 758 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP9]]
 759 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 760 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 761 // CHECK3:       omp.body.continue:
 762 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 763 // CHECK3:       omp.inner.for.inc:
 764 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 765 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 766 // CHECK3-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 767 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 768 // CHECK3:       omp.inner.for.end:
 769 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 770 // CHECK3:       omp.loop.exit:
 771 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 772 // CHECK3-NEXT:    ret void
 773 //
 774 //
 775 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32
 776 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 777 // CHECK3-NEXT:  entry:
 778 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 779 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 780 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 781 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32.omp_outlined, ptr [[TMP0]])
 782 // CHECK3-NEXT:    ret void
 783 //
 784 //
 785 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l32.omp_outlined
 786 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 787 // CHECK3-NEXT:  entry:
 788 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 789 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 790 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 791 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 792 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 793 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 794 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 795 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 796 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 797 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 798 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 799 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 800 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 801 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 802 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 803 // CHECK3-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 804 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 805 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 806 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 807 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 808 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 809 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 810 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 811 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 812 // CHECK3:       cond.true:
 813 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 814 // CHECK3:       cond.false:
 815 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 816 // CHECK3-NEXT:    br label [[COND_END]]
 817 // CHECK3:       cond.end:
 818 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 819 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 820 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 821 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 822 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 823 // CHECK3:       omp.inner.for.cond:
 824 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 825 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 826 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 827 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 828 // CHECK3:       omp.inner.for.body:
 829 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 830 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 831 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 832 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 833 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 834 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 835 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP9]]
 836 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 837 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 838 // CHECK3:       omp.body.continue:
 839 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 840 // CHECK3:       omp.inner.for.inc:
 841 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 842 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 843 // CHECK3-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 844 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 845 // CHECK3:       omp.inner.for.end:
 846 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 847 // CHECK3:       omp.loop.exit:
 848 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 849 // CHECK3-NEXT:    ret void
 850 //
 851 //
 852 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36
 853 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 854 // CHECK3-NEXT:  entry:
 855 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 856 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 857 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 858 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.omp_outlined, ptr [[TMP0]])
 859 // CHECK3-NEXT:    ret void
 860 //
 861 //
 862 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l36.omp_outlined
 863 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 864 // CHECK3-NEXT:  entry:
 865 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 866 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 867 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 868 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 869 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 870 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 871 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 872 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 873 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 874 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 875 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 876 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 877 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 878 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 879 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 880 // CHECK3-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 881 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 882 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 883 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 884 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 885 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 61)
 886 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 887 // CHECK3:       omp.dispatch.cond:
 888 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 889 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 890 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 891 // CHECK3:       cond.true:
 892 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 893 // CHECK3:       cond.false:
 894 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 895 // CHECK3-NEXT:    br label [[COND_END]]
 896 // CHECK3:       cond.end:
 897 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 898 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 899 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 900 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 901 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 902 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 903 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 904 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 905 // CHECK3:       omp.dispatch.body:
 906 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 907 // CHECK3:       omp.inner.for.cond:
 908 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
 909 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
 910 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 911 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 912 // CHECK3:       omp.inner.for.body:
 913 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 914 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 915 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 916 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 917 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 918 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 919 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP11]]
 920 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP9]]
 921 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 922 // CHECK3:       omp.body.continue:
 923 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 924 // CHECK3:       omp.inner.for.inc:
 925 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 926 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
 927 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 928 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
 929 // CHECK3:       omp.inner.for.end:
 930 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 931 // CHECK3:       omp.dispatch.inc:
 932 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 933 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 934 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
 935 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
 936 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 937 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 938 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
 939 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
 940 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 941 // CHECK3:       omp.dispatch.end:
 942 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 943 // CHECK3-NEXT:    ret void
 944 //
 945 //
 946 // CHECK9-LABEL: define {{[^@]+}}@main
 947 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 948 // CHECK9-NEXT:  entry:
 949 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 950 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 951 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
 952 // CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
 953 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
 954 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
 955 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 956 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
 957 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
 958 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
 959 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
 960 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 961 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 962 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 963 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 964 // CHECK9-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
 965 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x ptr], align 8
 966 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x ptr], align 8
 967 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x ptr], align 8
 968 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 8
 969 // CHECK9-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
 970 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
 971 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
 972 // CHECK9-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 973 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
 974 // CHECK9-NEXT:    [[N_CASTED19:%.*]] = alloca i64, align 8
 975 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
 976 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [4 x ptr], align 8
 977 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [4 x ptr], align 8
 978 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [4 x ptr], align 8
 979 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES23:%.*]] = alloca [4 x i64], align 8
 980 // CHECK9-NEXT:    [[_TMP24:%.*]] = alloca i32, align 4
 981 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
 982 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
 983 // CHECK9-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 984 // CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 985 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 986 // CHECK9-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
 987 // CHECK9-NEXT:    store i32 100, ptr [[N]], align 4
 988 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 989 // CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
 990 // CHECK9-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
 991 // CHECK9-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
 992 // CHECK9-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
 993 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
 994 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
 995 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
 996 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
 997 // CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
 998 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 24, i1 false)
 999 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1000 // CHECK9-NEXT:    store i64 [[TMP4]], ptr [[TMP6]], align 8
1001 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1002 // CHECK9-NEXT:    store i64 [[TMP4]], ptr [[TMP7]], align 8
1003 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1004 // CHECK9-NEXT:    store ptr null, ptr [[TMP8]], align 8
1005 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1006 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP9]], align 8
1007 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1008 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP10]], align 8
1009 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1010 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
1011 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1012 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 8
1013 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1014 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 8
1015 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1016 // CHECK9-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 8
1017 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1018 // CHECK9-NEXT:    store ptr null, ptr [[TMP15]], align 8
1019 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1020 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1021 // CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1022 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
1023 // CHECK9-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
1024 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1025 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
1026 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1027 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1028 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1029 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1030 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
1031 // CHECK9-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
1032 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1033 // CHECK9-NEXT:    store i32 3, ptr [[TMP23]], align 4
1034 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1035 // CHECK9-NEXT:    store i32 3, ptr [[TMP24]], align 4
1036 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1037 // CHECK9-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 8
1038 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1039 // CHECK9-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 8
1040 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1041 // CHECK9-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 8
1042 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1043 // CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 8
1044 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1045 // CHECK9-NEXT:    store ptr null, ptr [[TMP29]], align 8
1046 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1047 // CHECK9-NEXT:    store ptr null, ptr [[TMP30]], align 8
1048 // CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1049 // CHECK9-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
1050 // CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1051 // CHECK9-NEXT:    store i64 0, ptr [[TMP32]], align 8
1052 // CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1053 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
1054 // CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1055 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
1056 // CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1057 // CHECK9-NEXT:    store i32 0, ptr [[TMP35]], align 4
1058 // CHECK9-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, ptr [[KERNEL_ARGS]])
1059 // CHECK9-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
1060 // CHECK9-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1061 // CHECK9:       omp_offload.failed:
1062 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94(i64 [[TMP4]], i64 [[TMP1]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
1063 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1064 // CHECK9:       omp_offload.cont:
1065 // CHECK9-NEXT:    [[TMP38:%.*]] = load i32, ptr [[N]], align 4
1066 // CHECK9-NEXT:    store i32 [[TMP38]], ptr [[N_CASTED3]], align 4
1067 // CHECK9-NEXT:    [[TMP39:%.*]] = load i64, ptr [[N_CASTED3]], align 8
1068 // CHECK9-NEXT:    [[TMP40:%.*]] = mul nuw i64 [[TMP1]], 4
1069 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES7]], ptr align 8 @.offload_sizes.1, i64 24, i1 false)
1070 // CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
1071 // CHECK9-NEXT:    store i64 [[TMP39]], ptr [[TMP41]], align 8
1072 // CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
1073 // CHECK9-NEXT:    store i64 [[TMP39]], ptr [[TMP42]], align 8
1074 // CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
1075 // CHECK9-NEXT:    store ptr null, ptr [[TMP43]], align 8
1076 // CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
1077 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP44]], align 8
1078 // CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
1079 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP45]], align 8
1080 // CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 1
1081 // CHECK9-NEXT:    store ptr null, ptr [[TMP46]], align 8
1082 // CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
1083 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP47]], align 8
1084 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
1085 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP48]], align 8
1086 // CHECK9-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
1087 // CHECK9-NEXT:    store i64 [[TMP40]], ptr [[TMP49]], align 8
1088 // CHECK9-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 2
1089 // CHECK9-NEXT:    store ptr null, ptr [[TMP50]], align 8
1090 // CHECK9-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
1091 // CHECK9-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
1092 // CHECK9-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
1093 // CHECK9-NEXT:    [[TMP54:%.*]] = load i32, ptr [[N]], align 4
1094 // CHECK9-NEXT:    store i32 [[TMP54]], ptr [[DOTCAPTURE_EXPR_9]], align 4
1095 // CHECK9-NEXT:    [[TMP55:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_9]], align 4
1096 // CHECK9-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP55]], 0
1097 // CHECK9-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
1098 // CHECK9-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
1099 // CHECK9-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_10]], align 4
1100 // CHECK9-NEXT:    [[TMP56:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_10]], align 4
1101 // CHECK9-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP56]], 1
1102 // CHECK9-NEXT:    [[TMP57:%.*]] = zext i32 [[ADD14]] to i64
1103 // CHECK9-NEXT:    [[TMP58:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
1104 // CHECK9-NEXT:    store i32 3, ptr [[TMP58]], align 4
1105 // CHECK9-NEXT:    [[TMP59:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
1106 // CHECK9-NEXT:    store i32 3, ptr [[TMP59]], align 4
1107 // CHECK9-NEXT:    [[TMP60:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
1108 // CHECK9-NEXT:    store ptr [[TMP51]], ptr [[TMP60]], align 8
1109 // CHECK9-NEXT:    [[TMP61:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
1110 // CHECK9-NEXT:    store ptr [[TMP52]], ptr [[TMP61]], align 8
1111 // CHECK9-NEXT:    [[TMP62:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
1112 // CHECK9-NEXT:    store ptr [[TMP53]], ptr [[TMP62]], align 8
1113 // CHECK9-NEXT:    [[TMP63:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
1114 // CHECK9-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP63]], align 8
1115 // CHECK9-NEXT:    [[TMP64:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
1116 // CHECK9-NEXT:    store ptr null, ptr [[TMP64]], align 8
1117 // CHECK9-NEXT:    [[TMP65:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
1118 // CHECK9-NEXT:    store ptr null, ptr [[TMP65]], align 8
1119 // CHECK9-NEXT:    [[TMP66:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
1120 // CHECK9-NEXT:    store i64 [[TMP57]], ptr [[TMP66]], align 8
1121 // CHECK9-NEXT:    [[TMP67:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
1122 // CHECK9-NEXT:    store i64 0, ptr [[TMP67]], align 8
1123 // CHECK9-NEXT:    [[TMP68:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
1124 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP68]], align 4
1125 // CHECK9-NEXT:    [[TMP69:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
1126 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
1127 // CHECK9-NEXT:    [[TMP70:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
1128 // CHECK9-NEXT:    store i32 0, ptr [[TMP70]], align 4
1129 // CHECK9-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98.region_id, ptr [[KERNEL_ARGS15]])
1130 // CHECK9-NEXT:    [[TMP72:%.*]] = icmp ne i32 [[TMP71]], 0
1131 // CHECK9-NEXT:    br i1 [[TMP72]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
1132 // CHECK9:       omp_offload.failed16:
1133 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98(i64 [[TMP39]], i64 [[TMP1]], ptr [[VLA]]) #[[ATTR3]]
1134 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
1135 // CHECK9:       omp_offload.cont17:
1136 // CHECK9-NEXT:    [[TMP73:%.*]] = load i32, ptr [[N]], align 4
1137 // CHECK9-NEXT:    store i32 [[TMP73]], ptr [[DOTCAPTURE_EXPR_18]], align 4
1138 // CHECK9-NEXT:    [[TMP74:%.*]] = load i32, ptr [[N]], align 4
1139 // CHECK9-NEXT:    store i32 [[TMP74]], ptr [[N_CASTED19]], align 4
1140 // CHECK9-NEXT:    [[TMP75:%.*]] = load i64, ptr [[N_CASTED19]], align 8
1141 // CHECK9-NEXT:    [[TMP76:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_18]], align 4
1142 // CHECK9-NEXT:    store i32 [[TMP76]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
1143 // CHECK9-NEXT:    [[TMP77:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
1144 // CHECK9-NEXT:    [[TMP78:%.*]] = mul nuw i64 [[TMP1]], 4
1145 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES23]], ptr align 8 @.offload_sizes.3, i64 32, i1 false)
1146 // CHECK9-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
1147 // CHECK9-NEXT:    store i64 [[TMP75]], ptr [[TMP79]], align 8
1148 // CHECK9-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
1149 // CHECK9-NEXT:    store i64 [[TMP75]], ptr [[TMP80]], align 8
1150 // CHECK9-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 0
1151 // CHECK9-NEXT:    store ptr null, ptr [[TMP81]], align 8
1152 // CHECK9-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
1153 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP82]], align 8
1154 // CHECK9-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
1155 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP83]], align 8
1156 // CHECK9-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 1
1157 // CHECK9-NEXT:    store ptr null, ptr [[TMP84]], align 8
1158 // CHECK9-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
1159 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP85]], align 8
1160 // CHECK9-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
1161 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP86]], align 8
1162 // CHECK9-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x i64], ptr [[DOTOFFLOAD_SIZES23]], i32 0, i32 2
1163 // CHECK9-NEXT:    store i64 [[TMP78]], ptr [[TMP87]], align 8
1164 // CHECK9-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 2
1165 // CHECK9-NEXT:    store ptr null, ptr [[TMP88]], align 8
1166 // CHECK9-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
1167 // CHECK9-NEXT:    store i64 [[TMP77]], ptr [[TMP89]], align 8
1168 // CHECK9-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
1169 // CHECK9-NEXT:    store i64 [[TMP77]], ptr [[TMP90]], align 8
1170 // CHECK9-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 3
1171 // CHECK9-NEXT:    store ptr null, ptr [[TMP91]], align 8
1172 // CHECK9-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
1173 // CHECK9-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
1174 // CHECK9-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x i64], ptr [[DOTOFFLOAD_SIZES23]], i32 0, i32 0
1175 // CHECK9-NEXT:    [[TMP95:%.*]] = load i32, ptr [[N]], align 4
1176 // CHECK9-NEXT:    store i32 [[TMP95]], ptr [[DOTCAPTURE_EXPR_25]], align 4
1177 // CHECK9-NEXT:    [[TMP96:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_25]], align 4
1178 // CHECK9-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP96]], 0
1179 // CHECK9-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
1180 // CHECK9-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[DIV28]], 1
1181 // CHECK9-NEXT:    store i32 [[SUB29]], ptr [[DOTCAPTURE_EXPR_26]], align 4
1182 // CHECK9-NEXT:    [[TMP97:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_26]], align 4
1183 // CHECK9-NEXT:    [[ADD30:%.*]] = add nsw i32 [[TMP97]], 1
1184 // CHECK9-NEXT:    [[TMP98:%.*]] = zext i32 [[ADD30]] to i64
1185 // CHECK9-NEXT:    [[TMP99:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 0
1186 // CHECK9-NEXT:    store i32 3, ptr [[TMP99]], align 4
1187 // CHECK9-NEXT:    [[TMP100:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 1
1188 // CHECK9-NEXT:    store i32 4, ptr [[TMP100]], align 4
1189 // CHECK9-NEXT:    [[TMP101:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 2
1190 // CHECK9-NEXT:    store ptr [[TMP92]], ptr [[TMP101]], align 8
1191 // CHECK9-NEXT:    [[TMP102:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 3
1192 // CHECK9-NEXT:    store ptr [[TMP93]], ptr [[TMP102]], align 8
1193 // CHECK9-NEXT:    [[TMP103:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 4
1194 // CHECK9-NEXT:    store ptr [[TMP94]], ptr [[TMP103]], align 8
1195 // CHECK9-NEXT:    [[TMP104:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 5
1196 // CHECK9-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP104]], align 8
1197 // CHECK9-NEXT:    [[TMP105:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 6
1198 // CHECK9-NEXT:    store ptr null, ptr [[TMP105]], align 8
1199 // CHECK9-NEXT:    [[TMP106:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 7
1200 // CHECK9-NEXT:    store ptr null, ptr [[TMP106]], align 8
1201 // CHECK9-NEXT:    [[TMP107:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 8
1202 // CHECK9-NEXT:    store i64 [[TMP98]], ptr [[TMP107]], align 8
1203 // CHECK9-NEXT:    [[TMP108:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 9
1204 // CHECK9-NEXT:    store i64 0, ptr [[TMP108]], align 8
1205 // CHECK9-NEXT:    [[TMP109:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 10
1206 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP109]], align 4
1207 // CHECK9-NEXT:    [[TMP110:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 11
1208 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP110]], align 4
1209 // CHECK9-NEXT:    [[TMP111:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 12
1210 // CHECK9-NEXT:    store i32 0, ptr [[TMP111]], align 4
1211 // CHECK9-NEXT:    [[TMP112:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.region_id, ptr [[KERNEL_ARGS31]])
1212 // CHECK9-NEXT:    [[TMP113:%.*]] = icmp ne i32 [[TMP112]], 0
1213 // CHECK9-NEXT:    br i1 [[TMP113]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
1214 // CHECK9:       omp_offload.failed32:
1215 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102(i64 [[TMP75]], i64 [[TMP1]], ptr [[VLA]], i64 [[TMP77]]) #[[ATTR3]]
1216 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
1217 // CHECK9:       omp_offload.cont33:
1218 // CHECK9-NEXT:    [[TMP114:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
1219 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP114]])
1220 // CHECK9-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
1221 // CHECK9-NEXT:    [[TMP115:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
1222 // CHECK9-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP115]])
1223 // CHECK9-NEXT:    [[TMP116:%.*]] = load i32, ptr [[RETVAL]], align 4
1224 // CHECK9-NEXT:    ret i32 [[TMP116]]
1225 //
1226 //
1227 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
1228 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1229 // CHECK9-NEXT:  entry:
1230 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1231 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1232 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1233 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
1234 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1235 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1236 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1237 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1238 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1239 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
1240 // CHECK9-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
1241 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[N_CASTED]], align 8
1242 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.omp_outlined, i64 [[TMP3]], i64 [[TMP0]], ptr [[TMP1]])
1243 // CHECK9-NEXT:    ret void
1244 //
1245 //
1246 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.omp_outlined
1247 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1248 // CHECK9-NEXT:  entry:
1249 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1250 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1251 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1252 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1253 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1254 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1255 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1256 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1257 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1258 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1259 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1260 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1261 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1262 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1263 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
1264 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1265 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1266 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1267 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1268 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1269 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1270 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1271 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
1272 // CHECK9-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
1273 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1274 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1275 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1276 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1277 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1278 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1279 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1280 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1281 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1282 // CHECK9:       omp.precond.then:
1283 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1284 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1285 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
1286 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1287 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1288 // CHECK9-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1289 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
1290 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1291 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1292 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1293 // CHECK9-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1294 // CHECK9-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1295 // CHECK9:       cond.true:
1296 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1297 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1298 // CHECK9:       cond.false:
1299 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1300 // CHECK9-NEXT:    br label [[COND_END]]
1301 // CHECK9:       cond.end:
1302 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1303 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1304 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1305 // CHECK9-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
1306 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1307 // CHECK9:       omp.inner.for.cond:
1308 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1309 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1310 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1311 // CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1312 // CHECK9:       omp.inner.for.body:
1313 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1314 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1315 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1316 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
1317 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
1318 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
1319 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM]]
1320 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1321 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1322 // CHECK9:       omp.body.continue:
1323 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1324 // CHECK9:       omp.inner.for.inc:
1325 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1326 // CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
1327 // CHECK9-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
1328 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1329 // CHECK9:       omp.inner.for.end:
1330 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1331 // CHECK9:       omp.loop.exit:
1332 // CHECK9-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1333 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
1334 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
1335 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1336 // CHECK9:       omp.precond.end:
1337 // CHECK9-NEXT:    ret void
1338 //
1339 //
1340 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98
1341 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1342 // CHECK9-NEXT:  entry:
1343 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1344 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1345 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1346 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
1347 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1348 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1349 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1350 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1351 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1352 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
1353 // CHECK9-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
1354 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[N_CASTED]], align 8
1355 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98.omp_outlined, i64 [[TMP3]], i64 [[TMP0]], ptr [[TMP1]])
1356 // CHECK9-NEXT:    ret void
1357 //
1358 //
1359 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98.omp_outlined
1360 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1361 // CHECK9-NEXT:  entry:
1362 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1363 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1364 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1365 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1366 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1367 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1368 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1369 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1370 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1371 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1372 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1373 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1374 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1375 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1376 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
1377 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1378 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1379 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1380 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1381 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1382 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1383 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1384 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
1385 // CHECK9-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
1386 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1387 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1388 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1389 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1390 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1391 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1392 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1393 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1394 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1395 // CHECK9:       omp.precond.then:
1396 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1397 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1398 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
1399 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1400 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1401 // CHECK9-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1402 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
1403 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1404 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1405 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1406 // CHECK9-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1407 // CHECK9-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1408 // CHECK9:       cond.true:
1409 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1410 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1411 // CHECK9:       cond.false:
1412 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1413 // CHECK9-NEXT:    br label [[COND_END]]
1414 // CHECK9:       cond.end:
1415 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1416 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1417 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1418 // CHECK9-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
1419 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1420 // CHECK9:       omp.inner.for.cond:
1421 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1422 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1423 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1424 // CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1425 // CHECK9:       omp.inner.for.body:
1426 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1427 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1428 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1429 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
1430 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
1431 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
1432 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM]]
1433 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1434 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1435 // CHECK9:       omp.body.continue:
1436 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1437 // CHECK9:       omp.inner.for.inc:
1438 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1439 // CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
1440 // CHECK9-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
1441 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1442 // CHECK9:       omp.inner.for.end:
1443 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1444 // CHECK9:       omp.loop.exit:
1445 // CHECK9-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1446 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
1447 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
1448 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1449 // CHECK9:       omp.precond.end:
1450 // CHECK9-NEXT:    ret void
1451 //
1452 //
1453 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102
1454 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1455 // CHECK9-NEXT:  entry:
1456 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1457 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1458 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1459 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1460 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
1461 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1462 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1463 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1464 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1465 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
1466 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1467 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1468 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
1469 // CHECK9-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
1470 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[N_CASTED]], align 8
1471 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
1472 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
1473 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
1474 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined, i64 [[TMP3]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP5]])
1475 // CHECK9-NEXT:    ret void
1476 //
1477 //
1478 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined
1479 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1480 // CHECK9-NEXT:  entry:
1481 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1482 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1483 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1484 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1485 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1486 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1487 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1488 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1489 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1490 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1491 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1492 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1493 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1494 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1495 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1496 // CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
1497 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1498 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1499 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1500 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1501 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1502 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
1503 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1504 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1505 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
1506 // CHECK9-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1507 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1508 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1509 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1510 // CHECK9-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
1511 // CHECK9-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
1512 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1513 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1514 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1515 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1516 // CHECK9:       omp.precond.then:
1517 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1518 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1519 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
1520 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1521 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1522 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
1523 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1524 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1525 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP8]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[TMP6]])
1526 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1527 // CHECK9:       omp.dispatch.cond:
1528 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1529 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1530 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1531 // CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1532 // CHECK9:       cond.true:
1533 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1534 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1535 // CHECK9:       cond.false:
1536 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1537 // CHECK9-NEXT:    br label [[COND_END]]
1538 // CHECK9:       cond.end:
1539 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1540 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1541 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1542 // CHECK9-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
1543 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1544 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1545 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1546 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1547 // CHECK9:       omp.dispatch.body:
1548 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1549 // CHECK9:       omp.inner.for.cond:
1550 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11:![0-9]+]]
1551 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP11]]
1552 // CHECK9-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
1553 // CHECK9-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1554 // CHECK9:       omp.inner.for.body:
1555 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
1556 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
1557 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1558 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP11]]
1559 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP11]]
1560 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
1561 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 [[IDXPROM]]
1562 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP11]]
1563 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1564 // CHECK9:       omp.body.continue:
1565 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1566 // CHECK9:       omp.inner.for.inc:
1567 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
1568 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
1569 // CHECK9-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
1570 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1571 // CHECK9:       omp.inner.for.end:
1572 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1573 // CHECK9:       omp.dispatch.inc:
1574 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1575 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1576 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
1577 // CHECK9-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_LB]], align 4
1578 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1579 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1580 // CHECK9-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
1581 // CHECK9-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_UB]], align 4
1582 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
1583 // CHECK9:       omp.dispatch.end:
1584 // CHECK9-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1585 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
1586 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP26]])
1587 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1588 // CHECK9:       omp.precond.end:
1589 // CHECK9-NEXT:    ret void
1590 //
1591 //
1592 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
1593 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1594 // CHECK9-NEXT:  entry:
1595 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1596 // CHECK9-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
1597 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
1598 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
1599 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
1600 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1601 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1602 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x ptr], align 8
1603 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x ptr], align 8
1604 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x ptr], align 8
1605 // CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
1606 // CHECK9-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1607 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x ptr], align 8
1608 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x ptr], align 8
1609 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x ptr], align 8
1610 // CHECK9-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
1611 // CHECK9-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1612 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1613 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1614 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP0]], align 8
1615 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1616 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
1617 // CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1618 // CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
1619 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1620 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1621 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1622 // CHECK9-NEXT:    store i32 3, ptr [[TMP5]], align 4
1623 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1624 // CHECK9-NEXT:    store i32 1, ptr [[TMP6]], align 4
1625 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1626 // CHECK9-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
1627 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1628 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
1629 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1630 // CHECK9-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP9]], align 8
1631 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1632 // CHECK9-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP10]], align 8
1633 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1634 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
1635 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1636 // CHECK9-NEXT:    store ptr null, ptr [[TMP12]], align 8
1637 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1638 // CHECK9-NEXT:    store i64 10, ptr [[TMP13]], align 8
1639 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1640 // CHECK9-NEXT:    store i64 0, ptr [[TMP14]], align 8
1641 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1642 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
1643 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1644 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
1645 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1646 // CHECK9-NEXT:    store i32 0, ptr [[TMP17]], align 4
1647 // CHECK9-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76.region_id, ptr [[KERNEL_ARGS]])
1648 // CHECK9-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1649 // CHECK9-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1650 // CHECK9:       omp_offload.failed:
1651 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76(ptr [[A]]) #[[ATTR3]]
1652 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1653 // CHECK9:       omp_offload.cont:
1654 // CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1655 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP20]], align 8
1656 // CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1657 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP21]], align 8
1658 // CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
1659 // CHECK9-NEXT:    store ptr null, ptr [[TMP22]], align 8
1660 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1661 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1662 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
1663 // CHECK9-NEXT:    store i32 3, ptr [[TMP25]], align 4
1664 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
1665 // CHECK9-NEXT:    store i32 1, ptr [[TMP26]], align 4
1666 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
1667 // CHECK9-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
1668 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
1669 // CHECK9-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
1670 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
1671 // CHECK9-NEXT:    store ptr @.offload_sizes.7, ptr [[TMP29]], align 8
1672 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
1673 // CHECK9-NEXT:    store ptr @.offload_maptypes.8, ptr [[TMP30]], align 8
1674 // CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
1675 // CHECK9-NEXT:    store ptr null, ptr [[TMP31]], align 8
1676 // CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
1677 // CHECK9-NEXT:    store ptr null, ptr [[TMP32]], align 8
1678 // CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
1679 // CHECK9-NEXT:    store i64 10, ptr [[TMP33]], align 8
1680 // CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
1681 // CHECK9-NEXT:    store i64 0, ptr [[TMP34]], align 8
1682 // CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
1683 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
1684 // CHECK9-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
1685 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
1686 // CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
1687 // CHECK9-NEXT:    store i32 0, ptr [[TMP37]], align 4
1688 // CHECK9-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80.region_id, ptr [[KERNEL_ARGS5]])
1689 // CHECK9-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
1690 // CHECK9-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
1691 // CHECK9:       omp_offload.failed6:
1692 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80(ptr [[A]]) #[[ATTR3]]
1693 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
1694 // CHECK9:       omp_offload.cont7:
1695 // CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
1696 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP40]], align 8
1697 // CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
1698 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP41]], align 8
1699 // CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
1700 // CHECK9-NEXT:    store ptr null, ptr [[TMP42]], align 8
1701 // CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
1702 // CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
1703 // CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
1704 // CHECK9-NEXT:    store i32 3, ptr [[TMP45]], align 4
1705 // CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
1706 // CHECK9-NEXT:    store i32 1, ptr [[TMP46]], align 4
1707 // CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
1708 // CHECK9-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 8
1709 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
1710 // CHECK9-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 8
1711 // CHECK9-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
1712 // CHECK9-NEXT:    store ptr @.offload_sizes.9, ptr [[TMP49]], align 8
1713 // CHECK9-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
1714 // CHECK9-NEXT:    store ptr @.offload_maptypes.10, ptr [[TMP50]], align 8
1715 // CHECK9-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
1716 // CHECK9-NEXT:    store ptr null, ptr [[TMP51]], align 8
1717 // CHECK9-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
1718 // CHECK9-NEXT:    store ptr null, ptr [[TMP52]], align 8
1719 // CHECK9-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
1720 // CHECK9-NEXT:    store i64 10, ptr [[TMP53]], align 8
1721 // CHECK9-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
1722 // CHECK9-NEXT:    store i64 0, ptr [[TMP54]], align 8
1723 // CHECK9-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
1724 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
1725 // CHECK9-NEXT:    [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
1726 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
1727 // CHECK9-NEXT:    [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
1728 // CHECK9-NEXT:    store i32 0, ptr [[TMP57]], align 4
1729 // CHECK9-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, ptr [[KERNEL_ARGS12]])
1730 // CHECK9-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
1731 // CHECK9-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
1732 // CHECK9:       omp_offload.failed13:
1733 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84(ptr [[A]]) #[[ATTR3]]
1734 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
1735 // CHECK9:       omp_offload.cont14:
1736 // CHECK9-NEXT:    ret i32 0
1737 //
1738 //
1739 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76
1740 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1741 // CHECK9-NEXT:  entry:
1742 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1743 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1744 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1745 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76.omp_outlined, ptr [[TMP0]])
1746 // CHECK9-NEXT:    ret void
1747 //
1748 //
1749 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76.omp_outlined
1750 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1751 // CHECK9-NEXT:  entry:
1752 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1753 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1754 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1755 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1756 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1757 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1758 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1759 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1760 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1761 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1762 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1763 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1764 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1765 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1766 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1767 // CHECK9-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
1768 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1769 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1770 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1771 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1772 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1773 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1774 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1775 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1776 // CHECK9:       cond.true:
1777 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1778 // CHECK9:       cond.false:
1779 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1780 // CHECK9-NEXT:    br label [[COND_END]]
1781 // CHECK9:       cond.end:
1782 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1783 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1784 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1785 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1786 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1787 // CHECK9:       omp.inner.for.cond:
1788 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1789 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1790 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1791 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1792 // CHECK9:       omp.inner.for.body:
1793 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1794 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1795 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1796 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1797 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
1798 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
1799 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1800 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1801 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1802 // CHECK9:       omp.body.continue:
1803 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1804 // CHECK9:       omp.inner.for.inc:
1805 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1806 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1807 // CHECK9-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
1808 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1809 // CHECK9:       omp.inner.for.end:
1810 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1811 // CHECK9:       omp.loop.exit:
1812 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1813 // CHECK9-NEXT:    ret void
1814 //
1815 //
1816 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80
1817 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1818 // CHECK9-NEXT:  entry:
1819 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1820 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1821 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1822 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80.omp_outlined, ptr [[TMP0]])
1823 // CHECK9-NEXT:    ret void
1824 //
1825 //
1826 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80.omp_outlined
1827 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1828 // CHECK9-NEXT:  entry:
1829 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1830 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1831 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1832 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1833 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1834 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1835 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1836 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1837 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1838 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1839 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1840 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1841 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1842 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1843 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1844 // CHECK9-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
1845 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1846 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1847 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1848 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1849 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1850 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1851 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1852 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1853 // CHECK9:       cond.true:
1854 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1855 // CHECK9:       cond.false:
1856 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1857 // CHECK9-NEXT:    br label [[COND_END]]
1858 // CHECK9:       cond.end:
1859 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1860 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1861 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1862 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1863 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1864 // CHECK9:       omp.inner.for.cond:
1865 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1866 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1867 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1868 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1869 // CHECK9:       omp.inner.for.body:
1870 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1871 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1872 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1873 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1874 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
1875 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
1876 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1877 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1878 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1879 // CHECK9:       omp.body.continue:
1880 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1881 // CHECK9:       omp.inner.for.inc:
1882 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1883 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1884 // CHECK9-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
1885 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1886 // CHECK9:       omp.inner.for.end:
1887 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1888 // CHECK9:       omp.loop.exit:
1889 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1890 // CHECK9-NEXT:    ret void
1891 //
1892 //
1893 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
1894 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1895 // CHECK9-NEXT:  entry:
1896 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1897 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1898 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1899 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined, ptr [[TMP0]])
1900 // CHECK9-NEXT:    ret void
1901 //
1902 //
1903 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined
1904 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1905 // CHECK9-NEXT:  entry:
1906 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1907 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1908 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1909 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1910 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1911 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1912 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1913 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1914 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1915 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1916 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1917 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1918 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1919 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1920 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1921 // CHECK9-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
1922 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1923 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1924 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1925 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1926 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 10)
1927 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1928 // CHECK9:       omp.dispatch.cond:
1929 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1930 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1931 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1932 // CHECK9:       cond.true:
1933 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1934 // CHECK9:       cond.false:
1935 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1936 // CHECK9-NEXT:    br label [[COND_END]]
1937 // CHECK9:       cond.end:
1938 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1939 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1940 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1941 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1942 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1943 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1944 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1945 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1946 // CHECK9:       omp.dispatch.body:
1947 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1948 // CHECK9:       omp.inner.for.cond:
1949 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14:![0-9]+]]
1950 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
1951 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1952 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1953 // CHECK9:       omp.inner.for.body:
1954 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
1955 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1956 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1957 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP14]]
1958 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP14]]
1959 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1960 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1961 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP14]]
1962 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1963 // CHECK9:       omp.body.continue:
1964 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1965 // CHECK9:       omp.inner.for.inc:
1966 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
1967 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
1968 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
1969 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
1970 // CHECK9:       omp.inner.for.end:
1971 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1972 // CHECK9:       omp.dispatch.inc:
1973 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1974 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1975 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1976 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
1977 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1978 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1979 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
1980 // CHECK9-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
1981 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
1982 // CHECK9:       omp.dispatch.end:
1983 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1984 // CHECK9-NEXT:    ret void
1985 //
1986 //
1987 // CHECK11-LABEL: define {{[^@]+}}@main
1988 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1989 // CHECK11-NEXT:  entry:
1990 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1991 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1992 // CHECK11-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 4
1993 // CHECK11-NEXT:    [[N:%.*]] = alloca i32, align 4
1994 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
1995 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1996 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
1997 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
1998 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
1999 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
2000 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
2001 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2002 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2003 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2004 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2005 // CHECK11-NEXT:    [[N_CASTED3:%.*]] = alloca i32, align 4
2006 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x ptr], align 4
2007 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x ptr], align 4
2008 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x ptr], align 4
2009 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
2010 // CHECK11-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
2011 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
2012 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
2013 // CHECK11-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2014 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
2015 // CHECK11-NEXT:    [[N_CASTED19:%.*]] = alloca i32, align 4
2016 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
2017 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [4 x ptr], align 4
2018 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [4 x ptr], align 4
2019 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [4 x ptr], align 4
2020 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES23:%.*]] = alloca [4 x i64], align 4
2021 // CHECK11-NEXT:    [[_TMP24:%.*]] = alloca i32, align 4
2022 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
2023 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
2024 // CHECK11-NEXT:    [[KERNEL_ARGS31:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2025 // CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
2026 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
2027 // CHECK11-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 4
2028 // CHECK11-NEXT:    store i32 100, ptr [[N]], align 4
2029 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
2030 // CHECK11-NEXT:    [[TMP1:%.*]] = call ptr @llvm.stacksave.p0()
2031 // CHECK11-NEXT:    store ptr [[TMP1]], ptr [[SAVED_STACK]], align 4
2032 // CHECK11-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
2033 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[__VLA_EXPR0]], align 4
2034 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N]], align 4
2035 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
2036 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
2037 // CHECK11-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
2038 // CHECK11-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
2039 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 24, i1 false)
2040 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2041 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP6]], align 4
2042 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2043 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP7]], align 4
2044 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2045 // CHECK11-NEXT:    store ptr null, ptr [[TMP8]], align 4
2046 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2047 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP9]], align 4
2048 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2049 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP10]], align 4
2050 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2051 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
2052 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2053 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 4
2054 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2055 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 4
2056 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2057 // CHECK11-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 4
2058 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2059 // CHECK11-NEXT:    store ptr null, ptr [[TMP15]], align 4
2060 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2061 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2062 // CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2063 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
2064 // CHECK11-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
2065 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2066 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
2067 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2068 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2069 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2070 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2071 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
2072 // CHECK11-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
2073 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2074 // CHECK11-NEXT:    store i32 3, ptr [[TMP23]], align 4
2075 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2076 // CHECK11-NEXT:    store i32 3, ptr [[TMP24]], align 4
2077 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2078 // CHECK11-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 4
2079 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2080 // CHECK11-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 4
2081 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2082 // CHECK11-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 4
2083 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2084 // CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 4
2085 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2086 // CHECK11-NEXT:    store ptr null, ptr [[TMP29]], align 4
2087 // CHECK11-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2088 // CHECK11-NEXT:    store ptr null, ptr [[TMP30]], align 4
2089 // CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2090 // CHECK11-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
2091 // CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2092 // CHECK11-NEXT:    store i64 0, ptr [[TMP32]], align 8
2093 // CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2094 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
2095 // CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2096 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
2097 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2098 // CHECK11-NEXT:    store i32 0, ptr [[TMP35]], align 4
2099 // CHECK11-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, ptr [[KERNEL_ARGS]])
2100 // CHECK11-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2101 // CHECK11-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2102 // CHECK11:       omp_offload.failed:
2103 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94(i32 [[TMP3]], i32 [[TMP0]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
2104 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2105 // CHECK11:       omp_offload.cont:
2106 // CHECK11-NEXT:    [[TMP38:%.*]] = load i32, ptr [[N]], align 4
2107 // CHECK11-NEXT:    store i32 [[TMP38]], ptr [[N_CASTED3]], align 4
2108 // CHECK11-NEXT:    [[TMP39:%.*]] = load i32, ptr [[N_CASTED3]], align 4
2109 // CHECK11-NEXT:    [[TMP40:%.*]] = mul nuw i32 [[TMP0]], 4
2110 // CHECK11-NEXT:    [[TMP41:%.*]] = sext i32 [[TMP40]] to i64
2111 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES7]], ptr align 4 @.offload_sizes.1, i32 24, i1 false)
2112 // CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
2113 // CHECK11-NEXT:    store i32 [[TMP39]], ptr [[TMP42]], align 4
2114 // CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
2115 // CHECK11-NEXT:    store i32 [[TMP39]], ptr [[TMP43]], align 4
2116 // CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
2117 // CHECK11-NEXT:    store ptr null, ptr [[TMP44]], align 4
2118 // CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
2119 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP45]], align 4
2120 // CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
2121 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP46]], align 4
2122 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
2123 // CHECK11-NEXT:    store ptr null, ptr [[TMP47]], align 4
2124 // CHECK11-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
2125 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP48]], align 4
2126 // CHECK11-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
2127 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP49]], align 4
2128 // CHECK11-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
2129 // CHECK11-NEXT:    store i64 [[TMP41]], ptr [[TMP50]], align 4
2130 // CHECK11-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
2131 // CHECK11-NEXT:    store ptr null, ptr [[TMP51]], align 4
2132 // CHECK11-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
2133 // CHECK11-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
2134 // CHECK11-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
2135 // CHECK11-NEXT:    [[TMP55:%.*]] = load i32, ptr [[N]], align 4
2136 // CHECK11-NEXT:    store i32 [[TMP55]], ptr [[DOTCAPTURE_EXPR_9]], align 4
2137 // CHECK11-NEXT:    [[TMP56:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_9]], align 4
2138 // CHECK11-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP56]], 0
2139 // CHECK11-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
2140 // CHECK11-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
2141 // CHECK11-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_10]], align 4
2142 // CHECK11-NEXT:    [[TMP57:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_10]], align 4
2143 // CHECK11-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP57]], 1
2144 // CHECK11-NEXT:    [[TMP58:%.*]] = zext i32 [[ADD14]] to i64
2145 // CHECK11-NEXT:    [[TMP59:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
2146 // CHECK11-NEXT:    store i32 3, ptr [[TMP59]], align 4
2147 // CHECK11-NEXT:    [[TMP60:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
2148 // CHECK11-NEXT:    store i32 3, ptr [[TMP60]], align 4
2149 // CHECK11-NEXT:    [[TMP61:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
2150 // CHECK11-NEXT:    store ptr [[TMP52]], ptr [[TMP61]], align 4
2151 // CHECK11-NEXT:    [[TMP62:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
2152 // CHECK11-NEXT:    store ptr [[TMP53]], ptr [[TMP62]], align 4
2153 // CHECK11-NEXT:    [[TMP63:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
2154 // CHECK11-NEXT:    store ptr [[TMP54]], ptr [[TMP63]], align 4
2155 // CHECK11-NEXT:    [[TMP64:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
2156 // CHECK11-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP64]], align 4
2157 // CHECK11-NEXT:    [[TMP65:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
2158 // CHECK11-NEXT:    store ptr null, ptr [[TMP65]], align 4
2159 // CHECK11-NEXT:    [[TMP66:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
2160 // CHECK11-NEXT:    store ptr null, ptr [[TMP66]], align 4
2161 // CHECK11-NEXT:    [[TMP67:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
2162 // CHECK11-NEXT:    store i64 [[TMP58]], ptr [[TMP67]], align 8
2163 // CHECK11-NEXT:    [[TMP68:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
2164 // CHECK11-NEXT:    store i64 0, ptr [[TMP68]], align 8
2165 // CHECK11-NEXT:    [[TMP69:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
2166 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
2167 // CHECK11-NEXT:    [[TMP70:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
2168 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP70]], align 4
2169 // CHECK11-NEXT:    [[TMP71:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
2170 // CHECK11-NEXT:    store i32 0, ptr [[TMP71]], align 4
2171 // CHECK11-NEXT:    [[TMP72:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98.region_id, ptr [[KERNEL_ARGS15]])
2172 // CHECK11-NEXT:    [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0
2173 // CHECK11-NEXT:    br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
2174 // CHECK11:       omp_offload.failed16:
2175 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98(i32 [[TMP39]], i32 [[TMP0]], ptr [[VLA]]) #[[ATTR3]]
2176 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
2177 // CHECK11:       omp_offload.cont17:
2178 // CHECK11-NEXT:    [[TMP74:%.*]] = load i32, ptr [[N]], align 4
2179 // CHECK11-NEXT:    store i32 [[TMP74]], ptr [[DOTCAPTURE_EXPR_18]], align 4
2180 // CHECK11-NEXT:    [[TMP75:%.*]] = load i32, ptr [[N]], align 4
2181 // CHECK11-NEXT:    store i32 [[TMP75]], ptr [[N_CASTED19]], align 4
2182 // CHECK11-NEXT:    [[TMP76:%.*]] = load i32, ptr [[N_CASTED19]], align 4
2183 // CHECK11-NEXT:    [[TMP77:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_18]], align 4
2184 // CHECK11-NEXT:    store i32 [[TMP77]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
2185 // CHECK11-NEXT:    [[TMP78:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
2186 // CHECK11-NEXT:    [[TMP79:%.*]] = mul nuw i32 [[TMP0]], 4
2187 // CHECK11-NEXT:    [[TMP80:%.*]] = sext i32 [[TMP79]] to i64
2188 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES23]], ptr align 4 @.offload_sizes.3, i32 32, i1 false)
2189 // CHECK11-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
2190 // CHECK11-NEXT:    store i32 [[TMP76]], ptr [[TMP81]], align 4
2191 // CHECK11-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
2192 // CHECK11-NEXT:    store i32 [[TMP76]], ptr [[TMP82]], align 4
2193 // CHECK11-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
2194 // CHECK11-NEXT:    store ptr null, ptr [[TMP83]], align 4
2195 // CHECK11-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
2196 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP84]], align 4
2197 // CHECK11-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
2198 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP85]], align 4
2199 // CHECK11-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
2200 // CHECK11-NEXT:    store ptr null, ptr [[TMP86]], align 4
2201 // CHECK11-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
2202 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP87]], align 4
2203 // CHECK11-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
2204 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP88]], align 4
2205 // CHECK11-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [4 x i64], ptr [[DOTOFFLOAD_SIZES23]], i32 0, i32 2
2206 // CHECK11-NEXT:    store i64 [[TMP80]], ptr [[TMP89]], align 4
2207 // CHECK11-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 2
2208 // CHECK11-NEXT:    store ptr null, ptr [[TMP90]], align 4
2209 // CHECK11-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
2210 // CHECK11-NEXT:    store i32 [[TMP78]], ptr [[TMP91]], align 4
2211 // CHECK11-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
2212 // CHECK11-NEXT:    store i32 [[TMP78]], ptr [[TMP92]], align 4
2213 // CHECK11-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 3
2214 // CHECK11-NEXT:    store ptr null, ptr [[TMP93]], align 4
2215 // CHECK11-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
2216 // CHECK11-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
2217 // CHECK11-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [4 x i64], ptr [[DOTOFFLOAD_SIZES23]], i32 0, i32 0
2218 // CHECK11-NEXT:    [[TMP97:%.*]] = load i32, ptr [[N]], align 4
2219 // CHECK11-NEXT:    store i32 [[TMP97]], ptr [[DOTCAPTURE_EXPR_25]], align 4
2220 // CHECK11-NEXT:    [[TMP98:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_25]], align 4
2221 // CHECK11-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP98]], 0
2222 // CHECK11-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
2223 // CHECK11-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[DIV28]], 1
2224 // CHECK11-NEXT:    store i32 [[SUB29]], ptr [[DOTCAPTURE_EXPR_26]], align 4
2225 // CHECK11-NEXT:    [[TMP99:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_26]], align 4
2226 // CHECK11-NEXT:    [[ADD30:%.*]] = add nsw i32 [[TMP99]], 1
2227 // CHECK11-NEXT:    [[TMP100:%.*]] = zext i32 [[ADD30]] to i64
2228 // CHECK11-NEXT:    [[TMP101:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 0
2229 // CHECK11-NEXT:    store i32 3, ptr [[TMP101]], align 4
2230 // CHECK11-NEXT:    [[TMP102:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 1
2231 // CHECK11-NEXT:    store i32 4, ptr [[TMP102]], align 4
2232 // CHECK11-NEXT:    [[TMP103:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 2
2233 // CHECK11-NEXT:    store ptr [[TMP94]], ptr [[TMP103]], align 4
2234 // CHECK11-NEXT:    [[TMP104:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 3
2235 // CHECK11-NEXT:    store ptr [[TMP95]], ptr [[TMP104]], align 4
2236 // CHECK11-NEXT:    [[TMP105:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 4
2237 // CHECK11-NEXT:    store ptr [[TMP96]], ptr [[TMP105]], align 4
2238 // CHECK11-NEXT:    [[TMP106:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 5
2239 // CHECK11-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP106]], align 4
2240 // CHECK11-NEXT:    [[TMP107:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 6
2241 // CHECK11-NEXT:    store ptr null, ptr [[TMP107]], align 4
2242 // CHECK11-NEXT:    [[TMP108:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 7
2243 // CHECK11-NEXT:    store ptr null, ptr [[TMP108]], align 4
2244 // CHECK11-NEXT:    [[TMP109:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 8
2245 // CHECK11-NEXT:    store i64 [[TMP100]], ptr [[TMP109]], align 8
2246 // CHECK11-NEXT:    [[TMP110:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 9
2247 // CHECK11-NEXT:    store i64 0, ptr [[TMP110]], align 8
2248 // CHECK11-NEXT:    [[TMP111:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 10
2249 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP111]], align 4
2250 // CHECK11-NEXT:    [[TMP112:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 11
2251 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP112]], align 4
2252 // CHECK11-NEXT:    [[TMP113:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS31]], i32 0, i32 12
2253 // CHECK11-NEXT:    store i32 0, ptr [[TMP113]], align 4
2254 // CHECK11-NEXT:    [[TMP114:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.region_id, ptr [[KERNEL_ARGS31]])
2255 // CHECK11-NEXT:    [[TMP115:%.*]] = icmp ne i32 [[TMP114]], 0
2256 // CHECK11-NEXT:    br i1 [[TMP115]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
2257 // CHECK11:       omp_offload.failed32:
2258 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102(i32 [[TMP76]], i32 [[TMP0]], ptr [[VLA]], i32 [[TMP78]]) #[[ATTR3]]
2259 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT33]]
2260 // CHECK11:       omp_offload.cont33:
2261 // CHECK11-NEXT:    [[TMP116:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
2262 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP116]])
2263 // CHECK11-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
2264 // CHECK11-NEXT:    [[TMP117:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
2265 // CHECK11-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP117]])
2266 // CHECK11-NEXT:    [[TMP118:%.*]] = load i32, ptr [[RETVAL]], align 4
2267 // CHECK11-NEXT:    ret i32 [[TMP118]]
2268 //
2269 //
2270 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
2271 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
2272 // CHECK11-NEXT:  entry:
2273 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2274 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2275 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2276 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
2277 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2278 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2279 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2280 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2281 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2282 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
2283 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
2284 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
2285 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.omp_outlined, i32 [[TMP3]], i32 [[TMP0]], ptr [[TMP1]])
2286 // CHECK11-NEXT:    ret void
2287 //
2288 //
2289 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.omp_outlined
2290 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2291 // CHECK11-NEXT:  entry:
2292 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2293 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2294 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2295 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2296 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2297 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2298 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2299 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2300 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2301 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2302 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2303 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2304 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2305 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2306 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
2307 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2308 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2309 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2310 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2311 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2312 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2313 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2314 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
2315 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
2316 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2317 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
2318 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2319 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2320 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2321 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
2322 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2323 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
2324 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2325 // CHECK11:       omp.precond.then:
2326 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2327 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2328 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
2329 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2330 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2331 // CHECK11-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2332 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
2333 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2334 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2335 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2336 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
2337 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2338 // CHECK11:       cond.true:
2339 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2340 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2341 // CHECK11:       cond.false:
2342 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2343 // CHECK11-NEXT:    br label [[COND_END]]
2344 // CHECK11:       cond.end:
2345 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
2346 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2347 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2348 // CHECK11-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
2349 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2350 // CHECK11:       omp.inner.for.cond:
2351 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2352 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2353 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2354 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2355 // CHECK11:       omp.inner.for.body:
2356 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2357 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
2358 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2359 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
2360 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
2361 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 [[TMP16]]
2362 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2363 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2364 // CHECK11:       omp.body.continue:
2365 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2366 // CHECK11:       omp.inner.for.inc:
2367 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2368 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
2369 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
2370 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2371 // CHECK11:       omp.inner.for.end:
2372 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2373 // CHECK11:       omp.loop.exit:
2374 // CHECK11-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2375 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
2376 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
2377 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
2378 // CHECK11:       omp.precond.end:
2379 // CHECK11-NEXT:    ret void
2380 //
2381 //
2382 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98
2383 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2384 // CHECK11-NEXT:  entry:
2385 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2386 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2387 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2388 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
2389 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2390 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2391 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2392 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2393 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2394 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
2395 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
2396 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
2397 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98.omp_outlined, i32 [[TMP3]], i32 [[TMP0]], ptr [[TMP1]])
2398 // CHECK11-NEXT:    ret void
2399 //
2400 //
2401 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l98.omp_outlined
2402 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2403 // CHECK11-NEXT:  entry:
2404 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2405 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2406 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2407 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2408 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2409 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2410 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2411 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2412 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2413 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2414 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2415 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2416 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2417 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2418 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
2419 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2420 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2421 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2422 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2423 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2424 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2425 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2426 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
2427 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
2428 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2429 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
2430 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2431 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2432 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2433 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
2434 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2435 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
2436 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2437 // CHECK11:       omp.precond.then:
2438 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2439 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2440 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
2441 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2442 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2443 // CHECK11-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2444 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
2445 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2446 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2447 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2448 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
2449 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2450 // CHECK11:       cond.true:
2451 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2452 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2453 // CHECK11:       cond.false:
2454 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2455 // CHECK11-NEXT:    br label [[COND_END]]
2456 // CHECK11:       cond.end:
2457 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
2458 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2459 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2460 // CHECK11-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
2461 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2462 // CHECK11:       omp.inner.for.cond:
2463 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2464 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2465 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2466 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2467 // CHECK11:       omp.inner.for.body:
2468 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2469 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
2470 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2471 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
2472 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I3]], align 4
2473 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 [[TMP16]]
2474 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2475 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2476 // CHECK11:       omp.body.continue:
2477 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2478 // CHECK11:       omp.inner.for.inc:
2479 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2480 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
2481 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
2482 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2483 // CHECK11:       omp.inner.for.end:
2484 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2485 // CHECK11:       omp.loop.exit:
2486 // CHECK11-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2487 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
2488 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
2489 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
2490 // CHECK11:       omp.precond.end:
2491 // CHECK11-NEXT:    ret void
2492 //
2493 //
2494 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102
2495 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2496 // CHECK11-NEXT:  entry:
2497 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2498 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2499 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2500 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2501 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
2502 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
2503 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2504 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2505 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2506 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
2507 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2508 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2509 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
2510 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
2511 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
2512 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
2513 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
2514 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
2515 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined, i32 [[TMP3]], i32 [[TMP0]], ptr [[TMP1]], i32 [[TMP5]])
2516 // CHECK11-NEXT:    ret void
2517 //
2518 //
2519 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l102.omp_outlined
2520 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2521 // CHECK11-NEXT:  entry:
2522 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2523 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2524 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2525 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2526 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2527 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2528 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2529 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2530 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2531 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2532 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2533 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2534 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2535 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2536 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2537 // CHECK11-NEXT:    [[I4:%.*]] = alloca i32, align 4
2538 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2539 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2540 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2541 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2542 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2543 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
2544 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2545 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2546 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
2547 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2548 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2549 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
2550 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2551 // CHECK11-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
2552 // CHECK11-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
2553 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
2554 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2555 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
2556 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2557 // CHECK11:       omp.precond.then:
2558 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2559 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
2560 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
2561 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2562 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2563 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
2564 // CHECK11-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2565 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2566 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP8]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[TMP6]])
2567 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2568 // CHECK11:       omp.dispatch.cond:
2569 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2570 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
2571 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2572 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2573 // CHECK11:       cond.true:
2574 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
2575 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2576 // CHECK11:       cond.false:
2577 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2578 // CHECK11-NEXT:    br label [[COND_END]]
2579 // CHECK11:       cond.end:
2580 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2581 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2582 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2583 // CHECK11-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
2584 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2585 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2586 // CHECK11-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2587 // CHECK11-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2588 // CHECK11:       omp.dispatch.body:
2589 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2590 // CHECK11:       omp.inner.for.cond:
2591 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12:![0-9]+]]
2592 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP12]]
2593 // CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2594 // CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2595 // CHECK11:       omp.inner.for.body:
2596 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
2597 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2598 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2599 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP12]]
2600 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP12]]
2601 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 [[TMP19]]
2602 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP12]]
2603 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2604 // CHECK11:       omp.body.continue:
2605 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2606 // CHECK11:       omp.inner.for.inc:
2607 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
2608 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP20]], 1
2609 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
2610 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
2611 // CHECK11:       omp.inner.for.end:
2612 // CHECK11-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2613 // CHECK11:       omp.dispatch.inc:
2614 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2615 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2616 // CHECK11-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
2617 // CHECK11-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_LB]], align 4
2618 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2619 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2620 // CHECK11-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
2621 // CHECK11-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_UB]], align 4
2622 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND]]
2623 // CHECK11:       omp.dispatch.end:
2624 // CHECK11-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2625 // CHECK11-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
2626 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP26]])
2627 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
2628 // CHECK11:       omp.precond.end:
2629 // CHECK11-NEXT:    ret void
2630 //
2631 //
2632 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
2633 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
2634 // CHECK11-NEXT:  entry:
2635 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
2636 // CHECK11-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
2637 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
2638 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
2639 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
2640 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2641 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2642 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x ptr], align 4
2643 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x ptr], align 4
2644 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x ptr], align 4
2645 // CHECK11-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
2646 // CHECK11-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2647 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x ptr], align 4
2648 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x ptr], align 4
2649 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x ptr], align 4
2650 // CHECK11-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
2651 // CHECK11-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2652 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
2653 // CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2654 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP0]], align 4
2655 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2656 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
2657 // CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2658 // CHECK11-NEXT:    store ptr null, ptr [[TMP2]], align 4
2659 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2660 // CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2661 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2662 // CHECK11-NEXT:    store i32 3, ptr [[TMP5]], align 4
2663 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2664 // CHECK11-NEXT:    store i32 1, ptr [[TMP6]], align 4
2665 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2666 // CHECK11-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
2667 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2668 // CHECK11-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
2669 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2670 // CHECK11-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP9]], align 4
2671 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2672 // CHECK11-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP10]], align 4
2673 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2674 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
2675 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2676 // CHECK11-NEXT:    store ptr null, ptr [[TMP12]], align 4
2677 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2678 // CHECK11-NEXT:    store i64 10, ptr [[TMP13]], align 8
2679 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2680 // CHECK11-NEXT:    store i64 0, ptr [[TMP14]], align 8
2681 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2682 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
2683 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2684 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
2685 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2686 // CHECK11-NEXT:    store i32 0, ptr [[TMP17]], align 4
2687 // CHECK11-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76.region_id, ptr [[KERNEL_ARGS]])
2688 // CHECK11-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2689 // CHECK11-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2690 // CHECK11:       omp_offload.failed:
2691 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76(ptr [[A]]) #[[ATTR3]]
2692 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2693 // CHECK11:       omp_offload.cont:
2694 // CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2695 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP20]], align 4
2696 // CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2697 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP21]], align 4
2698 // CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
2699 // CHECK11-NEXT:    store ptr null, ptr [[TMP22]], align 4
2700 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2701 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2702 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
2703 // CHECK11-NEXT:    store i32 3, ptr [[TMP25]], align 4
2704 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
2705 // CHECK11-NEXT:    store i32 1, ptr [[TMP26]], align 4
2706 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
2707 // CHECK11-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 4
2708 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
2709 // CHECK11-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 4
2710 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
2711 // CHECK11-NEXT:    store ptr @.offload_sizes.7, ptr [[TMP29]], align 4
2712 // CHECK11-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
2713 // CHECK11-NEXT:    store ptr @.offload_maptypes.8, ptr [[TMP30]], align 4
2714 // CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
2715 // CHECK11-NEXT:    store ptr null, ptr [[TMP31]], align 4
2716 // CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
2717 // CHECK11-NEXT:    store ptr null, ptr [[TMP32]], align 4
2718 // CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
2719 // CHECK11-NEXT:    store i64 10, ptr [[TMP33]], align 8
2720 // CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
2721 // CHECK11-NEXT:    store i64 0, ptr [[TMP34]], align 8
2722 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
2723 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
2724 // CHECK11-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
2725 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
2726 // CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
2727 // CHECK11-NEXT:    store i32 0, ptr [[TMP37]], align 4
2728 // CHECK11-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80.region_id, ptr [[KERNEL_ARGS5]])
2729 // CHECK11-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
2730 // CHECK11-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
2731 // CHECK11:       omp_offload.failed6:
2732 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80(ptr [[A]]) #[[ATTR3]]
2733 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
2734 // CHECK11:       omp_offload.cont7:
2735 // CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2736 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP40]], align 4
2737 // CHECK11-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2738 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP41]], align 4
2739 // CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
2740 // CHECK11-NEXT:    store ptr null, ptr [[TMP42]], align 4
2741 // CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2742 // CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2743 // CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
2744 // CHECK11-NEXT:    store i32 3, ptr [[TMP45]], align 4
2745 // CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
2746 // CHECK11-NEXT:    store i32 1, ptr [[TMP46]], align 4
2747 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
2748 // CHECK11-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 4
2749 // CHECK11-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
2750 // CHECK11-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 4
2751 // CHECK11-NEXT:    [[TMP49:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
2752 // CHECK11-NEXT:    store ptr @.offload_sizes.9, ptr [[TMP49]], align 4
2753 // CHECK11-NEXT:    [[TMP50:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
2754 // CHECK11-NEXT:    store ptr @.offload_maptypes.10, ptr [[TMP50]], align 4
2755 // CHECK11-NEXT:    [[TMP51:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
2756 // CHECK11-NEXT:    store ptr null, ptr [[TMP51]], align 4
2757 // CHECK11-NEXT:    [[TMP52:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
2758 // CHECK11-NEXT:    store ptr null, ptr [[TMP52]], align 4
2759 // CHECK11-NEXT:    [[TMP53:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
2760 // CHECK11-NEXT:    store i64 10, ptr [[TMP53]], align 8
2761 // CHECK11-NEXT:    [[TMP54:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
2762 // CHECK11-NEXT:    store i64 0, ptr [[TMP54]], align 8
2763 // CHECK11-NEXT:    [[TMP55:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
2764 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
2765 // CHECK11-NEXT:    [[TMP56:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
2766 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
2767 // CHECK11-NEXT:    [[TMP57:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
2768 // CHECK11-NEXT:    store i32 0, ptr [[TMP57]], align 4
2769 // CHECK11-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, ptr [[KERNEL_ARGS12]])
2770 // CHECK11-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
2771 // CHECK11-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
2772 // CHECK11:       omp_offload.failed13:
2773 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84(ptr [[A]]) #[[ATTR3]]
2774 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
2775 // CHECK11:       omp_offload.cont14:
2776 // CHECK11-NEXT:    ret i32 0
2777 //
2778 //
2779 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76
2780 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2781 // CHECK11-NEXT:  entry:
2782 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2783 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2784 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2785 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76.omp_outlined, ptr [[TMP0]])
2786 // CHECK11-NEXT:    ret void
2787 //
2788 //
2789 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l76.omp_outlined
2790 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2791 // CHECK11-NEXT:  entry:
2792 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2793 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2794 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2795 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2796 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2797 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2798 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2799 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2800 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2801 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2802 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2803 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2804 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2805 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2806 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2807 // CHECK11-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
2808 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2809 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2810 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2811 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2812 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2813 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2814 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2815 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2816 // CHECK11:       cond.true:
2817 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2818 // CHECK11:       cond.false:
2819 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2820 // CHECK11-NEXT:    br label [[COND_END]]
2821 // CHECK11:       cond.end:
2822 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2823 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2824 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2825 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2826 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2827 // CHECK11:       omp.inner.for.cond:
2828 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2829 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2830 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2831 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2832 // CHECK11:       omp.inner.for.body:
2833 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2834 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2835 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2836 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2837 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
2838 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP9]]
2839 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2840 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2841 // CHECK11:       omp.body.continue:
2842 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2843 // CHECK11:       omp.inner.for.inc:
2844 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2845 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2846 // CHECK11-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
2847 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2848 // CHECK11:       omp.inner.for.end:
2849 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2850 // CHECK11:       omp.loop.exit:
2851 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
2852 // CHECK11-NEXT:    ret void
2853 //
2854 //
2855 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80
2856 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2857 // CHECK11-NEXT:  entry:
2858 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2859 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2860 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2861 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80.omp_outlined, ptr [[TMP0]])
2862 // CHECK11-NEXT:    ret void
2863 //
2864 //
2865 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l80.omp_outlined
2866 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2867 // CHECK11-NEXT:  entry:
2868 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2869 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2870 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2871 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2872 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2873 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2874 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2875 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2876 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2877 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2878 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2879 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2880 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2881 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2882 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2883 // CHECK11-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
2884 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2885 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2886 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2887 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2888 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2889 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2890 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2891 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2892 // CHECK11:       cond.true:
2893 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2894 // CHECK11:       cond.false:
2895 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2896 // CHECK11-NEXT:    br label [[COND_END]]
2897 // CHECK11:       cond.end:
2898 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2899 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2900 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2901 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2902 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2903 // CHECK11:       omp.inner.for.cond:
2904 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2905 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2906 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2907 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2908 // CHECK11:       omp.inner.for.body:
2909 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2910 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2911 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2912 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2913 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
2914 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP9]]
2915 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2916 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2917 // CHECK11:       omp.body.continue:
2918 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2919 // CHECK11:       omp.inner.for.inc:
2920 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2921 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2922 // CHECK11-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
2923 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2924 // CHECK11:       omp.inner.for.end:
2925 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2926 // CHECK11:       omp.loop.exit:
2927 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
2928 // CHECK11-NEXT:    ret void
2929 //
2930 //
2931 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
2932 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2933 // CHECK11-NEXT:  entry:
2934 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2935 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2936 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2937 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined, ptr [[TMP0]])
2938 // CHECK11-NEXT:    ret void
2939 //
2940 //
2941 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined
2942 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2943 // CHECK11-NEXT:  entry:
2944 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2945 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2946 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2947 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2948 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2949 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2950 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2951 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2952 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2953 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2954 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2955 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2956 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2957 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2958 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2959 // CHECK11-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
2960 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2961 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2962 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2963 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2964 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 10)
2965 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2966 // CHECK11:       omp.dispatch.cond:
2967 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2968 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2969 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2970 // CHECK11:       cond.true:
2971 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2972 // CHECK11:       cond.false:
2973 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2974 // CHECK11-NEXT:    br label [[COND_END]]
2975 // CHECK11:       cond.end:
2976 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2977 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2978 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2979 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2980 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2981 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2982 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2983 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2984 // CHECK11:       omp.dispatch.body:
2985 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2986 // CHECK11:       omp.inner.for.cond:
2987 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
2988 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
2989 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2990 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2991 // CHECK11:       omp.inner.for.body:
2992 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
2993 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2994 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2995 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
2996 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
2997 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP11]]
2998 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP15]]
2999 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3000 // CHECK11:       omp.body.continue:
3001 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3002 // CHECK11:       omp.inner.for.inc:
3003 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
3004 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
3005 // CHECK11-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
3006 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
3007 // CHECK11:       omp.inner.for.end:
3008 // CHECK11-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3009 // CHECK11:       omp.dispatch.inc:
3010 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3011 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
3012 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
3013 // CHECK11-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
3014 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3015 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
3016 // CHECK11-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
3017 // CHECK11-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
3018 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND]]
3019 // CHECK11:       omp.dispatch.end:
3020 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
3021 // CHECK11-NEXT:    ret void
3022 //