clang/test/OpenMP/target_teams_distribute_reduction_codegen.cpp

   1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
   2 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
   3 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
   4 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
   5 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
   6 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
   7 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
   8
   9 // RUN: %clang_cc1 -DCHECK -verify -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}}"
  10 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  11 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  12 // RUN: %clang_cc1 -DCHECK -verify -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}}"
  13 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  14 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  15
  16 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
  17 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  18 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++  -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
  19
  20 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  21 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  22 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++  -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  23
  24 // expected-no-diagnostics
  25 #ifndef HEADER
  26 #define HEADER
  27
  28 template <typename T>
  29 T tmain() {
  30   T t_var = T();
  31   T vec[] = {1, 2};
  32 #pragma omp target teams distribute reduction(+: t_var)
  33   for (int i = 0; i < 2; ++i) {
  34     t_var += (T) i;
  35   }
  36
  37 #pragma omp target teams distribute reduction(-: t_var)
  38   for (int i = 0; i < 2; ++i) {
  39     t_var -= vec[i];
  40   }
  41
  42 t_var = T(1);
  43 #pragma omp target teams distribute reduction(*: t_var)
  44   for (int i = 0; i < 2; ++i) {
  45     t_var = t_var * vec[i];
  46   }
  47
  48   bool and_var = true;
  49 #pragma omp target teams distribute reduction(&&: and_var)
  50   for (int i = 0; i < 2; ++i) {
  51     and_var = and_var && (vec[i]%2 == 0);
  52   }
  53
  54   bool or_var = false;
  55 #pragma omp target teams distribute reduction(||: or_var)
  56   for (int i = 0; i < 2; ++i) {
  57     or_var = or_var || (vec[i]%2 == 0);
  58   }
  59
  60   unsigned int bit_var = 1;
  61 #pragma omp target teams distribute reduction(&: bit_var)
  62   for (int i = 0; i < 2; ++i) {
  63     bit_var = (bit_var & vec[i]);
  64   }
  65
  66 #pragma omp target teams distribute reduction(|: bit_var)
  67   for (int i = 0; i < 2; ++i) {
  68     bit_var = (bit_var | vec[i]);
  69   }
  70
  71 #pragma omp target teams distribute reduction(^: bit_var)
  72   for (int i = 0; i < 2; ++i) {
  73     bit_var = (bit_var ^ vec[i]);
  74   }
  75
  76   t_var = T(0);
  77 #pragma omp target teams distribute reduction(max: t_var)
  78   for (int i = 0; i < 2; ++i) {
  79     t_var = t_var >= vec[i] ? t_var : vec[i];
  80   }
  81
  82   t_var = T(10);
  83 #pragma omp target teams distribute reduction(min: t_var)
  84   for (int i = 0; i < 2; ++i) {
  85     t_var = t_var <= vec[i] ? t_var : vec[i];
  86   }
  87
  88   return T();
  89 }
  90
  91 int main() {
  92   static int sivar;
  93 #ifdef LAMBDA
  94
  95   [&]() {
  96 #pragma omp target teams distribute reduction(+: sivar)
  97   for (int i = 0; i < 2; ++i) {
  98
  99     // Skip global and bound tid vars
 100
 101     sivar += i;
 102
 103     [&]() {
 104
 105       sivar += 4;
 106
 107     }();
 108   }
 109   }();
 110
 111   [&]() {
 112 #pragma omp target teams distribute reduction(-: sivar)
 113   for (int i = 0; i < 2; ++i) {
 114     sivar -= i;
 115     [&]() {
 116       sivar += 4;
 117     }();
 118   }
 119   }();
 120
 121   [&]() {
 122 #pragma omp target teams distribute reduction(*: sivar)
 123   for (int i = 0; i < 2; ++i) {
 124     sivar *= i;
 125     [&]() {
 126       sivar += 4;
 127     }();
 128   }
 129   }();
 130
 131   [&]() {
 132   bool and_var = true;
 133 #pragma omp target teams distribute reduction(&&: and_var)
 134   for (int i = 0; i < 2; ++i) {
 135     and_var = and_var && (i == 0);
 136     [&]() {
 137       sivar += 4;
 138     }();
 139   }
 140   }();
 141
 142   [&]() {
 143   bool or_var = true;
 144 #pragma omp target teams distribute reduction(||: or_var)
 145   for (int i = 0; i < 2; ++i) {
 146     or_var = or_var || (i == 0);
 147     [&]() {
 148       sivar += 4;
 149     }();
 150   }
 151   }();
 152
 153   [&]() {
 154   unsigned int bit_var = true;
 155 #pragma omp target teams distribute reduction(&: bit_var)
 156   for (int i = 0; i < 2; ++i) {
 157     bit_var = (bit_var & i);
 158     [&]() {
 159       sivar += 4;
 160     }();
 161   }
 162   }();
 163
 164   [&]() {
 165   unsigned int bit_var = false;
 166 #pragma omp target teams distribute reduction(|: bit_var)
 167   for (int i = 0; i < 2; ++i) {
 168     bit_var = (bit_var | i);
 169     [&]() {
 170       sivar += 4;
 171     }();
 172   }
 173   }();
 174
 175   [&]() {
 176   unsigned int bit_var = false;
 177 #pragma omp target teams distribute reduction(^: bit_var)
 178   for (int i = 0; i < 2; ++i) {
 179     bit_var = (bit_var ^ i);
 180     [&]() {
 181       sivar += 4;
 182     }();
 183   }
 184   }();
 185
 186   [&]() {
 187   int max_var = 0;
 188 #pragma omp target teams distribute reduction(max: max_var)
 189   for (int i = 0; i < 2; ++i) {
 190     max_var = max_var >= i ? max_var : i;
 191     [&]() {
 192       max_var += 4;
 193     }();
 194   }
 195   }();
 196
 197   [&]() {
 198   int min_var = 10;
 199 #pragma omp target teams distribute reduction(min: min_var)
 200   for (int i = 0; i < 2; ++i) {
 201     min_var = min_var <= i ? min_var : i;
 202     [&]() {
 203       min_var += 4;
 204     }();
 205   }
 206   }();
 207   return 0;
 208 #else
 209 #pragma omp target teams distribute reduction(+: sivar)
 210   for (int i = 0; i < 2; ++i) {
 211     sivar += i;
 212   }
 213
 214 #pragma omp target teams distribute reduction(-: sivar)
 215   for (int i = 0; i < 2; ++i) {
 216     sivar -= i;
 217   }
 218
 219   sivar = 1;
 220 #pragma omp target teams distribute reduction(*: sivar)
 221   for (int i = 0; i < 2; ++i) {
 222     sivar = sivar * i;
 223   }
 224
 225   bool and_var = true;
 226 #pragma omp target teams distribute reduction(&&: and_var)
 227   for (int i = 0; i < 2; ++i) {
 228     and_var = and_var && (i == 0);
 229   }
 230
 231   bool or_var = false;
 232 #pragma omp target teams distribute reduction(||: or_var)
 233   for (int i = 0; i < 2; ++i) {
 234     or_var = or_var || (i == 0);
 235   }
 236
 237   unsigned int bit_var = 1;
 238 #pragma omp target teams distribute reduction(&: bit_var)
 239   for (int i = 0; i < 2; ++i) {
 240     bit_var = (bit_var & i);
 241   }
 242
 243 #pragma omp target teams distribute reduction(|: bit_var)
 244   for (int i = 0; i < 2; ++i) {
 245     bit_var = (bit_var | i);
 246   }
 247
 248 #pragma omp target teams distribute reduction(^: bit_var)
 249   for (int i = 0; i < 2; ++i) {
 250     bit_var = (bit_var ^ i);
 251   }
 252
 253   sivar = 0;
 254 #pragma omp target teams distribute reduction(max: sivar)
 255   for (int i = 0; i < 2; ++i) {
 256     sivar = sivar >= i ? sivar : i;
 257   }
 258
 259   sivar = 10;
 260 #pragma omp target teams distribute reduction(min: sivar)
 261   for (int i = 0; i < 2; ++i) {
 262     sivar = sivar <= i ? sivar : i;
 263   }
 264
 265   return tmain<int>();
 266 #endif
 267 }
 268
 269
 270 #endif
 271 // CHECK1-LABEL: define {{[^@]+}}@main
 272 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 273 // CHECK1-NEXT:  entry:
 274 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 275 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 276 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 277 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 278 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 279 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 280 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x ptr], align 8
 281 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x ptr], align 8
 282 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x ptr], align 8
 283 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
 284 // CHECK1-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 285 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x ptr], align 8
 286 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x ptr], align 8
 287 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x ptr], align 8
 288 // CHECK1-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
 289 // CHECK1-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 290 // CHECK1-NEXT:    [[AND_VAR:%.*]] = alloca i8, align 1
 291 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x ptr], align 8
 292 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x ptr], align 8
 293 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x ptr], align 8
 294 // CHECK1-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
 295 // CHECK1-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 296 // CHECK1-NEXT:    [[OR_VAR:%.*]] = alloca i8, align 1
 297 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [1 x ptr], align 8
 298 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS23:%.*]] = alloca [1 x ptr], align 8
 299 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [1 x ptr], align 8
 300 // CHECK1-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
 301 // CHECK1-NEXT:    [[KERNEL_ARGS26:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 302 // CHECK1-NEXT:    [[BIT_VAR:%.*]] = alloca i32, align 4
 303 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [1 x ptr], align 8
 304 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [1 x ptr], align 8
 305 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [1 x ptr], align 8
 306 // CHECK1-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
 307 // CHECK1-NEXT:    [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 308 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [1 x ptr], align 8
 309 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [1 x ptr], align 8
 310 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [1 x ptr], align 8
 311 // CHECK1-NEXT:    [[_TMP39:%.*]] = alloca i32, align 4
 312 // CHECK1-NEXT:    [[KERNEL_ARGS40:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 313 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS43:%.*]] = alloca [1 x ptr], align 8
 314 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS44:%.*]] = alloca [1 x ptr], align 8
 315 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS45:%.*]] = alloca [1 x ptr], align 8
 316 // CHECK1-NEXT:    [[_TMP46:%.*]] = alloca i32, align 4
 317 // CHECK1-NEXT:    [[KERNEL_ARGS47:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 318 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS50:%.*]] = alloca [1 x ptr], align 8
 319 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS51:%.*]] = alloca [1 x ptr], align 8
 320 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS52:%.*]] = alloca [1 x ptr], align 8
 321 // CHECK1-NEXT:    [[_TMP53:%.*]] = alloca i32, align 4
 322 // CHECK1-NEXT:    [[KERNEL_ARGS54:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 323 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS57:%.*]] = alloca [1 x ptr], align 8
 324 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS58:%.*]] = alloca [1 x ptr], align 8
 325 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS59:%.*]] = alloca [1 x ptr], align 8
 326 // CHECK1-NEXT:    [[_TMP60:%.*]] = alloca i32, align 4
 327 // CHECK1-NEXT:    [[KERNEL_ARGS61:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 328 // CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 329 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 330 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP0]], align 8
 331 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 332 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP1]], align 8
 333 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 334 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 335 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 336 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 337 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 338 // CHECK1-NEXT:    store i32 2, ptr [[TMP5]], align 4
 339 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 340 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
 341 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 342 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 343 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 344 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 345 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 346 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 8
 347 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 348 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 8
 349 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 350 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 351 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 352 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 353 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 354 // CHECK1-NEXT:    store i64 2, ptr [[TMP13]], align 8
 355 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 356 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
 357 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 358 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 359 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 360 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 361 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 362 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
 363 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.region_id, ptr [[KERNEL_ARGS]])
 364 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 365 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 366 // CHECK1:       omp_offload.failed:
 367 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209(ptr @_ZZ4mainE5sivar) #[[ATTR2:[0-9]+]]
 368 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 369 // CHECK1:       omp_offload.cont:
 370 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
 371 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP20]], align 8
 372 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
 373 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP21]], align 8
 374 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
 375 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 376 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
 377 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
 378 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
 379 // CHECK1-NEXT:    store i32 2, ptr [[TMP25]], align 4
 380 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
 381 // CHECK1-NEXT:    store i32 1, ptr [[TMP26]], align 4
 382 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
 383 // CHECK1-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
 384 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
 385 // CHECK1-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
 386 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
 387 // CHECK1-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP29]], align 8
 388 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
 389 // CHECK1-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP30]], align 8
 390 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
 391 // CHECK1-NEXT:    store ptr null, ptr [[TMP31]], align 8
 392 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
 393 // CHECK1-NEXT:    store ptr null, ptr [[TMP32]], align 8
 394 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
 395 // CHECK1-NEXT:    store i64 2, ptr [[TMP33]], align 8
 396 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
 397 // CHECK1-NEXT:    store i64 0, ptr [[TMP34]], align 8
 398 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
 399 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
 400 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
 401 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
 402 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
 403 // CHECK1-NEXT:    store i32 0, ptr [[TMP37]], align 4
 404 // CHECK1-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.region_id, ptr [[KERNEL_ARGS5]])
 405 // CHECK1-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 406 // CHECK1-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
 407 // CHECK1:       omp_offload.failed6:
 408 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
 409 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
 410 // CHECK1:       omp_offload.cont7:
 411 // CHECK1-NEXT:    store i32 1, ptr @_ZZ4mainE5sivar, align 4
 412 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
 413 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP40]], align 8
 414 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
 415 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP41]], align 8
 416 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
 417 // CHECK1-NEXT:    store ptr null, ptr [[TMP42]], align 8
 418 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
 419 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
 420 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
 421 // CHECK1-NEXT:    store i32 2, ptr [[TMP45]], align 4
 422 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
 423 // CHECK1-NEXT:    store i32 1, ptr [[TMP46]], align 4
 424 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
 425 // CHECK1-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 8
 426 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
 427 // CHECK1-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 8
 428 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
 429 // CHECK1-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP49]], align 8
 430 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
 431 // CHECK1-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP50]], align 8
 432 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
 433 // CHECK1-NEXT:    store ptr null, ptr [[TMP51]], align 8
 434 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
 435 // CHECK1-NEXT:    store ptr null, ptr [[TMP52]], align 8
 436 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
 437 // CHECK1-NEXT:    store i64 2, ptr [[TMP53]], align 8
 438 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
 439 // CHECK1-NEXT:    store i64 0, ptr [[TMP54]], align 8
 440 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
 441 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
 442 // CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
 443 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
 444 // CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
 445 // CHECK1-NEXT:    store i32 0, ptr [[TMP57]], align 4
 446 // CHECK1-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.region_id, ptr [[KERNEL_ARGS12]])
 447 // CHECK1-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
 448 // CHECK1-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
 449 // CHECK1:       omp_offload.failed13:
 450 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
 451 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
 452 // CHECK1:       omp_offload.cont14:
 453 // CHECK1-NEXT:    store i8 1, ptr [[AND_VAR]], align 1
 454 // CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
 455 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[TMP60]], align 8
 456 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
 457 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[TMP61]], align 8
 458 // CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS17]], i64 0, i64 0
 459 // CHECK1-NEXT:    store ptr null, ptr [[TMP62]], align 8
 460 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
 461 // CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
 462 // CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 0
 463 // CHECK1-NEXT:    store i32 2, ptr [[TMP65]], align 4
 464 // CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 1
 465 // CHECK1-NEXT:    store i32 1, ptr [[TMP66]], align 4
 466 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 2
 467 // CHECK1-NEXT:    store ptr [[TMP63]], ptr [[TMP67]], align 8
 468 // CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 3
 469 // CHECK1-NEXT:    store ptr [[TMP64]], ptr [[TMP68]], align 8
 470 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 4
 471 // CHECK1-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP69]], align 8
 472 // CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 5
 473 // CHECK1-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP70]], align 8
 474 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 6
 475 // CHECK1-NEXT:    store ptr null, ptr [[TMP71]], align 8
 476 // CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 7
 477 // CHECK1-NEXT:    store ptr null, ptr [[TMP72]], align 8
 478 // CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 8
 479 // CHECK1-NEXT:    store i64 2, ptr [[TMP73]], align 8
 480 // CHECK1-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 9
 481 // CHECK1-NEXT:    store i64 0, ptr [[TMP74]], align 8
 482 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 10
 483 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP75]], align 4
 484 // CHECK1-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 11
 485 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP76]], align 4
 486 // CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 12
 487 // CHECK1-NEXT:    store i32 0, ptr [[TMP77]], align 4
 488 // CHECK1-NEXT:    [[TMP78:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.region_id, ptr [[KERNEL_ARGS19]])
 489 // CHECK1-NEXT:    [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
 490 // CHECK1-NEXT:    br i1 [[TMP79]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
 491 // CHECK1:       omp_offload.failed20:
 492 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226(ptr [[AND_VAR]]) #[[ATTR2]]
 493 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
 494 // CHECK1:       omp_offload.cont21:
 495 // CHECK1-NEXT:    store i8 0, ptr [[OR_VAR]], align 1
 496 // CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
 497 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[TMP80]], align 8
 498 // CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
 499 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[TMP81]], align 8
 500 // CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 0
 501 // CHECK1-NEXT:    store ptr null, ptr [[TMP82]], align 8
 502 // CHECK1-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
 503 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
 504 // CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 0
 505 // CHECK1-NEXT:    store i32 2, ptr [[TMP85]], align 4
 506 // CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 1
 507 // CHECK1-NEXT:    store i32 1, ptr [[TMP86]], align 4
 508 // CHECK1-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 2
 509 // CHECK1-NEXT:    store ptr [[TMP83]], ptr [[TMP87]], align 8
 510 // CHECK1-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 3
 511 // CHECK1-NEXT:    store ptr [[TMP84]], ptr [[TMP88]], align 8
 512 // CHECK1-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 4
 513 // CHECK1-NEXT:    store ptr @.offload_sizes.7, ptr [[TMP89]], align 8
 514 // CHECK1-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 5
 515 // CHECK1-NEXT:    store ptr @.offload_maptypes.8, ptr [[TMP90]], align 8
 516 // CHECK1-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 6
 517 // CHECK1-NEXT:    store ptr null, ptr [[TMP91]], align 8
 518 // CHECK1-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 7
 519 // CHECK1-NEXT:    store ptr null, ptr [[TMP92]], align 8
 520 // CHECK1-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 8
 521 // CHECK1-NEXT:    store i64 2, ptr [[TMP93]], align 8
 522 // CHECK1-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 9
 523 // CHECK1-NEXT:    store i64 0, ptr [[TMP94]], align 8
 524 // CHECK1-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 10
 525 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP95]], align 4
 526 // CHECK1-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 11
 527 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP96]], align 4
 528 // CHECK1-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 12
 529 // CHECK1-NEXT:    store i32 0, ptr [[TMP97]], align 4
 530 // CHECK1-NEXT:    [[TMP98:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.region_id, ptr [[KERNEL_ARGS26]])
 531 // CHECK1-NEXT:    [[TMP99:%.*]] = icmp ne i32 [[TMP98]], 0
 532 // CHECK1-NEXT:    br i1 [[TMP99]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
 533 // CHECK1:       omp_offload.failed27:
 534 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232(ptr [[OR_VAR]]) #[[ATTR2]]
 535 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
 536 // CHECK1:       omp_offload.cont28:
 537 // CHECK1-NEXT:    store i32 1, ptr [[BIT_VAR]], align 4
 538 // CHECK1-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
 539 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP100]], align 8
 540 // CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
 541 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP101]], align 8
 542 // CHECK1-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 0
 543 // CHECK1-NEXT:    store ptr null, ptr [[TMP102]], align 8
 544 // CHECK1-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
 545 // CHECK1-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
 546 // CHECK1-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 0
 547 // CHECK1-NEXT:    store i32 2, ptr [[TMP105]], align 4
 548 // CHECK1-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 1
 549 // CHECK1-NEXT:    store i32 1, ptr [[TMP106]], align 4
 550 // CHECK1-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 2
 551 // CHECK1-NEXT:    store ptr [[TMP103]], ptr [[TMP107]], align 8
 552 // CHECK1-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 3
 553 // CHECK1-NEXT:    store ptr [[TMP104]], ptr [[TMP108]], align 8
 554 // CHECK1-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 4
 555 // CHECK1-NEXT:    store ptr @.offload_sizes.9, ptr [[TMP109]], align 8
 556 // CHECK1-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 5
 557 // CHECK1-NEXT:    store ptr @.offload_maptypes.10, ptr [[TMP110]], align 8
 558 // CHECK1-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 6
 559 // CHECK1-NEXT:    store ptr null, ptr [[TMP111]], align 8
 560 // CHECK1-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 7
 561 // CHECK1-NEXT:    store ptr null, ptr [[TMP112]], align 8
 562 // CHECK1-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 8
 563 // CHECK1-NEXT:    store i64 2, ptr [[TMP113]], align 8
 564 // CHECK1-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 9
 565 // CHECK1-NEXT:    store i64 0, ptr [[TMP114]], align 8
 566 // CHECK1-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 10
 567 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP115]], align 4
 568 // CHECK1-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 11
 569 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP116]], align 4
 570 // CHECK1-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 12
 571 // CHECK1-NEXT:    store i32 0, ptr [[TMP117]], align 4
 572 // CHECK1-NEXT:    [[TMP118:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.region_id, ptr [[KERNEL_ARGS33]])
 573 // CHECK1-NEXT:    [[TMP119:%.*]] = icmp ne i32 [[TMP118]], 0
 574 // CHECK1-NEXT:    br i1 [[TMP119]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
 575 // CHECK1:       omp_offload.failed34:
 576 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238(ptr [[BIT_VAR]]) #[[ATTR2]]
 577 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT35]]
 578 // CHECK1:       omp_offload.cont35:
 579 // CHECK1-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
 580 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP120]], align 8
 581 // CHECK1-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
 582 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP121]], align 8
 583 // CHECK1-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
 584 // CHECK1-NEXT:    store ptr null, ptr [[TMP122]], align 8
 585 // CHECK1-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
 586 // CHECK1-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
 587 // CHECK1-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 0
 588 // CHECK1-NEXT:    store i32 2, ptr [[TMP125]], align 4
 589 // CHECK1-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 1
 590 // CHECK1-NEXT:    store i32 1, ptr [[TMP126]], align 4
 591 // CHECK1-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 2
 592 // CHECK1-NEXT:    store ptr [[TMP123]], ptr [[TMP127]], align 8
 593 // CHECK1-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 3
 594 // CHECK1-NEXT:    store ptr [[TMP124]], ptr [[TMP128]], align 8
 595 // CHECK1-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 4
 596 // CHECK1-NEXT:    store ptr @.offload_sizes.11, ptr [[TMP129]], align 8
 597 // CHECK1-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 5
 598 // CHECK1-NEXT:    store ptr @.offload_maptypes.12, ptr [[TMP130]], align 8
 599 // CHECK1-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 6
 600 // CHECK1-NEXT:    store ptr null, ptr [[TMP131]], align 8
 601 // CHECK1-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 7
 602 // CHECK1-NEXT:    store ptr null, ptr [[TMP132]], align 8
 603 // CHECK1-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 8
 604 // CHECK1-NEXT:    store i64 2, ptr [[TMP133]], align 8
 605 // CHECK1-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 9
 606 // CHECK1-NEXT:    store i64 0, ptr [[TMP134]], align 8
 607 // CHECK1-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 10
 608 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP135]], align 4
 609 // CHECK1-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 11
 610 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP136]], align 4
 611 // CHECK1-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 12
 612 // CHECK1-NEXT:    store i32 0, ptr [[TMP137]], align 4
 613 // CHECK1-NEXT:    [[TMP138:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.region_id, ptr [[KERNEL_ARGS40]])
 614 // CHECK1-NEXT:    [[TMP139:%.*]] = icmp ne i32 [[TMP138]], 0
 615 // CHECK1-NEXT:    br i1 [[TMP139]], label [[OMP_OFFLOAD_FAILED41:%.*]], label [[OMP_OFFLOAD_CONT42:%.*]]
 616 // CHECK1:       omp_offload.failed41:
 617 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243(ptr [[BIT_VAR]]) #[[ATTR2]]
 618 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT42]]
 619 // CHECK1:       omp_offload.cont42:
 620 // CHECK1-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
 621 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP140]], align 8
 622 // CHECK1-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
 623 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP141]], align 8
 624 // CHECK1-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS45]], i64 0, i64 0
 625 // CHECK1-NEXT:    store ptr null, ptr [[TMP142]], align 8
 626 // CHECK1-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
 627 // CHECK1-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
 628 // CHECK1-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 0
 629 // CHECK1-NEXT:    store i32 2, ptr [[TMP145]], align 4
 630 // CHECK1-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 1
 631 // CHECK1-NEXT:    store i32 1, ptr [[TMP146]], align 4
 632 // CHECK1-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 2
 633 // CHECK1-NEXT:    store ptr [[TMP143]], ptr [[TMP147]], align 8
 634 // CHECK1-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 3
 635 // CHECK1-NEXT:    store ptr [[TMP144]], ptr [[TMP148]], align 8
 636 // CHECK1-NEXT:    [[TMP149:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 4
 637 // CHECK1-NEXT:    store ptr @.offload_sizes.13, ptr [[TMP149]], align 8
 638 // CHECK1-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 5
 639 // CHECK1-NEXT:    store ptr @.offload_maptypes.14, ptr [[TMP150]], align 8
 640 // CHECK1-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 6
 641 // CHECK1-NEXT:    store ptr null, ptr [[TMP151]], align 8
 642 // CHECK1-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 7
 643 // CHECK1-NEXT:    store ptr null, ptr [[TMP152]], align 8
 644 // CHECK1-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 8
 645 // CHECK1-NEXT:    store i64 2, ptr [[TMP153]], align 8
 646 // CHECK1-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 9
 647 // CHECK1-NEXT:    store i64 0, ptr [[TMP154]], align 8
 648 // CHECK1-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 10
 649 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP155]], align 4
 650 // CHECK1-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 11
 651 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP156]], align 4
 652 // CHECK1-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 12
 653 // CHECK1-NEXT:    store i32 0, ptr [[TMP157]], align 4
 654 // CHECK1-NEXT:    [[TMP158:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.region_id, ptr [[KERNEL_ARGS47]])
 655 // CHECK1-NEXT:    [[TMP159:%.*]] = icmp ne i32 [[TMP158]], 0
 656 // CHECK1-NEXT:    br i1 [[TMP159]], label [[OMP_OFFLOAD_FAILED48:%.*]], label [[OMP_OFFLOAD_CONT49:%.*]]
 657 // CHECK1:       omp_offload.failed48:
 658 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248(ptr [[BIT_VAR]]) #[[ATTR2]]
 659 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT49]]
 660 // CHECK1:       omp_offload.cont49:
 661 // CHECK1-NEXT:    store i32 0, ptr @_ZZ4mainE5sivar, align 4
 662 // CHECK1-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
 663 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP160]], align 8
 664 // CHECK1-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
 665 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP161]], align 8
 666 // CHECK1-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS52]], i64 0, i64 0
 667 // CHECK1-NEXT:    store ptr null, ptr [[TMP162]], align 8
 668 // CHECK1-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
 669 // CHECK1-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
 670 // CHECK1-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 0
 671 // CHECK1-NEXT:    store i32 2, ptr [[TMP165]], align 4
 672 // CHECK1-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 1
 673 // CHECK1-NEXT:    store i32 1, ptr [[TMP166]], align 4
 674 // CHECK1-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 2
 675 // CHECK1-NEXT:    store ptr [[TMP163]], ptr [[TMP167]], align 8
 676 // CHECK1-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 3
 677 // CHECK1-NEXT:    store ptr [[TMP164]], ptr [[TMP168]], align 8
 678 // CHECK1-NEXT:    [[TMP169:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 4
 679 // CHECK1-NEXT:    store ptr @.offload_sizes.15, ptr [[TMP169]], align 8
 680 // CHECK1-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 5
 681 // CHECK1-NEXT:    store ptr @.offload_maptypes.16, ptr [[TMP170]], align 8
 682 // CHECK1-NEXT:    [[TMP171:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 6
 683 // CHECK1-NEXT:    store ptr null, ptr [[TMP171]], align 8
 684 // CHECK1-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 7
 685 // CHECK1-NEXT:    store ptr null, ptr [[TMP172]], align 8
 686 // CHECK1-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 8
 687 // CHECK1-NEXT:    store i64 2, ptr [[TMP173]], align 8
 688 // CHECK1-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 9
 689 // CHECK1-NEXT:    store i64 0, ptr [[TMP174]], align 8
 690 // CHECK1-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 10
 691 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP175]], align 4
 692 // CHECK1-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 11
 693 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP176]], align 4
 694 // CHECK1-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 12
 695 // CHECK1-NEXT:    store i32 0, ptr [[TMP177]], align 4
 696 // CHECK1-NEXT:    [[TMP178:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.region_id, ptr [[KERNEL_ARGS54]])
 697 // CHECK1-NEXT:    [[TMP179:%.*]] = icmp ne i32 [[TMP178]], 0
 698 // CHECK1-NEXT:    br i1 [[TMP179]], label [[OMP_OFFLOAD_FAILED55:%.*]], label [[OMP_OFFLOAD_CONT56:%.*]]
 699 // CHECK1:       omp_offload.failed55:
 700 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
 701 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT56]]
 702 // CHECK1:       omp_offload.cont56:
 703 // CHECK1-NEXT:    store i32 10, ptr @_ZZ4mainE5sivar, align 4
 704 // CHECK1-NEXT:    [[TMP180:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
 705 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP180]], align 8
 706 // CHECK1-NEXT:    [[TMP181:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
 707 // CHECK1-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP181]], align 8
 708 // CHECK1-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS59]], i64 0, i64 0
 709 // CHECK1-NEXT:    store ptr null, ptr [[TMP182]], align 8
 710 // CHECK1-NEXT:    [[TMP183:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
 711 // CHECK1-NEXT:    [[TMP184:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
 712 // CHECK1-NEXT:    [[TMP185:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 0
 713 // CHECK1-NEXT:    store i32 2, ptr [[TMP185]], align 4
 714 // CHECK1-NEXT:    [[TMP186:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 1
 715 // CHECK1-NEXT:    store i32 1, ptr [[TMP186]], align 4
 716 // CHECK1-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 2
 717 // CHECK1-NEXT:    store ptr [[TMP183]], ptr [[TMP187]], align 8
 718 // CHECK1-NEXT:    [[TMP188:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 3
 719 // CHECK1-NEXT:    store ptr [[TMP184]], ptr [[TMP188]], align 8
 720 // CHECK1-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 4
 721 // CHECK1-NEXT:    store ptr @.offload_sizes.17, ptr [[TMP189]], align 8
 722 // CHECK1-NEXT:    [[TMP190:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 5
 723 // CHECK1-NEXT:    store ptr @.offload_maptypes.18, ptr [[TMP190]], align 8
 724 // CHECK1-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 6
 725 // CHECK1-NEXT:    store ptr null, ptr [[TMP191]], align 8
 726 // CHECK1-NEXT:    [[TMP192:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 7
 727 // CHECK1-NEXT:    store ptr null, ptr [[TMP192]], align 8
 728 // CHECK1-NEXT:    [[TMP193:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 8
 729 // CHECK1-NEXT:    store i64 2, ptr [[TMP193]], align 8
 730 // CHECK1-NEXT:    [[TMP194:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 9
 731 // CHECK1-NEXT:    store i64 0, ptr [[TMP194]], align 8
 732 // CHECK1-NEXT:    [[TMP195:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 10
 733 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP195]], align 4
 734 // CHECK1-NEXT:    [[TMP196:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 11
 735 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP196]], align 4
 736 // CHECK1-NEXT:    [[TMP197:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 12
 737 // CHECK1-NEXT:    store i32 0, ptr [[TMP197]], align 4
 738 // CHECK1-NEXT:    [[TMP198:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.region_id, ptr [[KERNEL_ARGS61]])
 739 // CHECK1-NEXT:    [[TMP199:%.*]] = icmp ne i32 [[TMP198]], 0
 740 // CHECK1-NEXT:    br i1 [[TMP199]], label [[OMP_OFFLOAD_FAILED62:%.*]], label [[OMP_OFFLOAD_CONT63:%.*]]
 741 // CHECK1:       omp_offload.failed62:
 742 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
 743 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT63]]
 744 // CHECK1:       omp_offload.cont63:
 745 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
 746 // CHECK1-NEXT:    ret i32 [[CALL]]
 747 //
 748 //
 749 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209
 750 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] {
 751 // CHECK1-NEXT:  entry:
 752 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
 753 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 754 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
 755 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined, ptr [[TMP0]])
 756 // CHECK1-NEXT:    ret void
 757 //
 758 //
 759 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined
 760 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
 761 // CHECK1-NEXT:  entry:
 762 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 763 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 764 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
 765 // CHECK1-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
 766 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 767 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 768 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 769 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 770 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 771 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 772 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 773 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 774 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 775 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 776 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 777 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
 778 // CHECK1-NEXT:    store i32 0, ptr [[SIVAR1]], align 4
 779 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 780 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 781 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 782 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 783 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 784 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 785 // 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)
 786 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 787 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
 788 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 789 // CHECK1:       cond.true:
 790 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 791 // CHECK1:       cond.false:
 792 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 793 // CHECK1-NEXT:    br label [[COND_END]]
 794 // CHECK1:       cond.end:
 795 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 796 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 797 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 798 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 799 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 800 // CHECK1:       omp.inner.for.cond:
 801 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 802 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 803 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 804 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 805 // CHECK1:       omp.inner.for.body:
 806 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 807 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 808 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 809 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 810 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 811 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
 812 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
 813 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[SIVAR1]], align 4
 814 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 815 // CHECK1:       omp.body.continue:
 816 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 817 // CHECK1:       omp.inner.for.inc:
 818 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 819 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
 820 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
 821 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 822 // CHECK1:       omp.inner.for.end:
 823 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 824 // CHECK1:       omp.loop.exit:
 825 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 826 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 827 // CHECK1-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 8
 828 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 829 // CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 830 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 831 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 832 // CHECK1-NEXT:    ]
 833 // CHECK1:       .omp.reduction.case1:
 834 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
 835 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
 836 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
 837 // CHECK1-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
 838 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 839 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 840 // CHECK1:       .omp.reduction.case2:
 841 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
 842 // CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
 843 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 844 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 845 // CHECK1:       .omp.reduction.default:
 846 // CHECK1-NEXT:    ret void
 847 //
 848 //
 849 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined.omp.reduction.reduction_func
 850 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 851 // CHECK1-NEXT:  entry:
 852 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 853 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 854 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 855 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 856 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 857 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 858 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 859 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 860 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 861 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 862 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 863 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
 864 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
 865 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
 866 // CHECK1-NEXT:    ret void
 867 //
 868 //
 869 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214
 870 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
 871 // CHECK1-NEXT:  entry:
 872 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
 873 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 874 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
 875 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined, ptr [[TMP0]])
 876 // CHECK1-NEXT:    ret void
 877 //
 878 //
 879 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined
 880 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
 881 // CHECK1-NEXT:  entry:
 882 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 883 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 884 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
 885 // CHECK1-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
 886 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 887 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 888 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 889 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 890 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 891 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 892 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 893 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 894 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 895 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 896 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 897 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
 898 // CHECK1-NEXT:    store i32 0, ptr [[SIVAR1]], align 4
 899 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 900 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
 901 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 902 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 903 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 904 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 905 // 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)
 906 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 907 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
 908 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 909 // CHECK1:       cond.true:
 910 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 911 // CHECK1:       cond.false:
 912 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 913 // CHECK1-NEXT:    br label [[COND_END]]
 914 // CHECK1:       cond.end:
 915 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 916 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 917 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 918 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 919 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 920 // CHECK1:       omp.inner.for.cond:
 921 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 922 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 923 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 924 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 925 // CHECK1:       omp.inner.for.body:
 926 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 927 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 928 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 929 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 930 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 931 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
 932 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP10]], [[TMP9]]
 933 // CHECK1-NEXT:    store i32 [[SUB]], ptr [[SIVAR1]], align 4
 934 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 935 // CHECK1:       omp.body.continue:
 936 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 937 // CHECK1:       omp.inner.for.inc:
 938 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 939 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
 940 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
 941 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 942 // CHECK1:       omp.inner.for.end:
 943 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 944 // CHECK1:       omp.loop.exit:
 945 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 946 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 947 // CHECK1-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 8
 948 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 949 // CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 950 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 951 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 952 // CHECK1-NEXT:    ]
 953 // CHECK1:       .omp.reduction.case1:
 954 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
 955 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
 956 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
 957 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[TMP0]], align 4
 958 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 959 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 960 // CHECK1:       .omp.reduction.case2:
 961 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
 962 // CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
 963 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
 964 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 965 // CHECK1:       .omp.reduction.default:
 966 // CHECK1-NEXT:    ret void
 967 //
 968 //
 969 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined.omp.reduction.reduction_func
 970 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
 971 // CHECK1-NEXT:  entry:
 972 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 973 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 974 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 975 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 976 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 977 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 978 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 979 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 980 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 981 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 982 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 983 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
 984 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
 985 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
 986 // CHECK1-NEXT:    ret void
 987 //
 988 //
 989 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220
 990 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
 991 // CHECK1-NEXT:  entry:
 992 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
 993 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
 994 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
 995 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined, ptr [[TMP0]])
 996 // CHECK1-NEXT:    ret void
 997 //
 998 //
 999 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined
1000 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
1001 // CHECK1-NEXT:  entry:
1002 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1003 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1004 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
1005 // CHECK1-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
1006 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1007 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1008 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1009 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1010 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1011 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1012 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1013 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1014 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i32, align 4
1015 // CHECK1-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
1016 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1017 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1018 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
1019 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
1020 // CHECK1-NEXT:    store i32 1, ptr [[SIVAR1]], align 4
1021 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1022 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1023 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1024 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1025 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1026 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1027 // 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)
1028 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1029 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
1030 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1031 // CHECK1:       cond.true:
1032 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1033 // CHECK1:       cond.false:
1034 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1035 // CHECK1-NEXT:    br label [[COND_END]]
1036 // CHECK1:       cond.end:
1037 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1038 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1039 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1040 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1041 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1042 // CHECK1:       omp.inner.for.cond:
1043 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1044 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1045 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1046 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1047 // CHECK1:       omp.inner.for.body:
1048 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1049 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1050 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1051 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1052 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[SIVAR1]], align 4
1053 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
1054 // CHECK1-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[TMP9]], [[TMP10]]
1055 // CHECK1-NEXT:    store i32 [[MUL3]], ptr [[SIVAR1]], align 4
1056 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1057 // CHECK1:       omp.body.continue:
1058 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1059 // CHECK1:       omp.inner.for.inc:
1060 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1061 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
1062 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
1063 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1064 // CHECK1:       omp.inner.for.end:
1065 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1066 // CHECK1:       omp.loop.exit:
1067 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1068 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1069 // CHECK1-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 8
1070 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
1071 // CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1072 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1073 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1074 // CHECK1-NEXT:    ]
1075 // CHECK1:       .omp.reduction.case1:
1076 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
1077 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
1078 // CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP14]], [[TMP15]]
1079 // CHECK1-NEXT:    store i32 [[MUL5]], ptr [[TMP0]], align 4
1080 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1081 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1082 // CHECK1:       .omp.reduction.case2:
1083 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
1084 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, ptr [[TMP0]] monotonic, align 4
1085 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
1086 // CHECK1:       atomic_cont:
1087 // CHECK1-NEXT:    [[TMP17:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP22:%.*]], [[ATOMIC_CONT]] ]
1088 // CHECK1-NEXT:    store i32 [[TMP17]], ptr [[_TMP6]], align 4
1089 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[_TMP6]], align 4
1090 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[SIVAR1]], align 4
1091 // CHECK1-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[TMP18]], [[TMP19]]
1092 // CHECK1-NEXT:    store i32 [[MUL7]], ptr [[ATOMIC_TEMP]], align 4
1093 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
1094 // CHECK1-NEXT:    [[TMP21:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP17]], i32 [[TMP20]] monotonic monotonic, align 4
1095 // CHECK1-NEXT:    [[TMP22]] = extractvalue { i32, i1 } [[TMP21]], 0
1096 // CHECK1-NEXT:    [[TMP23:%.*]] = extractvalue { i32, i1 } [[TMP21]], 1
1097 // CHECK1-NEXT:    br i1 [[TMP23]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
1098 // CHECK1:       atomic_exit:
1099 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1100 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1101 // CHECK1:       .omp.reduction.default:
1102 // CHECK1-NEXT:    ret void
1103 //
1104 //
1105 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined.omp.reduction.reduction_func
1106 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
1107 // CHECK1-NEXT:  entry:
1108 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
1109 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
1110 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
1111 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
1112 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
1113 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
1114 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
1115 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
1116 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
1117 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
1118 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1119 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
1120 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], [[TMP9]]
1121 // CHECK1-NEXT:    store i32 [[MUL]], ptr [[TMP7]], align 4
1122 // CHECK1-NEXT:    ret void
1123 //
1124 //
1125 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226
1126 // CHECK1-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]]) #[[ATTR1]] {
1127 // CHECK1-NEXT:  entry:
1128 // CHECK1-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 8
1129 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 8
1130 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 8
1131 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined, ptr [[TMP0]])
1132 // CHECK1-NEXT:    ret void
1133 //
1134 //
1135 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined
1136 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]]) #[[ATTR1]] {
1137 // CHECK1-NEXT:  entry:
1138 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1139 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1140 // CHECK1-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 8
1141 // CHECK1-NEXT:    [[AND_VAR1:%.*]] = alloca i8, align 1
1142 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1143 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1144 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1145 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1146 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1147 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1148 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1149 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1150 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
1151 // CHECK1-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
1152 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1153 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1154 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 8
1155 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 8
1156 // CHECK1-NEXT:    store i8 1, ptr [[AND_VAR1]], align 1
1157 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1158 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1159 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1160 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1161 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1162 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1163 // 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)
1164 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1165 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
1166 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1167 // CHECK1:       cond.true:
1168 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1169 // CHECK1:       cond.false:
1170 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1171 // CHECK1-NEXT:    br label [[COND_END]]
1172 // CHECK1:       cond.end:
1173 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1174 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1175 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1176 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1177 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1178 // CHECK1:       omp.inner.for.cond:
1179 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1180 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1181 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1182 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1183 // CHECK1:       omp.inner.for.body:
1184 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1185 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1186 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1187 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1188 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[AND_VAR1]], align 1
1189 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP9]] to i1
1190 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
1191 // CHECK1:       land.rhs:
1192 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
1193 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP10]], 0
1194 // CHECK1-NEXT:    br label [[LAND_END]]
1195 // CHECK1:       land.end:
1196 // CHECK1-NEXT:    [[TMP11:%.*]] = phi i1 [ false, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LAND_RHS]] ]
1197 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP11]] to i8
1198 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[AND_VAR1]], align 1
1199 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1200 // CHECK1:       omp.body.continue:
1201 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1202 // CHECK1:       omp.inner.for.inc:
1203 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1204 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
1205 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
1206 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1207 // CHECK1:       omp.inner.for.end:
1208 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1209 // CHECK1:       omp.loop.exit:
1210 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1211 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1212 // CHECK1-NEXT:    store ptr [[AND_VAR1]], ptr [[TMP13]], align 8
1213 // CHECK1-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
1214 // CHECK1-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1215 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1216 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1217 // CHECK1-NEXT:    ]
1218 // CHECK1:       .omp.reduction.case1:
1219 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8, ptr [[TMP0]], align 1
1220 // CHECK1-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP15]] to i1
1221 // CHECK1-NEXT:    br i1 [[TOBOOL5]], label [[LAND_RHS6:%.*]], label [[LAND_END8:%.*]]
1222 // CHECK1:       land.rhs6:
1223 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8, ptr [[AND_VAR1]], align 1
1224 // CHECK1-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP16]] to i1
1225 // CHECK1-NEXT:    br label [[LAND_END8]]
1226 // CHECK1:       land.end8:
1227 // CHECK1-NEXT:    [[TMP17:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LAND_RHS6]] ]
1228 // CHECK1-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP17]] to i8
1229 // CHECK1-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
1230 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1231 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1232 // CHECK1:       .omp.reduction.case2:
1233 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[AND_VAR1]], align 1
1234 // CHECK1-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP18]] to i1
1235 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
1236 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
1237 // CHECK1:       atomic_cont:
1238 // CHECK1-NEXT:    [[TMP19:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP25:%.*]], [[LAND_END17:%.*]] ]
1239 // CHECK1-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP19]] to i1
1240 // CHECK1-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
1241 // CHECK1-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
1242 // CHECK1-NEXT:    [[TMP20:%.*]] = load i8, ptr [[_TMP12]], align 1
1243 // CHECK1-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP20]] to i1
1244 // CHECK1-NEXT:    br i1 [[TOBOOL14]], label [[LAND_RHS15:%.*]], label [[LAND_END17]]
1245 // CHECK1:       land.rhs15:
1246 // CHECK1-NEXT:    [[TMP21:%.*]] = load i8, ptr [[AND_VAR1]], align 1
1247 // CHECK1-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP21]] to i1
1248 // CHECK1-NEXT:    br label [[LAND_END17]]
1249 // CHECK1:       land.end17:
1250 // CHECK1-NEXT:    [[TMP22:%.*]] = phi i1 [ false, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LAND_RHS15]] ]
1251 // CHECK1-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP22]] to i8
1252 // CHECK1-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
1253 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
1254 // CHECK1-NEXT:    [[TMP24:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP19]], i8 [[TMP23]] monotonic monotonic, align 1
1255 // CHECK1-NEXT:    [[TMP25]] = extractvalue { i8, i1 } [[TMP24]], 0
1256 // CHECK1-NEXT:    [[TMP26:%.*]] = extractvalue { i8, i1 } [[TMP24]], 1
1257 // CHECK1-NEXT:    br i1 [[TMP26]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
1258 // CHECK1:       atomic_exit:
1259 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1260 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1261 // CHECK1:       .omp.reduction.default:
1262 // CHECK1-NEXT:    ret void
1263 //
1264 //
1265 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined.omp.reduction.reduction_func
1266 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
1267 // CHECK1-NEXT:  entry:
1268 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
1269 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
1270 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
1271 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
1272 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
1273 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
1274 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
1275 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
1276 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
1277 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
1278 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
1279 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
1280 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
1281 // CHECK1:       land.rhs:
1282 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
1283 // CHECK1-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
1284 // CHECK1-NEXT:    br label [[LAND_END]]
1285 // CHECK1:       land.end:
1286 // CHECK1-NEXT:    [[TMP10:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LAND_RHS]] ]
1287 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
1288 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
1289 // CHECK1-NEXT:    ret void
1290 //
1291 //
1292 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232
1293 // CHECK1-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]]) #[[ATTR1]] {
1294 // CHECK1-NEXT:  entry:
1295 // CHECK1-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 8
1296 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 8
1297 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 8
1298 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined, ptr [[TMP0]])
1299 // CHECK1-NEXT:    ret void
1300 //
1301 //
1302 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined
1303 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]]) #[[ATTR1]] {
1304 // CHECK1-NEXT:  entry:
1305 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1306 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1307 // CHECK1-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 8
1308 // CHECK1-NEXT:    [[OR_VAR1:%.*]] = alloca i8, align 1
1309 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1310 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1311 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1312 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1313 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1314 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1315 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1316 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1317 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
1318 // CHECK1-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
1319 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1320 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1321 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 8
1322 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 8
1323 // CHECK1-NEXT:    store i8 0, ptr [[OR_VAR1]], align 1
1324 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1325 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1326 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1327 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1328 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1329 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1330 // 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)
1331 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1332 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
1333 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1334 // CHECK1:       cond.true:
1335 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1336 // CHECK1:       cond.false:
1337 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1338 // CHECK1-NEXT:    br label [[COND_END]]
1339 // CHECK1:       cond.end:
1340 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1341 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1342 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1343 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1344 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1345 // CHECK1:       omp.inner.for.cond:
1346 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1347 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1348 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1349 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1350 // CHECK1:       omp.inner.for.body:
1351 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1352 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1353 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1354 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1355 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[OR_VAR1]], align 1
1356 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP9]] to i1
1357 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
1358 // CHECK1:       lor.rhs:
1359 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
1360 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP10]], 0
1361 // CHECK1-NEXT:    br label [[LOR_END]]
1362 // CHECK1:       lor.end:
1363 // CHECK1-NEXT:    [[TMP11:%.*]] = phi i1 [ true, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LOR_RHS]] ]
1364 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP11]] to i8
1365 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[OR_VAR1]], align 1
1366 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1367 // CHECK1:       omp.body.continue:
1368 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1369 // CHECK1:       omp.inner.for.inc:
1370 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1371 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
1372 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
1373 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1374 // CHECK1:       omp.inner.for.end:
1375 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1376 // CHECK1:       omp.loop.exit:
1377 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1378 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1379 // CHECK1-NEXT:    store ptr [[OR_VAR1]], ptr [[TMP13]], align 8
1380 // CHECK1-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
1381 // CHECK1-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1382 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1383 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1384 // CHECK1-NEXT:    ]
1385 // CHECK1:       .omp.reduction.case1:
1386 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8, ptr [[TMP0]], align 1
1387 // CHECK1-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP15]] to i1
1388 // CHECK1-NEXT:    br i1 [[TOBOOL5]], label [[LOR_END8:%.*]], label [[LOR_RHS6:%.*]]
1389 // CHECK1:       lor.rhs6:
1390 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8, ptr [[OR_VAR1]], align 1
1391 // CHECK1-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP16]] to i1
1392 // CHECK1-NEXT:    br label [[LOR_END8]]
1393 // CHECK1:       lor.end8:
1394 // CHECK1-NEXT:    [[TMP17:%.*]] = phi i1 [ true, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LOR_RHS6]] ]
1395 // CHECK1-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP17]] to i8
1396 // CHECK1-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
1397 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1398 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1399 // CHECK1:       .omp.reduction.case2:
1400 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[OR_VAR1]], align 1
1401 // CHECK1-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP18]] to i1
1402 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
1403 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
1404 // CHECK1:       atomic_cont:
1405 // CHECK1-NEXT:    [[TMP19:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP25:%.*]], [[LOR_END17:%.*]] ]
1406 // CHECK1-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP19]] to i1
1407 // CHECK1-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
1408 // CHECK1-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
1409 // CHECK1-NEXT:    [[TMP20:%.*]] = load i8, ptr [[_TMP12]], align 1
1410 // CHECK1-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP20]] to i1
1411 // CHECK1-NEXT:    br i1 [[TOBOOL14]], label [[LOR_END17]], label [[LOR_RHS15:%.*]]
1412 // CHECK1:       lor.rhs15:
1413 // CHECK1-NEXT:    [[TMP21:%.*]] = load i8, ptr [[OR_VAR1]], align 1
1414 // CHECK1-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP21]] to i1
1415 // CHECK1-NEXT:    br label [[LOR_END17]]
1416 // CHECK1:       lor.end17:
1417 // CHECK1-NEXT:    [[TMP22:%.*]] = phi i1 [ true, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LOR_RHS15]] ]
1418 // CHECK1-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP22]] to i8
1419 // CHECK1-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
1420 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
1421 // CHECK1-NEXT:    [[TMP24:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP19]], i8 [[TMP23]] monotonic monotonic, align 1
1422 // CHECK1-NEXT:    [[TMP25]] = extractvalue { i8, i1 } [[TMP24]], 0
1423 // CHECK1-NEXT:    [[TMP26:%.*]] = extractvalue { i8, i1 } [[TMP24]], 1
1424 // CHECK1-NEXT:    br i1 [[TMP26]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
1425 // CHECK1:       atomic_exit:
1426 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1427 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1428 // CHECK1:       .omp.reduction.default:
1429 // CHECK1-NEXT:    ret void
1430 //
1431 //
1432 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined.omp.reduction.reduction_func
1433 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
1434 // CHECK1-NEXT:  entry:
1435 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
1436 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
1437 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
1438 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
1439 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
1440 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
1441 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
1442 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
1443 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
1444 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
1445 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
1446 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
1447 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
1448 // CHECK1:       lor.rhs:
1449 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
1450 // CHECK1-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
1451 // CHECK1-NEXT:    br label [[LOR_END]]
1452 // CHECK1:       lor.end:
1453 // CHECK1-NEXT:    [[TMP10:%.*]] = phi i1 [ true, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LOR_RHS]] ]
1454 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
1455 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
1456 // CHECK1-NEXT:    ret void
1457 //
1458 //
1459 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238
1460 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
1461 // CHECK1-NEXT:  entry:
1462 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
1463 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
1464 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
1465 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined, ptr [[TMP0]])
1466 // CHECK1-NEXT:    ret void
1467 //
1468 //
1469 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined
1470 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
1471 // CHECK1-NEXT:  entry:
1472 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1473 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1474 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
1475 // CHECK1-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
1476 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1477 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1478 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1479 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1480 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1481 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1482 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1483 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1484 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1485 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1486 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
1487 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
1488 // CHECK1-NEXT:    store i32 -1, ptr [[BIT_VAR1]], align 4
1489 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1490 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1491 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1492 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1493 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1494 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1495 // 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)
1496 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1497 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
1498 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1499 // CHECK1:       cond.true:
1500 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1501 // CHECK1:       cond.false:
1502 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1503 // CHECK1-NEXT:    br label [[COND_END]]
1504 // CHECK1:       cond.end:
1505 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1506 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1507 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1508 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1509 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1510 // CHECK1:       omp.inner.for.cond:
1511 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1512 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1513 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1514 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1515 // CHECK1:       omp.inner.for.body:
1516 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1517 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1518 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1519 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1520 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1521 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
1522 // CHECK1-NEXT:    [[AND:%.*]] = and i32 [[TMP9]], [[TMP10]]
1523 // CHECK1-NEXT:    store i32 [[AND]], ptr [[BIT_VAR1]], align 4
1524 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1525 // CHECK1:       omp.body.continue:
1526 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1527 // CHECK1:       omp.inner.for.inc:
1528 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1529 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
1530 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1531 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1532 // CHECK1:       omp.inner.for.end:
1533 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1534 // CHECK1:       omp.loop.exit:
1535 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1536 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1537 // CHECK1-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP12]], align 8
1538 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
1539 // CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1540 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1541 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1542 // CHECK1-NEXT:    ]
1543 // CHECK1:       .omp.reduction.case1:
1544 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
1545 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1546 // CHECK1-NEXT:    [[AND4:%.*]] = and i32 [[TMP14]], [[TMP15]]
1547 // CHECK1-NEXT:    store i32 [[AND4]], ptr [[TMP0]], align 4
1548 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1549 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1550 // CHECK1:       .omp.reduction.case2:
1551 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1552 // CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw and ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
1553 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1554 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1555 // CHECK1:       .omp.reduction.default:
1556 // CHECK1-NEXT:    ret void
1557 //
1558 //
1559 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined.omp.reduction.reduction_func
1560 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
1561 // CHECK1-NEXT:  entry:
1562 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
1563 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
1564 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
1565 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
1566 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
1567 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
1568 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
1569 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
1570 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
1571 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
1572 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1573 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
1574 // CHECK1-NEXT:    [[AND:%.*]] = and i32 [[TMP8]], [[TMP9]]
1575 // CHECK1-NEXT:    store i32 [[AND]], ptr [[TMP7]], align 4
1576 // CHECK1-NEXT:    ret void
1577 //
1578 //
1579 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243
1580 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
1581 // CHECK1-NEXT:  entry:
1582 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
1583 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
1584 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
1585 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined, ptr [[TMP0]])
1586 // CHECK1-NEXT:    ret void
1587 //
1588 //
1589 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined
1590 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
1591 // CHECK1-NEXT:  entry:
1592 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1593 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1594 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
1595 // CHECK1-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
1596 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1597 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1598 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1599 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1600 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1601 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1602 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1603 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1604 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1605 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1606 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
1607 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
1608 // CHECK1-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
1609 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1610 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1611 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1612 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1613 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1614 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1615 // 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)
1616 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1617 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
1618 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1619 // CHECK1:       cond.true:
1620 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1621 // CHECK1:       cond.false:
1622 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1623 // CHECK1-NEXT:    br label [[COND_END]]
1624 // CHECK1:       cond.end:
1625 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1626 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1627 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1628 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1629 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1630 // CHECK1:       omp.inner.for.cond:
1631 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1632 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1633 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1634 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1635 // CHECK1:       omp.inner.for.body:
1636 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1637 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1638 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1639 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1640 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1641 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
1642 // CHECK1-NEXT:    [[OR:%.*]] = or i32 [[TMP9]], [[TMP10]]
1643 // CHECK1-NEXT:    store i32 [[OR]], ptr [[BIT_VAR1]], align 4
1644 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1645 // CHECK1:       omp.body.continue:
1646 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1647 // CHECK1:       omp.inner.for.inc:
1648 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1649 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
1650 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1651 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1652 // CHECK1:       omp.inner.for.end:
1653 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1654 // CHECK1:       omp.loop.exit:
1655 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1656 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1657 // CHECK1-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP12]], align 8
1658 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
1659 // CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1660 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1661 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1662 // CHECK1-NEXT:    ]
1663 // CHECK1:       .omp.reduction.case1:
1664 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
1665 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1666 // CHECK1-NEXT:    [[OR4:%.*]] = or i32 [[TMP14]], [[TMP15]]
1667 // CHECK1-NEXT:    store i32 [[OR4]], ptr [[TMP0]], align 4
1668 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1669 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1670 // CHECK1:       .omp.reduction.case2:
1671 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1672 // CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw or ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
1673 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1674 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1675 // CHECK1:       .omp.reduction.default:
1676 // CHECK1-NEXT:    ret void
1677 //
1678 //
1679 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined.omp.reduction.reduction_func
1680 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
1681 // CHECK1-NEXT:  entry:
1682 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
1683 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
1684 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
1685 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
1686 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
1687 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
1688 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
1689 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
1690 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
1691 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
1692 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1693 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
1694 // CHECK1-NEXT:    [[OR:%.*]] = or i32 [[TMP8]], [[TMP9]]
1695 // CHECK1-NEXT:    store i32 [[OR]], ptr [[TMP7]], align 4
1696 // CHECK1-NEXT:    ret void
1697 //
1698 //
1699 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248
1700 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
1701 // CHECK1-NEXT:  entry:
1702 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
1703 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
1704 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
1705 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined, ptr [[TMP0]])
1706 // CHECK1-NEXT:    ret void
1707 //
1708 //
1709 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined
1710 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
1711 // CHECK1-NEXT:  entry:
1712 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1713 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1714 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
1715 // CHECK1-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
1716 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1717 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1718 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1719 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1720 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1721 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1722 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1723 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1724 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1725 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1726 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
1727 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
1728 // CHECK1-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
1729 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1730 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1731 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1732 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1733 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1734 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1735 // 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)
1736 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1737 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
1738 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1739 // CHECK1:       cond.true:
1740 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1741 // CHECK1:       cond.false:
1742 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1743 // CHECK1-NEXT:    br label [[COND_END]]
1744 // CHECK1:       cond.end:
1745 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1746 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1747 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1748 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1749 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1750 // CHECK1:       omp.inner.for.cond:
1751 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1752 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1753 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1754 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1755 // CHECK1:       omp.inner.for.body:
1756 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1757 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1758 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1759 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1760 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1761 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
1762 // CHECK1-NEXT:    [[XOR:%.*]] = xor i32 [[TMP9]], [[TMP10]]
1763 // CHECK1-NEXT:    store i32 [[XOR]], ptr [[BIT_VAR1]], align 4
1764 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1765 // CHECK1:       omp.body.continue:
1766 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1767 // CHECK1:       omp.inner.for.inc:
1768 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1769 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
1770 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
1771 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1772 // CHECK1:       omp.inner.for.end:
1773 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1774 // CHECK1:       omp.loop.exit:
1775 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1776 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1777 // CHECK1-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP12]], align 8
1778 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
1779 // CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1780 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1781 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1782 // CHECK1-NEXT:    ]
1783 // CHECK1:       .omp.reduction.case1:
1784 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
1785 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1786 // CHECK1-NEXT:    [[XOR4:%.*]] = xor i32 [[TMP14]], [[TMP15]]
1787 // CHECK1-NEXT:    store i32 [[XOR4]], ptr [[TMP0]], align 4
1788 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1789 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1790 // CHECK1:       .omp.reduction.case2:
1791 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
1792 // CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw xor ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
1793 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1794 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1795 // CHECK1:       .omp.reduction.default:
1796 // CHECK1-NEXT:    ret void
1797 //
1798 //
1799 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined.omp.reduction.reduction_func
1800 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
1801 // CHECK1-NEXT:  entry:
1802 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
1803 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
1804 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
1805 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
1806 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
1807 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
1808 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
1809 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
1810 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
1811 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
1812 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1813 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
1814 // CHECK1-NEXT:    [[XOR:%.*]] = xor i32 [[TMP8]], [[TMP9]]
1815 // CHECK1-NEXT:    store i32 [[XOR]], ptr [[TMP7]], align 4
1816 // CHECK1-NEXT:    ret void
1817 //
1818 //
1819 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254
1820 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
1821 // CHECK1-NEXT:  entry:
1822 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
1823 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
1824 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
1825 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined, ptr [[TMP0]])
1826 // CHECK1-NEXT:    ret void
1827 //
1828 //
1829 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined
1830 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
1831 // CHECK1-NEXT:  entry:
1832 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1833 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1834 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
1835 // CHECK1-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
1836 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1837 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1838 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1839 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1840 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1841 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1842 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1843 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1844 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1845 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1846 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
1847 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
1848 // CHECK1-NEXT:    store i32 -2147483648, ptr [[SIVAR1]], align 4
1849 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1850 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1851 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1852 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1853 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1854 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1855 // 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)
1856 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1857 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
1858 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1859 // CHECK1:       cond.true:
1860 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1861 // CHECK1:       cond.false:
1862 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1863 // CHECK1-NEXT:    br label [[COND_END]]
1864 // CHECK1:       cond.end:
1865 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1866 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1867 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1868 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1869 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1870 // CHECK1:       omp.inner.for.cond:
1871 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1872 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1873 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1874 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1875 // CHECK1:       omp.inner.for.body:
1876 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1877 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1878 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1879 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1880 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[SIVAR1]], align 4
1881 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
1882 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sge i32 [[TMP9]], [[TMP10]]
1883 // CHECK1-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
1884 // CHECK1:       cond.true4:
1885 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[SIVAR1]], align 4
1886 // CHECK1-NEXT:    br label [[COND_END6:%.*]]
1887 // CHECK1:       cond.false5:
1888 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
1889 // CHECK1-NEXT:    br label [[COND_END6]]
1890 // CHECK1:       cond.end6:
1891 // CHECK1-NEXT:    [[COND7:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE4]] ], [ [[TMP12]], [[COND_FALSE5]] ]
1892 // CHECK1-NEXT:    store i32 [[COND7]], ptr [[SIVAR1]], align 4
1893 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1894 // CHECK1:       omp.body.continue:
1895 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1896 // CHECK1:       omp.inner.for.inc:
1897 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1898 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
1899 // CHECK1-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
1900 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1901 // CHECK1:       omp.inner.for.end:
1902 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1903 // CHECK1:       omp.loop.exit:
1904 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1905 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1906 // CHECK1-NEXT:    store ptr [[SIVAR1]], ptr [[TMP14]], align 8
1907 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
1908 // CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1909 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1910 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1911 // CHECK1-NEXT:    ]
1912 // CHECK1:       .omp.reduction.case1:
1913 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
1914 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
1915 // CHECK1-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP16]], [[TMP17]]
1916 // CHECK1-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
1917 // CHECK1:       cond.true10:
1918 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP0]], align 4
1919 // CHECK1-NEXT:    br label [[COND_END12:%.*]]
1920 // CHECK1:       cond.false11:
1921 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[SIVAR1]], align 4
1922 // CHECK1-NEXT:    br label [[COND_END12]]
1923 // CHECK1:       cond.end12:
1924 // CHECK1-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP18]], [[COND_TRUE10]] ], [ [[TMP19]], [[COND_FALSE11]] ]
1925 // CHECK1-NEXT:    store i32 [[COND13]], ptr [[TMP0]], align 4
1926 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1927 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1928 // CHECK1:       .omp.reduction.case2:
1929 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[SIVAR1]], align 4
1930 // CHECK1-NEXT:    [[TMP21:%.*]] = atomicrmw max ptr [[TMP0]], i32 [[TMP20]] monotonic, align 4
1931 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
1932 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
1933 // CHECK1:       .omp.reduction.default:
1934 // CHECK1-NEXT:    ret void
1935 //
1936 //
1937 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined.omp.reduction.reduction_func
1938 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
1939 // CHECK1-NEXT:  entry:
1940 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
1941 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
1942 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
1943 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
1944 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
1945 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
1946 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
1947 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
1948 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
1949 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
1950 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1951 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
1952 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1953 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1954 // CHECK1:       cond.true:
1955 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
1956 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1957 // CHECK1:       cond.false:
1958 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
1959 // CHECK1-NEXT:    br label [[COND_END]]
1960 // CHECK1:       cond.end:
1961 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1962 // CHECK1-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
1963 // CHECK1-NEXT:    ret void
1964 //
1965 //
1966 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260
1967 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
1968 // CHECK1-NEXT:  entry:
1969 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
1970 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
1971 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
1972 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined, ptr [[TMP0]])
1973 // CHECK1-NEXT:    ret void
1974 //
1975 //
1976 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined
1977 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
1978 // CHECK1-NEXT:  entry:
1979 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1980 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1981 // CHECK1-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
1982 // CHECK1-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
1983 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1984 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1985 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1986 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1987 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1988 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1989 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1990 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
1991 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1992 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1993 // CHECK1-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
1994 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
1995 // CHECK1-NEXT:    store i32 2147483647, ptr [[SIVAR1]], align 4
1996 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1997 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1998 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1999 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2000 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2001 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2002 // 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)
2003 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2004 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
2005 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2006 // CHECK1:       cond.true:
2007 // CHECK1-NEXT:    br label [[COND_END:%.*]]
2008 // CHECK1:       cond.false:
2009 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2010 // CHECK1-NEXT:    br label [[COND_END]]
2011 // CHECK1:       cond.end:
2012 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2013 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2014 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2015 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2016 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2017 // CHECK1:       omp.inner.for.cond:
2018 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2019 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2020 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2021 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2022 // CHECK1:       omp.inner.for.body:
2023 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2024 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2025 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2026 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2027 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[SIVAR1]], align 4
2028 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
2029 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
2030 // CHECK1-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
2031 // CHECK1:       cond.true4:
2032 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[SIVAR1]], align 4
2033 // CHECK1-NEXT:    br label [[COND_END6:%.*]]
2034 // CHECK1:       cond.false5:
2035 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
2036 // CHECK1-NEXT:    br label [[COND_END6]]
2037 // CHECK1:       cond.end6:
2038 // CHECK1-NEXT:    [[COND7:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE4]] ], [ [[TMP12]], [[COND_FALSE5]] ]
2039 // CHECK1-NEXT:    store i32 [[COND7]], ptr [[SIVAR1]], align 4
2040 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2041 // CHECK1:       omp.body.continue:
2042 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2043 // CHECK1:       omp.inner.for.inc:
2044 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2045 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
2046 // CHECK1-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
2047 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
2048 // CHECK1:       omp.inner.for.end:
2049 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2050 // CHECK1:       omp.loop.exit:
2051 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
2052 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2053 // CHECK1-NEXT:    store ptr [[SIVAR1]], ptr [[TMP14]], align 8
2054 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
2055 // CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2056 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2057 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2058 // CHECK1-NEXT:    ]
2059 // CHECK1:       .omp.reduction.case1:
2060 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
2061 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
2062 // CHECK1-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP16]], [[TMP17]]
2063 // CHECK1-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
2064 // CHECK1:       cond.true10:
2065 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP0]], align 4
2066 // CHECK1-NEXT:    br label [[COND_END12:%.*]]
2067 // CHECK1:       cond.false11:
2068 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[SIVAR1]], align 4
2069 // CHECK1-NEXT:    br label [[COND_END12]]
2070 // CHECK1:       cond.end12:
2071 // CHECK1-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP18]], [[COND_TRUE10]] ], [ [[TMP19]], [[COND_FALSE11]] ]
2072 // CHECK1-NEXT:    store i32 [[COND13]], ptr [[TMP0]], align 4
2073 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
2074 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2075 // CHECK1:       .omp.reduction.case2:
2076 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[SIVAR1]], align 4
2077 // CHECK1-NEXT:    [[TMP21:%.*]] = atomicrmw min ptr [[TMP0]], i32 [[TMP20]] monotonic, align 4
2078 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
2079 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2080 // CHECK1:       .omp.reduction.default:
2081 // CHECK1-NEXT:    ret void
2082 //
2083 //
2084 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined.omp.reduction.reduction_func
2085 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
2086 // CHECK1-NEXT:  entry:
2087 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
2088 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
2089 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
2090 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
2091 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
2092 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
2093 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
2094 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
2095 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
2096 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
2097 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2098 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
2099 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], [[TMP9]]
2100 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2101 // CHECK1:       cond.true:
2102 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
2103 // CHECK1-NEXT:    br label [[COND_END:%.*]]
2104 // CHECK1:       cond.false:
2105 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
2106 // CHECK1-NEXT:    br label [[COND_END]]
2107 // CHECK1:       cond.end:
2108 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
2109 // CHECK1-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
2110 // CHECK1-NEXT:    ret void
2111 //
2112 //
2113 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
2114 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] comdat {
2115 // CHECK1-NEXT:  entry:
2116 // CHECK1-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
2117 // CHECK1-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
2118 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
2119 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
2120 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
2121 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2122 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2123 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [2 x ptr], align 8
2124 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [2 x ptr], align 8
2125 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [2 x ptr], align 8
2126 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
2127 // CHECK1-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2128 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x ptr], align 8
2129 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x ptr], align 8
2130 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x ptr], align 8
2131 // CHECK1-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
2132 // CHECK1-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2133 // CHECK1-NEXT:    [[AND_VAR:%.*]] = alloca i8, align 1
2134 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [2 x ptr], align 8
2135 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [2 x ptr], align 8
2136 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [2 x ptr], align 8
2137 // CHECK1-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
2138 // CHECK1-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2139 // CHECK1-NEXT:    [[OR_VAR:%.*]] = alloca i8, align 1
2140 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [2 x ptr], align 8
2141 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS23:%.*]] = alloca [2 x ptr], align 8
2142 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [2 x ptr], align 8
2143 // CHECK1-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
2144 // CHECK1-NEXT:    [[KERNEL_ARGS26:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2145 // CHECK1-NEXT:    [[BIT_VAR:%.*]] = alloca i32, align 4
2146 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [2 x ptr], align 8
2147 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [2 x ptr], align 8
2148 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [2 x ptr], align 8
2149 // CHECK1-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
2150 // CHECK1-NEXT:    [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2151 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [2 x ptr], align 8
2152 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [2 x ptr], align 8
2153 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [2 x ptr], align 8
2154 // CHECK1-NEXT:    [[_TMP39:%.*]] = alloca i32, align 4
2155 // CHECK1-NEXT:    [[KERNEL_ARGS40:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2156 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS43:%.*]] = alloca [2 x ptr], align 8
2157 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS44:%.*]] = alloca [2 x ptr], align 8
2158 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS45:%.*]] = alloca [2 x ptr], align 8
2159 // CHECK1-NEXT:    [[_TMP46:%.*]] = alloca i32, align 4
2160 // CHECK1-NEXT:    [[KERNEL_ARGS47:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2161 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS50:%.*]] = alloca [2 x ptr], align 8
2162 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS51:%.*]] = alloca [2 x ptr], align 8
2163 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS52:%.*]] = alloca [2 x ptr], align 8
2164 // CHECK1-NEXT:    [[_TMP53:%.*]] = alloca i32, align 4
2165 // CHECK1-NEXT:    [[KERNEL_ARGS54:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2166 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS57:%.*]] = alloca [2 x ptr], align 8
2167 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS58:%.*]] = alloca [2 x ptr], align 8
2168 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS59:%.*]] = alloca [2 x ptr], align 8
2169 // CHECK1-NEXT:    [[_TMP60:%.*]] = alloca i32, align 4
2170 // CHECK1-NEXT:    [[KERNEL_ARGS61:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2171 // CHECK1-NEXT:    store i32 0, ptr [[T_VAR]], align 4
2172 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i64 8, i1 false)
2173 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2174 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP0]], align 8
2175 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2176 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP1]], align 8
2177 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2178 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
2179 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2180 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2181 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2182 // CHECK1-NEXT:    store i32 2, ptr [[TMP5]], align 4
2183 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2184 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
2185 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2186 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
2187 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2188 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
2189 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2190 // CHECK1-NEXT:    store ptr @.offload_sizes.19, ptr [[TMP9]], align 8
2191 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2192 // CHECK1-NEXT:    store ptr @.offload_maptypes.20, ptr [[TMP10]], align 8
2193 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2194 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
2195 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2196 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
2197 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2198 // CHECK1-NEXT:    store i64 2, ptr [[TMP13]], align 8
2199 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2200 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
2201 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2202 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
2203 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2204 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
2205 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2206 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
2207 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, ptr [[KERNEL_ARGS]])
2208 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2209 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2210 // CHECK1:       omp_offload.failed:
2211 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(ptr [[T_VAR]]) #[[ATTR2]]
2212 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2213 // CHECK1:       omp_offload.cont:
2214 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2215 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP20]], align 8
2216 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2217 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP21]], align 8
2218 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
2219 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
2220 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 1
2221 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP23]], align 8
2222 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 1
2223 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP24]], align 8
2224 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 1
2225 // CHECK1-NEXT:    store ptr null, ptr [[TMP25]], align 8
2226 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2227 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2228 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
2229 // CHECK1-NEXT:    store i32 2, ptr [[TMP28]], align 4
2230 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
2231 // CHECK1-NEXT:    store i32 2, ptr [[TMP29]], align 4
2232 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
2233 // CHECK1-NEXT:    store ptr [[TMP26]], ptr [[TMP30]], align 8
2234 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
2235 // CHECK1-NEXT:    store ptr [[TMP27]], ptr [[TMP31]], align 8
2236 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
2237 // CHECK1-NEXT:    store ptr @.offload_sizes.21, ptr [[TMP32]], align 8
2238 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
2239 // CHECK1-NEXT:    store ptr @.offload_maptypes.22, ptr [[TMP33]], align 8
2240 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
2241 // CHECK1-NEXT:    store ptr null, ptr [[TMP34]], align 8
2242 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
2243 // CHECK1-NEXT:    store ptr null, ptr [[TMP35]], align 8
2244 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
2245 // CHECK1-NEXT:    store i64 2, ptr [[TMP36]], align 8
2246 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
2247 // CHECK1-NEXT:    store i64 0, ptr [[TMP37]], align 8
2248 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
2249 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP38]], align 4
2250 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
2251 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP39]], align 4
2252 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
2253 // CHECK1-NEXT:    store i32 0, ptr [[TMP40]], align 4
2254 // CHECK1-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.region_id, ptr [[KERNEL_ARGS5]])
2255 // CHECK1-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
2256 // CHECK1-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
2257 // CHECK1:       omp_offload.failed6:
2258 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
2259 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
2260 // CHECK1:       omp_offload.cont7:
2261 // CHECK1-NEXT:    store i32 1, ptr [[T_VAR]], align 4
2262 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2263 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP43]], align 8
2264 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2265 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP44]], align 8
2266 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
2267 // CHECK1-NEXT:    store ptr null, ptr [[TMP45]], align 8
2268 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
2269 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP46]], align 8
2270 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
2271 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP47]], align 8
2272 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
2273 // CHECK1-NEXT:    store ptr null, ptr [[TMP48]], align 8
2274 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2275 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2276 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
2277 // CHECK1-NEXT:    store i32 2, ptr [[TMP51]], align 4
2278 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
2279 // CHECK1-NEXT:    store i32 2, ptr [[TMP52]], align 4
2280 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
2281 // CHECK1-NEXT:    store ptr [[TMP49]], ptr [[TMP53]], align 8
2282 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
2283 // CHECK1-NEXT:    store ptr [[TMP50]], ptr [[TMP54]], align 8
2284 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
2285 // CHECK1-NEXT:    store ptr @.offload_sizes.23, ptr [[TMP55]], align 8
2286 // CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
2287 // CHECK1-NEXT:    store ptr @.offload_maptypes.24, ptr [[TMP56]], align 8
2288 // CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
2289 // CHECK1-NEXT:    store ptr null, ptr [[TMP57]], align 8
2290 // CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
2291 // CHECK1-NEXT:    store ptr null, ptr [[TMP58]], align 8
2292 // CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
2293 // CHECK1-NEXT:    store i64 2, ptr [[TMP59]], align 8
2294 // CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
2295 // CHECK1-NEXT:    store i64 0, ptr [[TMP60]], align 8
2296 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
2297 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP61]], align 4
2298 // CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
2299 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP62]], align 4
2300 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
2301 // CHECK1-NEXT:    store i32 0, ptr [[TMP63]], align 4
2302 // CHECK1-NEXT:    [[TMP64:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.region_id, ptr [[KERNEL_ARGS12]])
2303 // CHECK1-NEXT:    [[TMP65:%.*]] = icmp ne i32 [[TMP64]], 0
2304 // CHECK1-NEXT:    br i1 [[TMP65]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
2305 // CHECK1:       omp_offload.failed13:
2306 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
2307 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
2308 // CHECK1:       omp_offload.cont14:
2309 // CHECK1-NEXT:    store i8 1, ptr [[AND_VAR]], align 1
2310 // CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
2311 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[TMP66]], align 8
2312 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
2313 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[TMP67]], align 8
2314 // CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS17]], i64 0, i64 0
2315 // CHECK1-NEXT:    store ptr null, ptr [[TMP68]], align 8
2316 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 1
2317 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP69]], align 8
2318 // CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 1
2319 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP70]], align 8
2320 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS17]], i64 0, i64 1
2321 // CHECK1-NEXT:    store ptr null, ptr [[TMP71]], align 8
2322 // CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
2323 // CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
2324 // CHECK1-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 0
2325 // CHECK1-NEXT:    store i32 2, ptr [[TMP74]], align 4
2326 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 1
2327 // CHECK1-NEXT:    store i32 2, ptr [[TMP75]], align 4
2328 // CHECK1-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 2
2329 // CHECK1-NEXT:    store ptr [[TMP72]], ptr [[TMP76]], align 8
2330 // CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 3
2331 // CHECK1-NEXT:    store ptr [[TMP73]], ptr [[TMP77]], align 8
2332 // CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 4
2333 // CHECK1-NEXT:    store ptr @.offload_sizes.25, ptr [[TMP78]], align 8
2334 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 5
2335 // CHECK1-NEXT:    store ptr @.offload_maptypes.26, ptr [[TMP79]], align 8
2336 // CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 6
2337 // CHECK1-NEXT:    store ptr null, ptr [[TMP80]], align 8
2338 // CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 7
2339 // CHECK1-NEXT:    store ptr null, ptr [[TMP81]], align 8
2340 // CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 8
2341 // CHECK1-NEXT:    store i64 2, ptr [[TMP82]], align 8
2342 // CHECK1-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 9
2343 // CHECK1-NEXT:    store i64 0, ptr [[TMP83]], align 8
2344 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 10
2345 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP84]], align 4
2346 // CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 11
2347 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP85]], align 4
2348 // CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 12
2349 // CHECK1-NEXT:    store i32 0, ptr [[TMP86]], align 4
2350 // CHECK1-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.region_id, ptr [[KERNEL_ARGS19]])
2351 // CHECK1-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
2352 // CHECK1-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
2353 // CHECK1:       omp_offload.failed20:
2354 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49(ptr [[AND_VAR]], ptr [[VEC]]) #[[ATTR2]]
2355 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
2356 // CHECK1:       omp_offload.cont21:
2357 // CHECK1-NEXT:    store i8 0, ptr [[OR_VAR]], align 1
2358 // CHECK1-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
2359 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[TMP89]], align 8
2360 // CHECK1-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
2361 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[TMP90]], align 8
2362 // CHECK1-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 0
2363 // CHECK1-NEXT:    store ptr null, ptr [[TMP91]], align 8
2364 // CHECK1-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 1
2365 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP92]], align 8
2366 // CHECK1-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 1
2367 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP93]], align 8
2368 // CHECK1-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 1
2369 // CHECK1-NEXT:    store ptr null, ptr [[TMP94]], align 8
2370 // CHECK1-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
2371 // CHECK1-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
2372 // CHECK1-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 0
2373 // CHECK1-NEXT:    store i32 2, ptr [[TMP97]], align 4
2374 // CHECK1-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 1
2375 // CHECK1-NEXT:    store i32 2, ptr [[TMP98]], align 4
2376 // CHECK1-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 2
2377 // CHECK1-NEXT:    store ptr [[TMP95]], ptr [[TMP99]], align 8
2378 // CHECK1-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 3
2379 // CHECK1-NEXT:    store ptr [[TMP96]], ptr [[TMP100]], align 8
2380 // CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 4
2381 // CHECK1-NEXT:    store ptr @.offload_sizes.27, ptr [[TMP101]], align 8
2382 // CHECK1-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 5
2383 // CHECK1-NEXT:    store ptr @.offload_maptypes.28, ptr [[TMP102]], align 8
2384 // CHECK1-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 6
2385 // CHECK1-NEXT:    store ptr null, ptr [[TMP103]], align 8
2386 // CHECK1-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 7
2387 // CHECK1-NEXT:    store ptr null, ptr [[TMP104]], align 8
2388 // CHECK1-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 8
2389 // CHECK1-NEXT:    store i64 2, ptr [[TMP105]], align 8
2390 // CHECK1-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 9
2391 // CHECK1-NEXT:    store i64 0, ptr [[TMP106]], align 8
2392 // CHECK1-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 10
2393 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP107]], align 4
2394 // CHECK1-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 11
2395 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP108]], align 4
2396 // CHECK1-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 12
2397 // CHECK1-NEXT:    store i32 0, ptr [[TMP109]], align 4
2398 // CHECK1-NEXT:    [[TMP110:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.region_id, ptr [[KERNEL_ARGS26]])
2399 // CHECK1-NEXT:    [[TMP111:%.*]] = icmp ne i32 [[TMP110]], 0
2400 // CHECK1-NEXT:    br i1 [[TMP111]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
2401 // CHECK1:       omp_offload.failed27:
2402 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55(ptr [[OR_VAR]], ptr [[VEC]]) #[[ATTR2]]
2403 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
2404 // CHECK1:       omp_offload.cont28:
2405 // CHECK1-NEXT:    store i32 1, ptr [[BIT_VAR]], align 4
2406 // CHECK1-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
2407 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP112]], align 8
2408 // CHECK1-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
2409 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP113]], align 8
2410 // CHECK1-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 0
2411 // CHECK1-NEXT:    store ptr null, ptr [[TMP114]], align 8
2412 // CHECK1-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 1
2413 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP115]], align 8
2414 // CHECK1-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 1
2415 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP116]], align 8
2416 // CHECK1-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 1
2417 // CHECK1-NEXT:    store ptr null, ptr [[TMP117]], align 8
2418 // CHECK1-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
2419 // CHECK1-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
2420 // CHECK1-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 0
2421 // CHECK1-NEXT:    store i32 2, ptr [[TMP120]], align 4
2422 // CHECK1-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 1
2423 // CHECK1-NEXT:    store i32 2, ptr [[TMP121]], align 4
2424 // CHECK1-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 2
2425 // CHECK1-NEXT:    store ptr [[TMP118]], ptr [[TMP122]], align 8
2426 // CHECK1-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 3
2427 // CHECK1-NEXT:    store ptr [[TMP119]], ptr [[TMP123]], align 8
2428 // CHECK1-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 4
2429 // CHECK1-NEXT:    store ptr @.offload_sizes.29, ptr [[TMP124]], align 8
2430 // CHECK1-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 5
2431 // CHECK1-NEXT:    store ptr @.offload_maptypes.30, ptr [[TMP125]], align 8
2432 // CHECK1-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 6
2433 // CHECK1-NEXT:    store ptr null, ptr [[TMP126]], align 8
2434 // CHECK1-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 7
2435 // CHECK1-NEXT:    store ptr null, ptr [[TMP127]], align 8
2436 // CHECK1-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 8
2437 // CHECK1-NEXT:    store i64 2, ptr [[TMP128]], align 8
2438 // CHECK1-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 9
2439 // CHECK1-NEXT:    store i64 0, ptr [[TMP129]], align 8
2440 // CHECK1-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 10
2441 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP130]], align 4
2442 // CHECK1-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 11
2443 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP131]], align 4
2444 // CHECK1-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 12
2445 // CHECK1-NEXT:    store i32 0, ptr [[TMP132]], align 4
2446 // CHECK1-NEXT:    [[TMP133:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.region_id, ptr [[KERNEL_ARGS33]])
2447 // CHECK1-NEXT:    [[TMP134:%.*]] = icmp ne i32 [[TMP133]], 0
2448 // CHECK1-NEXT:    br i1 [[TMP134]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
2449 // CHECK1:       omp_offload.failed34:
2450 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61(ptr [[BIT_VAR]], ptr [[VEC]]) #[[ATTR2]]
2451 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT35]]
2452 // CHECK1:       omp_offload.cont35:
2453 // CHECK1-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
2454 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP135]], align 8
2455 // CHECK1-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
2456 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP136]], align 8
2457 // CHECK1-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
2458 // CHECK1-NEXT:    store ptr null, ptr [[TMP137]], align 8
2459 // CHECK1-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
2460 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP138]], align 8
2461 // CHECK1-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
2462 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP139]], align 8
2463 // CHECK1-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
2464 // CHECK1-NEXT:    store ptr null, ptr [[TMP140]], align 8
2465 // CHECK1-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
2466 // CHECK1-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
2467 // CHECK1-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 0
2468 // CHECK1-NEXT:    store i32 2, ptr [[TMP143]], align 4
2469 // CHECK1-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 1
2470 // CHECK1-NEXT:    store i32 2, ptr [[TMP144]], align 4
2471 // CHECK1-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 2
2472 // CHECK1-NEXT:    store ptr [[TMP141]], ptr [[TMP145]], align 8
2473 // CHECK1-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 3
2474 // CHECK1-NEXT:    store ptr [[TMP142]], ptr [[TMP146]], align 8
2475 // CHECK1-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 4
2476 // CHECK1-NEXT:    store ptr @.offload_sizes.31, ptr [[TMP147]], align 8
2477 // CHECK1-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 5
2478 // CHECK1-NEXT:    store ptr @.offload_maptypes.32, ptr [[TMP148]], align 8
2479 // CHECK1-NEXT:    [[TMP149:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 6
2480 // CHECK1-NEXT:    store ptr null, ptr [[TMP149]], align 8
2481 // CHECK1-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 7
2482 // CHECK1-NEXT:    store ptr null, ptr [[TMP150]], align 8
2483 // CHECK1-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 8
2484 // CHECK1-NEXT:    store i64 2, ptr [[TMP151]], align 8
2485 // CHECK1-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 9
2486 // CHECK1-NEXT:    store i64 0, ptr [[TMP152]], align 8
2487 // CHECK1-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 10
2488 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP153]], align 4
2489 // CHECK1-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 11
2490 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP154]], align 4
2491 // CHECK1-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 12
2492 // CHECK1-NEXT:    store i32 0, ptr [[TMP155]], align 4
2493 // CHECK1-NEXT:    [[TMP156:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.region_id, ptr [[KERNEL_ARGS40]])
2494 // CHECK1-NEXT:    [[TMP157:%.*]] = icmp ne i32 [[TMP156]], 0
2495 // CHECK1-NEXT:    br i1 [[TMP157]], label [[OMP_OFFLOAD_FAILED41:%.*]], label [[OMP_OFFLOAD_CONT42:%.*]]
2496 // CHECK1:       omp_offload.failed41:
2497 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66(ptr [[BIT_VAR]], ptr [[VEC]]) #[[ATTR2]]
2498 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT42]]
2499 // CHECK1:       omp_offload.cont42:
2500 // CHECK1-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
2501 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP158]], align 8
2502 // CHECK1-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
2503 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP159]], align 8
2504 // CHECK1-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS45]], i64 0, i64 0
2505 // CHECK1-NEXT:    store ptr null, ptr [[TMP160]], align 8
2506 // CHECK1-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 1
2507 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP161]], align 8
2508 // CHECK1-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 1
2509 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP162]], align 8
2510 // CHECK1-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS45]], i64 0, i64 1
2511 // CHECK1-NEXT:    store ptr null, ptr [[TMP163]], align 8
2512 // CHECK1-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
2513 // CHECK1-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
2514 // CHECK1-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 0
2515 // CHECK1-NEXT:    store i32 2, ptr [[TMP166]], align 4
2516 // CHECK1-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 1
2517 // CHECK1-NEXT:    store i32 2, ptr [[TMP167]], align 4
2518 // CHECK1-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 2
2519 // CHECK1-NEXT:    store ptr [[TMP164]], ptr [[TMP168]], align 8
2520 // CHECK1-NEXT:    [[TMP169:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 3
2521 // CHECK1-NEXT:    store ptr [[TMP165]], ptr [[TMP169]], align 8
2522 // CHECK1-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 4
2523 // CHECK1-NEXT:    store ptr @.offload_sizes.33, ptr [[TMP170]], align 8
2524 // CHECK1-NEXT:    [[TMP171:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 5
2525 // CHECK1-NEXT:    store ptr @.offload_maptypes.34, ptr [[TMP171]], align 8
2526 // CHECK1-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 6
2527 // CHECK1-NEXT:    store ptr null, ptr [[TMP172]], align 8
2528 // CHECK1-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 7
2529 // CHECK1-NEXT:    store ptr null, ptr [[TMP173]], align 8
2530 // CHECK1-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 8
2531 // CHECK1-NEXT:    store i64 2, ptr [[TMP174]], align 8
2532 // CHECK1-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 9
2533 // CHECK1-NEXT:    store i64 0, ptr [[TMP175]], align 8
2534 // CHECK1-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 10
2535 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP176]], align 4
2536 // CHECK1-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 11
2537 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP177]], align 4
2538 // CHECK1-NEXT:    [[TMP178:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 12
2539 // CHECK1-NEXT:    store i32 0, ptr [[TMP178]], align 4
2540 // CHECK1-NEXT:    [[TMP179:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.region_id, ptr [[KERNEL_ARGS47]])
2541 // CHECK1-NEXT:    [[TMP180:%.*]] = icmp ne i32 [[TMP179]], 0
2542 // CHECK1-NEXT:    br i1 [[TMP180]], label [[OMP_OFFLOAD_FAILED48:%.*]], label [[OMP_OFFLOAD_CONT49:%.*]]
2543 // CHECK1:       omp_offload.failed48:
2544 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71(ptr [[BIT_VAR]], ptr [[VEC]]) #[[ATTR2]]
2545 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT49]]
2546 // CHECK1:       omp_offload.cont49:
2547 // CHECK1-NEXT:    store i32 0, ptr [[T_VAR]], align 4
2548 // CHECK1-NEXT:    [[TMP181:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
2549 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP181]], align 8
2550 // CHECK1-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
2551 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP182]], align 8
2552 // CHECK1-NEXT:    [[TMP183:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS52]], i64 0, i64 0
2553 // CHECK1-NEXT:    store ptr null, ptr [[TMP183]], align 8
2554 // CHECK1-NEXT:    [[TMP184:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 1
2555 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP184]], align 8
2556 // CHECK1-NEXT:    [[TMP185:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 1
2557 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP185]], align 8
2558 // CHECK1-NEXT:    [[TMP186:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS52]], i64 0, i64 1
2559 // CHECK1-NEXT:    store ptr null, ptr [[TMP186]], align 8
2560 // CHECK1-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
2561 // CHECK1-NEXT:    [[TMP188:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
2562 // CHECK1-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 0
2563 // CHECK1-NEXT:    store i32 2, ptr [[TMP189]], align 4
2564 // CHECK1-NEXT:    [[TMP190:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 1
2565 // CHECK1-NEXT:    store i32 2, ptr [[TMP190]], align 4
2566 // CHECK1-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 2
2567 // CHECK1-NEXT:    store ptr [[TMP187]], ptr [[TMP191]], align 8
2568 // CHECK1-NEXT:    [[TMP192:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 3
2569 // CHECK1-NEXT:    store ptr [[TMP188]], ptr [[TMP192]], align 8
2570 // CHECK1-NEXT:    [[TMP193:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 4
2571 // CHECK1-NEXT:    store ptr @.offload_sizes.35, ptr [[TMP193]], align 8
2572 // CHECK1-NEXT:    [[TMP194:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 5
2573 // CHECK1-NEXT:    store ptr @.offload_maptypes.36, ptr [[TMP194]], align 8
2574 // CHECK1-NEXT:    [[TMP195:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 6
2575 // CHECK1-NEXT:    store ptr null, ptr [[TMP195]], align 8
2576 // CHECK1-NEXT:    [[TMP196:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 7
2577 // CHECK1-NEXT:    store ptr null, ptr [[TMP196]], align 8
2578 // CHECK1-NEXT:    [[TMP197:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 8
2579 // CHECK1-NEXT:    store i64 2, ptr [[TMP197]], align 8
2580 // CHECK1-NEXT:    [[TMP198:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 9
2581 // CHECK1-NEXT:    store i64 0, ptr [[TMP198]], align 8
2582 // CHECK1-NEXT:    [[TMP199:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 10
2583 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP199]], align 4
2584 // CHECK1-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 11
2585 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP200]], align 4
2586 // CHECK1-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 12
2587 // CHECK1-NEXT:    store i32 0, ptr [[TMP201]], align 4
2588 // CHECK1-NEXT:    [[TMP202:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.region_id, ptr [[KERNEL_ARGS54]])
2589 // CHECK1-NEXT:    [[TMP203:%.*]] = icmp ne i32 [[TMP202]], 0
2590 // CHECK1-NEXT:    br i1 [[TMP203]], label [[OMP_OFFLOAD_FAILED55:%.*]], label [[OMP_OFFLOAD_CONT56:%.*]]
2591 // CHECK1:       omp_offload.failed55:
2592 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
2593 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT56]]
2594 // CHECK1:       omp_offload.cont56:
2595 // CHECK1-NEXT:    store i32 10, ptr [[T_VAR]], align 4
2596 // CHECK1-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
2597 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP204]], align 8
2598 // CHECK1-NEXT:    [[TMP205:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
2599 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[TMP205]], align 8
2600 // CHECK1-NEXT:    [[TMP206:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS59]], i64 0, i64 0
2601 // CHECK1-NEXT:    store ptr null, ptr [[TMP206]], align 8
2602 // CHECK1-NEXT:    [[TMP207:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 1
2603 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP207]], align 8
2604 // CHECK1-NEXT:    [[TMP208:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 1
2605 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[TMP208]], align 8
2606 // CHECK1-NEXT:    [[TMP209:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS59]], i64 0, i64 1
2607 // CHECK1-NEXT:    store ptr null, ptr [[TMP209]], align 8
2608 // CHECK1-NEXT:    [[TMP210:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
2609 // CHECK1-NEXT:    [[TMP211:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
2610 // CHECK1-NEXT:    [[TMP212:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 0
2611 // CHECK1-NEXT:    store i32 2, ptr [[TMP212]], align 4
2612 // CHECK1-NEXT:    [[TMP213:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 1
2613 // CHECK1-NEXT:    store i32 2, ptr [[TMP213]], align 4
2614 // CHECK1-NEXT:    [[TMP214:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 2
2615 // CHECK1-NEXT:    store ptr [[TMP210]], ptr [[TMP214]], align 8
2616 // CHECK1-NEXT:    [[TMP215:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 3
2617 // CHECK1-NEXT:    store ptr [[TMP211]], ptr [[TMP215]], align 8
2618 // CHECK1-NEXT:    [[TMP216:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 4
2619 // CHECK1-NEXT:    store ptr @.offload_sizes.37, ptr [[TMP216]], align 8
2620 // CHECK1-NEXT:    [[TMP217:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 5
2621 // CHECK1-NEXT:    store ptr @.offload_maptypes.38, ptr [[TMP217]], align 8
2622 // CHECK1-NEXT:    [[TMP218:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 6
2623 // CHECK1-NEXT:    store ptr null, ptr [[TMP218]], align 8
2624 // CHECK1-NEXT:    [[TMP219:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 7
2625 // CHECK1-NEXT:    store ptr null, ptr [[TMP219]], align 8
2626 // CHECK1-NEXT:    [[TMP220:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 8
2627 // CHECK1-NEXT:    store i64 2, ptr [[TMP220]], align 8
2628 // CHECK1-NEXT:    [[TMP221:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 9
2629 // CHECK1-NEXT:    store i64 0, ptr [[TMP221]], align 8
2630 // CHECK1-NEXT:    [[TMP222:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 10
2631 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP222]], align 4
2632 // CHECK1-NEXT:    [[TMP223:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 11
2633 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP223]], align 4
2634 // CHECK1-NEXT:    [[TMP224:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 12
2635 // CHECK1-NEXT:    store i32 0, ptr [[TMP224]], align 4
2636 // CHECK1-NEXT:    [[TMP225:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.region_id, ptr [[KERNEL_ARGS61]])
2637 // CHECK1-NEXT:    [[TMP226:%.*]] = icmp ne i32 [[TMP225]], 0
2638 // CHECK1-NEXT:    br i1 [[TMP226]], label [[OMP_OFFLOAD_FAILED62:%.*]], label [[OMP_OFFLOAD_CONT63:%.*]]
2639 // CHECK1:       omp_offload.failed62:
2640 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
2641 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT63]]
2642 // CHECK1:       omp_offload.cont63:
2643 // CHECK1-NEXT:    ret i32 0
2644 //
2645 //
2646 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32
2647 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] {
2648 // CHECK1-NEXT:  entry:
2649 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
2650 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
2651 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
2652 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined, ptr [[TMP0]])
2653 // CHECK1-NEXT:    ret void
2654 //
2655 //
2656 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined
2657 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] {
2658 // CHECK1-NEXT:  entry:
2659 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2660 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2661 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
2662 // CHECK1-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
2663 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2664 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2665 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2666 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2667 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2668 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2669 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
2670 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
2671 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2672 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2673 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
2674 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
2675 // CHECK1-NEXT:    store i32 0, ptr [[T_VAR1]], align 4
2676 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2677 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
2678 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2679 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2680 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2681 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2682 // 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)
2683 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2684 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
2685 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2686 // CHECK1:       cond.true:
2687 // CHECK1-NEXT:    br label [[COND_END:%.*]]
2688 // CHECK1:       cond.false:
2689 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2690 // CHECK1-NEXT:    br label [[COND_END]]
2691 // CHECK1:       cond.end:
2692 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2693 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2694 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2695 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2696 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2697 // CHECK1:       omp.inner.for.cond:
2698 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2699 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2700 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2701 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2702 // CHECK1:       omp.inner.for.body:
2703 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2704 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2705 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2706 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2707 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
2708 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
2709 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
2710 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[T_VAR1]], align 4
2711 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2712 // CHECK1:       omp.body.continue:
2713 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2714 // CHECK1:       omp.inner.for.inc:
2715 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2716 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
2717 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
2718 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
2719 // CHECK1:       omp.inner.for.end:
2720 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2721 // CHECK1:       omp.loop.exit:
2722 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
2723 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2724 // CHECK1-NEXT:    store ptr [[T_VAR1]], ptr [[TMP12]], align 8
2725 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
2726 // CHECK1-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2727 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2728 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2729 // CHECK1-NEXT:    ]
2730 // CHECK1:       .omp.reduction.case1:
2731 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
2732 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR1]], align 4
2733 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
2734 // CHECK1-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
2735 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
2736 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2737 // CHECK1:       .omp.reduction.case2:
2738 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[T_VAR1]], align 4
2739 // CHECK1-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
2740 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
2741 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2742 // CHECK1:       .omp.reduction.default:
2743 // CHECK1-NEXT:    ret void
2744 //
2745 //
2746 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func
2747 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
2748 // CHECK1-NEXT:  entry:
2749 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
2750 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
2751 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
2752 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
2753 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
2754 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
2755 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
2756 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
2757 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
2758 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
2759 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2760 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
2761 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
2762 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
2763 // CHECK1-NEXT:    ret void
2764 //
2765 //
2766 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37
2767 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
2768 // CHECK1-NEXT:  entry:
2769 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
2770 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
2771 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
2772 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
2773 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
2774 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
2775 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
2776 // CHECK1-NEXT:    ret void
2777 //
2778 //
2779 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined
2780 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
2781 // CHECK1-NEXT:  entry:
2782 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2783 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2784 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
2785 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
2786 // CHECK1-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
2787 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2788 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2789 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2790 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2791 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2792 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2793 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
2794 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
2795 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2796 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2797 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
2798 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
2799 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
2800 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
2801 // CHECK1-NEXT:    store i32 0, ptr [[T_VAR1]], align 4
2802 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2803 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
2804 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2805 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2806 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2807 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
2808 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2809 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2810 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
2811 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2812 // CHECK1:       cond.true:
2813 // CHECK1-NEXT:    br label [[COND_END:%.*]]
2814 // CHECK1:       cond.false:
2815 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2816 // CHECK1-NEXT:    br label [[COND_END]]
2817 // CHECK1:       cond.end:
2818 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2819 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2820 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2821 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
2822 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2823 // CHECK1:       omp.inner.for.cond:
2824 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2825 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2826 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2827 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2828 // CHECK1:       omp.inner.for.body:
2829 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2830 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2831 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2832 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2833 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
2834 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP10]] to i64
2835 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
2836 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
2837 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[T_VAR1]], align 4
2838 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP12]], [[TMP11]]
2839 // CHECK1-NEXT:    store i32 [[SUB]], ptr [[T_VAR1]], align 4
2840 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2841 // CHECK1:       omp.body.continue:
2842 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2843 // CHECK1:       omp.inner.for.inc:
2844 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2845 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
2846 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
2847 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
2848 // CHECK1:       omp.inner.for.end:
2849 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2850 // CHECK1:       omp.loop.exit:
2851 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
2852 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2853 // CHECK1-NEXT:    store ptr [[T_VAR1]], ptr [[TMP14]], align 8
2854 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
2855 // CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2856 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2857 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2858 // CHECK1-NEXT:    ]
2859 // CHECK1:       .omp.reduction.case1:
2860 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
2861 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR1]], align 4
2862 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
2863 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[TMP0]], align 4
2864 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
2865 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2866 // CHECK1:       .omp.reduction.case2:
2867 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR1]], align 4
2868 // CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
2869 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
2870 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2871 // CHECK1:       .omp.reduction.default:
2872 // CHECK1-NEXT:    ret void
2873 //
2874 //
2875 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined.omp.reduction.reduction_func
2876 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
2877 // CHECK1-NEXT:  entry:
2878 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
2879 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
2880 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
2881 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
2882 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
2883 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
2884 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
2885 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
2886 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
2887 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
2888 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2889 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
2890 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
2891 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
2892 // CHECK1-NEXT:    ret void
2893 //
2894 //
2895 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43
2896 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
2897 // CHECK1-NEXT:  entry:
2898 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
2899 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
2900 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
2901 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
2902 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
2903 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
2904 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
2905 // CHECK1-NEXT:    ret void
2906 //
2907 //
2908 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined
2909 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
2910 // CHECK1-NEXT:  entry:
2911 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2912 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2913 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
2914 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
2915 // CHECK1-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
2916 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2917 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2918 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2919 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2920 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2921 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2922 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
2923 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
2924 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i32, align 4
2925 // CHECK1-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
2926 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2927 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2928 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
2929 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
2930 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
2931 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
2932 // CHECK1-NEXT:    store i32 1, ptr [[T_VAR1]], align 4
2933 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2934 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
2935 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2936 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2937 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2938 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
2939 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2940 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2941 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
2942 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2943 // CHECK1:       cond.true:
2944 // CHECK1-NEXT:    br label [[COND_END:%.*]]
2945 // CHECK1:       cond.false:
2946 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2947 // CHECK1-NEXT:    br label [[COND_END]]
2948 // CHECK1:       cond.end:
2949 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2950 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2951 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2952 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
2953 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2954 // CHECK1:       omp.inner.for.cond:
2955 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2956 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2957 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2958 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2959 // CHECK1:       omp.inner.for.body:
2960 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2961 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2962 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2963 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2964 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
2965 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
2966 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
2967 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
2968 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
2969 // CHECK1-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[TMP10]], [[TMP12]]
2970 // CHECK1-NEXT:    store i32 [[MUL3]], ptr [[T_VAR1]], align 4
2971 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2972 // CHECK1:       omp.body.continue:
2973 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2974 // CHECK1:       omp.inner.for.inc:
2975 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2976 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
2977 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
2978 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
2979 // CHECK1:       omp.inner.for.end:
2980 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2981 // CHECK1:       omp.loop.exit:
2982 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
2983 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
2984 // CHECK1-NEXT:    store ptr [[T_VAR1]], ptr [[TMP14]], align 8
2985 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
2986 // CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
2987 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
2988 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
2989 // CHECK1-NEXT:    ]
2990 // CHECK1:       .omp.reduction.case1:
2991 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
2992 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR1]], align 4
2993 // CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP16]], [[TMP17]]
2994 // CHECK1-NEXT:    store i32 [[MUL5]], ptr [[TMP0]], align 4
2995 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
2996 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
2997 // CHECK1:       .omp.reduction.case2:
2998 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR1]], align 4
2999 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, ptr [[TMP0]] monotonic, align 4
3000 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
3001 // CHECK1:       atomic_cont:
3002 // CHECK1-NEXT:    [[TMP19:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP24:%.*]], [[ATOMIC_CONT]] ]
3003 // CHECK1-NEXT:    store i32 [[TMP19]], ptr [[_TMP6]], align 4
3004 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[_TMP6]], align 4
3005 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[T_VAR1]], align 4
3006 // CHECK1-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[TMP20]], [[TMP21]]
3007 // CHECK1-NEXT:    store i32 [[MUL7]], ptr [[ATOMIC_TEMP]], align 4
3008 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
3009 // CHECK1-NEXT:    [[TMP23:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP19]], i32 [[TMP22]] monotonic monotonic, align 4
3010 // CHECK1-NEXT:    [[TMP24]] = extractvalue { i32, i1 } [[TMP23]], 0
3011 // CHECK1-NEXT:    [[TMP25:%.*]] = extractvalue { i32, i1 } [[TMP23]], 1
3012 // CHECK1-NEXT:    br i1 [[TMP25]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
3013 // CHECK1:       atomic_exit:
3014 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3015 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3016 // CHECK1:       .omp.reduction.default:
3017 // CHECK1-NEXT:    ret void
3018 //
3019 //
3020 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined.omp.reduction.reduction_func
3021 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
3022 // CHECK1-NEXT:  entry:
3023 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
3024 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
3025 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
3026 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
3027 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
3028 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
3029 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
3030 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
3031 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
3032 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
3033 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
3034 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
3035 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], [[TMP9]]
3036 // CHECK1-NEXT:    store i32 [[MUL]], ptr [[TMP7]], align 4
3037 // CHECK1-NEXT:    ret void
3038 //
3039 //
3040 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49
3041 // CHECK1-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3042 // CHECK1-NEXT:  entry:
3043 // CHECK1-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 8
3044 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3045 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 8
3046 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3047 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 8
3048 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3049 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
3050 // CHECK1-NEXT:    ret void
3051 //
3052 //
3053 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined
3054 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3055 // CHECK1-NEXT:  entry:
3056 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
3057 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
3058 // CHECK1-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 8
3059 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3060 // CHECK1-NEXT:    [[AND_VAR1:%.*]] = alloca i8, align 1
3061 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3062 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3063 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3064 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3065 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3066 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3067 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
3068 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
3069 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
3070 // CHECK1-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
3071 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
3072 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
3073 // CHECK1-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 8
3074 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3075 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 8
3076 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3077 // CHECK1-NEXT:    store i8 1, ptr [[AND_VAR1]], align 1
3078 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3079 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
3080 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3081 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3082 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
3083 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
3084 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3085 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3086 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
3087 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3088 // CHECK1:       cond.true:
3089 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3090 // CHECK1:       cond.false:
3091 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3092 // CHECK1-NEXT:    br label [[COND_END]]
3093 // CHECK1:       cond.end:
3094 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3095 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3096 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3097 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
3098 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3099 // CHECK1:       omp.inner.for.cond:
3100 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3101 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3102 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3103 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3104 // CHECK1:       omp.inner.for.body:
3105 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3106 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3107 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3108 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
3109 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8, ptr [[AND_VAR1]], align 1
3110 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP10]] to i1
3111 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
3112 // CHECK1:       land.rhs:
3113 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
3114 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3115 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
3116 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
3117 // CHECK1-NEXT:    [[REM:%.*]] = srem i32 [[TMP12]], 2
3118 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[REM]], 0
3119 // CHECK1-NEXT:    br label [[LAND_END]]
3120 // CHECK1:       land.end:
3121 // CHECK1-NEXT:    [[TMP13:%.*]] = phi i1 [ false, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LAND_RHS]] ]
3122 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP13]] to i8
3123 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[AND_VAR1]], align 1
3124 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3125 // CHECK1:       omp.body.continue:
3126 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3127 // CHECK1:       omp.inner.for.inc:
3128 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3129 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
3130 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
3131 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
3132 // CHECK1:       omp.inner.for.end:
3133 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3134 // CHECK1:       omp.loop.exit:
3135 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
3136 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3137 // CHECK1-NEXT:    store ptr [[AND_VAR1]], ptr [[TMP15]], align 8
3138 // CHECK1-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
3139 // CHECK1-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3140 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3141 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3142 // CHECK1-NEXT:    ]
3143 // CHECK1:       .omp.reduction.case1:
3144 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8, ptr [[TMP0]], align 1
3145 // CHECK1-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP17]] to i1
3146 // CHECK1-NEXT:    br i1 [[TOBOOL5]], label [[LAND_RHS6:%.*]], label [[LAND_END8:%.*]]
3147 // CHECK1:       land.rhs6:
3148 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[AND_VAR1]], align 1
3149 // CHECK1-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP18]] to i1
3150 // CHECK1-NEXT:    br label [[LAND_END8]]
3151 // CHECK1:       land.end8:
3152 // CHECK1-NEXT:    [[TMP19:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LAND_RHS6]] ]
3153 // CHECK1-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP19]] to i8
3154 // CHECK1-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
3155 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3156 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3157 // CHECK1:       .omp.reduction.case2:
3158 // CHECK1-NEXT:    [[TMP20:%.*]] = load i8, ptr [[AND_VAR1]], align 1
3159 // CHECK1-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP20]] to i1
3160 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
3161 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
3162 // CHECK1:       atomic_cont:
3163 // CHECK1-NEXT:    [[TMP21:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP27:%.*]], [[LAND_END17:%.*]] ]
3164 // CHECK1-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP21]] to i1
3165 // CHECK1-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
3166 // CHECK1-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
3167 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, ptr [[_TMP12]], align 1
3168 // CHECK1-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP22]] to i1
3169 // CHECK1-NEXT:    br i1 [[TOBOOL14]], label [[LAND_RHS15:%.*]], label [[LAND_END17]]
3170 // CHECK1:       land.rhs15:
3171 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, ptr [[AND_VAR1]], align 1
3172 // CHECK1-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP23]] to i1
3173 // CHECK1-NEXT:    br label [[LAND_END17]]
3174 // CHECK1:       land.end17:
3175 // CHECK1-NEXT:    [[TMP24:%.*]] = phi i1 [ false, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LAND_RHS15]] ]
3176 // CHECK1-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP24]] to i8
3177 // CHECK1-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
3178 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
3179 // CHECK1-NEXT:    [[TMP26:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP21]], i8 [[TMP25]] monotonic monotonic, align 1
3180 // CHECK1-NEXT:    [[TMP27]] = extractvalue { i8, i1 } [[TMP26]], 0
3181 // CHECK1-NEXT:    [[TMP28:%.*]] = extractvalue { i8, i1 } [[TMP26]], 1
3182 // CHECK1-NEXT:    br i1 [[TMP28]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
3183 // CHECK1:       atomic_exit:
3184 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3185 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3186 // CHECK1:       .omp.reduction.default:
3187 // CHECK1-NEXT:    ret void
3188 //
3189 //
3190 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined.omp.reduction.reduction_func
3191 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
3192 // CHECK1-NEXT:  entry:
3193 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
3194 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
3195 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
3196 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
3197 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
3198 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
3199 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
3200 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
3201 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
3202 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
3203 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
3204 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
3205 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
3206 // CHECK1:       land.rhs:
3207 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
3208 // CHECK1-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
3209 // CHECK1-NEXT:    br label [[LAND_END]]
3210 // CHECK1:       land.end:
3211 // CHECK1-NEXT:    [[TMP10:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LAND_RHS]] ]
3212 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
3213 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
3214 // CHECK1-NEXT:    ret void
3215 //
3216 //
3217 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55
3218 // CHECK1-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3219 // CHECK1-NEXT:  entry:
3220 // CHECK1-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 8
3221 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3222 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 8
3223 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3224 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 8
3225 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3226 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
3227 // CHECK1-NEXT:    ret void
3228 //
3229 //
3230 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined
3231 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3232 // CHECK1-NEXT:  entry:
3233 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
3234 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
3235 // CHECK1-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 8
3236 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3237 // CHECK1-NEXT:    [[OR_VAR1:%.*]] = alloca i8, align 1
3238 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3239 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3240 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3241 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3242 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3243 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3244 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
3245 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
3246 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
3247 // CHECK1-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
3248 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
3249 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
3250 // CHECK1-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 8
3251 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3252 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 8
3253 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3254 // CHECK1-NEXT:    store i8 0, ptr [[OR_VAR1]], align 1
3255 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3256 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
3257 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3258 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3259 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
3260 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
3261 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3262 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3263 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
3264 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3265 // CHECK1:       cond.true:
3266 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3267 // CHECK1:       cond.false:
3268 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3269 // CHECK1-NEXT:    br label [[COND_END]]
3270 // CHECK1:       cond.end:
3271 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3272 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3273 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3274 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
3275 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3276 // CHECK1:       omp.inner.for.cond:
3277 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3278 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3279 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3280 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3281 // CHECK1:       omp.inner.for.body:
3282 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3283 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3284 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3285 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
3286 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8, ptr [[OR_VAR1]], align 1
3287 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP10]] to i1
3288 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
3289 // CHECK1:       lor.rhs:
3290 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
3291 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3292 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
3293 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
3294 // CHECK1-NEXT:    [[REM:%.*]] = srem i32 [[TMP12]], 2
3295 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[REM]], 0
3296 // CHECK1-NEXT:    br label [[LOR_END]]
3297 // CHECK1:       lor.end:
3298 // CHECK1-NEXT:    [[TMP13:%.*]] = phi i1 [ true, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LOR_RHS]] ]
3299 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP13]] to i8
3300 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[OR_VAR1]], align 1
3301 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3302 // CHECK1:       omp.body.continue:
3303 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3304 // CHECK1:       omp.inner.for.inc:
3305 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3306 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
3307 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
3308 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
3309 // CHECK1:       omp.inner.for.end:
3310 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3311 // CHECK1:       omp.loop.exit:
3312 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
3313 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3314 // CHECK1-NEXT:    store ptr [[OR_VAR1]], ptr [[TMP15]], align 8
3315 // CHECK1-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
3316 // CHECK1-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3317 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3318 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3319 // CHECK1-NEXT:    ]
3320 // CHECK1:       .omp.reduction.case1:
3321 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8, ptr [[TMP0]], align 1
3322 // CHECK1-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP17]] to i1
3323 // CHECK1-NEXT:    br i1 [[TOBOOL5]], label [[LOR_END8:%.*]], label [[LOR_RHS6:%.*]]
3324 // CHECK1:       lor.rhs6:
3325 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[OR_VAR1]], align 1
3326 // CHECK1-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP18]] to i1
3327 // CHECK1-NEXT:    br label [[LOR_END8]]
3328 // CHECK1:       lor.end8:
3329 // CHECK1-NEXT:    [[TMP19:%.*]] = phi i1 [ true, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LOR_RHS6]] ]
3330 // CHECK1-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP19]] to i8
3331 // CHECK1-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
3332 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3333 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3334 // CHECK1:       .omp.reduction.case2:
3335 // CHECK1-NEXT:    [[TMP20:%.*]] = load i8, ptr [[OR_VAR1]], align 1
3336 // CHECK1-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP20]] to i1
3337 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
3338 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
3339 // CHECK1:       atomic_cont:
3340 // CHECK1-NEXT:    [[TMP21:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP27:%.*]], [[LOR_END17:%.*]] ]
3341 // CHECK1-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP21]] to i1
3342 // CHECK1-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
3343 // CHECK1-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
3344 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, ptr [[_TMP12]], align 1
3345 // CHECK1-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP22]] to i1
3346 // CHECK1-NEXT:    br i1 [[TOBOOL14]], label [[LOR_END17]], label [[LOR_RHS15:%.*]]
3347 // CHECK1:       lor.rhs15:
3348 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, ptr [[OR_VAR1]], align 1
3349 // CHECK1-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP23]] to i1
3350 // CHECK1-NEXT:    br label [[LOR_END17]]
3351 // CHECK1:       lor.end17:
3352 // CHECK1-NEXT:    [[TMP24:%.*]] = phi i1 [ true, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LOR_RHS15]] ]
3353 // CHECK1-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP24]] to i8
3354 // CHECK1-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
3355 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
3356 // CHECK1-NEXT:    [[TMP26:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP21]], i8 [[TMP25]] monotonic monotonic, align 1
3357 // CHECK1-NEXT:    [[TMP27]] = extractvalue { i8, i1 } [[TMP26]], 0
3358 // CHECK1-NEXT:    [[TMP28:%.*]] = extractvalue { i8, i1 } [[TMP26]], 1
3359 // CHECK1-NEXT:    br i1 [[TMP28]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
3360 // CHECK1:       atomic_exit:
3361 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3362 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3363 // CHECK1:       .omp.reduction.default:
3364 // CHECK1-NEXT:    ret void
3365 //
3366 //
3367 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined.omp.reduction.reduction_func
3368 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
3369 // CHECK1-NEXT:  entry:
3370 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
3371 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
3372 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
3373 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
3374 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
3375 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
3376 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
3377 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
3378 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
3379 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
3380 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
3381 // CHECK1-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
3382 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
3383 // CHECK1:       lor.rhs:
3384 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
3385 // CHECK1-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
3386 // CHECK1-NEXT:    br label [[LOR_END]]
3387 // CHECK1:       lor.end:
3388 // CHECK1-NEXT:    [[TMP10:%.*]] = phi i1 [ true, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LOR_RHS]] ]
3389 // CHECK1-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
3390 // CHECK1-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
3391 // CHECK1-NEXT:    ret void
3392 //
3393 //
3394 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61
3395 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3396 // CHECK1-NEXT:  entry:
3397 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
3398 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3399 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
3400 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3401 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
3402 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3403 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
3404 // CHECK1-NEXT:    ret void
3405 //
3406 //
3407 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined
3408 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3409 // CHECK1-NEXT:  entry:
3410 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
3411 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
3412 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
3413 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3414 // CHECK1-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
3415 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3416 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3417 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3418 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3419 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3420 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3421 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
3422 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
3423 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
3424 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
3425 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
3426 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3427 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
3428 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3429 // CHECK1-NEXT:    store i32 -1, ptr [[BIT_VAR1]], align 4
3430 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3431 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
3432 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3433 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3434 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
3435 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
3436 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3437 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3438 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
3439 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3440 // CHECK1:       cond.true:
3441 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3442 // CHECK1:       cond.false:
3443 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3444 // CHECK1-NEXT:    br label [[COND_END]]
3445 // CHECK1:       cond.end:
3446 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3447 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3448 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3449 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
3450 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3451 // CHECK1:       omp.inner.for.cond:
3452 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3453 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3454 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3455 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3456 // CHECK1:       omp.inner.for.body:
3457 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3458 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3459 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3460 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
3461 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3462 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
3463 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3464 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
3465 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
3466 // CHECK1-NEXT:    [[AND:%.*]] = and i32 [[TMP10]], [[TMP12]]
3467 // CHECK1-NEXT:    store i32 [[AND]], ptr [[BIT_VAR1]], align 4
3468 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3469 // CHECK1:       omp.body.continue:
3470 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3471 // CHECK1:       omp.inner.for.inc:
3472 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3473 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
3474 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
3475 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
3476 // CHECK1:       omp.inner.for.end:
3477 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3478 // CHECK1:       omp.loop.exit:
3479 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
3480 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3481 // CHECK1-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP14]], align 8
3482 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
3483 // CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3484 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3485 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3486 // CHECK1-NEXT:    ]
3487 // CHECK1:       .omp.reduction.case1:
3488 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
3489 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3490 // CHECK1-NEXT:    [[AND4:%.*]] = and i32 [[TMP16]], [[TMP17]]
3491 // CHECK1-NEXT:    store i32 [[AND4]], ptr [[TMP0]], align 4
3492 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3493 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3494 // CHECK1:       .omp.reduction.case2:
3495 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3496 // CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw and ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
3497 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3498 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3499 // CHECK1:       .omp.reduction.default:
3500 // CHECK1-NEXT:    ret void
3501 //
3502 //
3503 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined.omp.reduction.reduction_func
3504 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
3505 // CHECK1-NEXT:  entry:
3506 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
3507 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
3508 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
3509 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
3510 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
3511 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
3512 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
3513 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
3514 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
3515 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
3516 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
3517 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
3518 // CHECK1-NEXT:    [[AND:%.*]] = and i32 [[TMP8]], [[TMP9]]
3519 // CHECK1-NEXT:    store i32 [[AND]], ptr [[TMP7]], align 4
3520 // CHECK1-NEXT:    ret void
3521 //
3522 //
3523 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66
3524 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3525 // CHECK1-NEXT:  entry:
3526 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
3527 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3528 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
3529 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3530 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
3531 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3532 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
3533 // CHECK1-NEXT:    ret void
3534 //
3535 //
3536 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined
3537 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3538 // CHECK1-NEXT:  entry:
3539 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
3540 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
3541 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
3542 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3543 // CHECK1-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
3544 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3545 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3546 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3547 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3548 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3549 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3550 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
3551 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
3552 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
3553 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
3554 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
3555 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3556 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
3557 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3558 // CHECK1-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
3559 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3560 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
3561 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3562 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3563 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
3564 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
3565 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3566 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3567 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
3568 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3569 // CHECK1:       cond.true:
3570 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3571 // CHECK1:       cond.false:
3572 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3573 // CHECK1-NEXT:    br label [[COND_END]]
3574 // CHECK1:       cond.end:
3575 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3576 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3577 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3578 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
3579 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3580 // CHECK1:       omp.inner.for.cond:
3581 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3582 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3583 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3584 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3585 // CHECK1:       omp.inner.for.body:
3586 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3587 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3588 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3589 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
3590 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3591 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
3592 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3593 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
3594 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
3595 // CHECK1-NEXT:    [[OR:%.*]] = or i32 [[TMP10]], [[TMP12]]
3596 // CHECK1-NEXT:    store i32 [[OR]], ptr [[BIT_VAR1]], align 4
3597 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3598 // CHECK1:       omp.body.continue:
3599 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3600 // CHECK1:       omp.inner.for.inc:
3601 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3602 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
3603 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
3604 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
3605 // CHECK1:       omp.inner.for.end:
3606 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3607 // CHECK1:       omp.loop.exit:
3608 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
3609 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3610 // CHECK1-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP14]], align 8
3611 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
3612 // CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3613 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3614 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3615 // CHECK1-NEXT:    ]
3616 // CHECK1:       .omp.reduction.case1:
3617 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
3618 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3619 // CHECK1-NEXT:    [[OR4:%.*]] = or i32 [[TMP16]], [[TMP17]]
3620 // CHECK1-NEXT:    store i32 [[OR4]], ptr [[TMP0]], align 4
3621 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3622 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3623 // CHECK1:       .omp.reduction.case2:
3624 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3625 // CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw or ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
3626 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3627 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3628 // CHECK1:       .omp.reduction.default:
3629 // CHECK1-NEXT:    ret void
3630 //
3631 //
3632 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined.omp.reduction.reduction_func
3633 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
3634 // CHECK1-NEXT:  entry:
3635 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
3636 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
3637 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
3638 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
3639 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
3640 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
3641 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
3642 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
3643 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
3644 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
3645 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
3646 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
3647 // CHECK1-NEXT:    [[OR:%.*]] = or i32 [[TMP8]], [[TMP9]]
3648 // CHECK1-NEXT:    store i32 [[OR]], ptr [[TMP7]], align 4
3649 // CHECK1-NEXT:    ret void
3650 //
3651 //
3652 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71
3653 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3654 // CHECK1-NEXT:  entry:
3655 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
3656 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3657 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
3658 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3659 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
3660 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3661 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
3662 // CHECK1-NEXT:    ret void
3663 //
3664 //
3665 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined
3666 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3667 // CHECK1-NEXT:  entry:
3668 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
3669 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
3670 // CHECK1-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
3671 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3672 // CHECK1-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
3673 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3674 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3675 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3676 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3677 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3678 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3679 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
3680 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
3681 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
3682 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
3683 // CHECK1-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
3684 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3685 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
3686 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3687 // CHECK1-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
3688 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3689 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
3690 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3691 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3692 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
3693 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
3694 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3695 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3696 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
3697 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3698 // CHECK1:       cond.true:
3699 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3700 // CHECK1:       cond.false:
3701 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3702 // CHECK1-NEXT:    br label [[COND_END]]
3703 // CHECK1:       cond.end:
3704 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3705 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3706 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3707 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
3708 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3709 // CHECK1:       omp.inner.for.cond:
3710 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3711 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3712 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3713 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3714 // CHECK1:       omp.inner.for.body:
3715 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3716 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3717 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3718 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
3719 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3720 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
3721 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3722 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
3723 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
3724 // CHECK1-NEXT:    [[XOR:%.*]] = xor i32 [[TMP10]], [[TMP12]]
3725 // CHECK1-NEXT:    store i32 [[XOR]], ptr [[BIT_VAR1]], align 4
3726 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3727 // CHECK1:       omp.body.continue:
3728 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3729 // CHECK1:       omp.inner.for.inc:
3730 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3731 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
3732 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
3733 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
3734 // CHECK1:       omp.inner.for.end:
3735 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3736 // CHECK1:       omp.loop.exit:
3737 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
3738 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3739 // CHECK1-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP14]], align 8
3740 // CHECK1-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
3741 // CHECK1-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3742 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3743 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3744 // CHECK1-NEXT:    ]
3745 // CHECK1:       .omp.reduction.case1:
3746 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
3747 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3748 // CHECK1-NEXT:    [[XOR4:%.*]] = xor i32 [[TMP16]], [[TMP17]]
3749 // CHECK1-NEXT:    store i32 [[XOR4]], ptr [[TMP0]], align 4
3750 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3751 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3752 // CHECK1:       .omp.reduction.case2:
3753 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
3754 // CHECK1-NEXT:    [[TMP19:%.*]] = atomicrmw xor ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
3755 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3756 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3757 // CHECK1:       .omp.reduction.default:
3758 // CHECK1-NEXT:    ret void
3759 //
3760 //
3761 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined.omp.reduction.reduction_func
3762 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
3763 // CHECK1-NEXT:  entry:
3764 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
3765 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
3766 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
3767 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
3768 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
3769 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
3770 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
3771 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
3772 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
3773 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
3774 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
3775 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
3776 // CHECK1-NEXT:    [[XOR:%.*]] = xor i32 [[TMP8]], [[TMP9]]
3777 // CHECK1-NEXT:    store i32 [[XOR]], ptr [[TMP7]], align 4
3778 // CHECK1-NEXT:    ret void
3779 //
3780 //
3781 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77
3782 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3783 // CHECK1-NEXT:  entry:
3784 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
3785 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3786 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
3787 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3788 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
3789 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3790 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
3791 // CHECK1-NEXT:    ret void
3792 //
3793 //
3794 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined
3795 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3796 // CHECK1-NEXT:  entry:
3797 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
3798 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
3799 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
3800 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3801 // CHECK1-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
3802 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3803 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3804 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3805 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3806 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3807 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3808 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
3809 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
3810 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
3811 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
3812 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
3813 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3814 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
3815 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3816 // CHECK1-NEXT:    store i32 -2147483648, ptr [[T_VAR1]], align 4
3817 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3818 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
3819 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3820 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3821 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
3822 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
3823 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3824 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3825 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
3826 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3827 // CHECK1:       cond.true:
3828 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3829 // CHECK1:       cond.false:
3830 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3831 // CHECK1-NEXT:    br label [[COND_END]]
3832 // CHECK1:       cond.end:
3833 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3834 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3835 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3836 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
3837 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3838 // CHECK1:       omp.inner.for.cond:
3839 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3840 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3841 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3842 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3843 // CHECK1:       omp.inner.for.body:
3844 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3845 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3846 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3847 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
3848 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
3849 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
3850 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3851 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
3852 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
3853 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sge i32 [[TMP10]], [[TMP12]]
3854 // CHECK1-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
3855 // CHECK1:       cond.true4:
3856 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[T_VAR1]], align 4
3857 // CHECK1-NEXT:    br label [[COND_END8:%.*]]
3858 // CHECK1:       cond.false5:
3859 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
3860 // CHECK1-NEXT:    [[IDXPROM6:%.*]] = sext i32 [[TMP14]] to i64
3861 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM6]]
3862 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[ARRAYIDX7]], align 4
3863 // CHECK1-NEXT:    br label [[COND_END8]]
3864 // CHECK1:       cond.end8:
3865 // CHECK1-NEXT:    [[COND9:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE4]] ], [ [[TMP15]], [[COND_FALSE5]] ]
3866 // CHECK1-NEXT:    store i32 [[COND9]], ptr [[T_VAR1]], align 4
3867 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3868 // CHECK1:       omp.body.continue:
3869 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3870 // CHECK1:       omp.inner.for.inc:
3871 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3872 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP16]], 1
3873 // CHECK1-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
3874 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
3875 // CHECK1:       omp.inner.for.end:
3876 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3877 // CHECK1:       omp.loop.exit:
3878 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
3879 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
3880 // CHECK1-NEXT:    store ptr [[T_VAR1]], ptr [[TMP17]], align 8
3881 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
3882 // CHECK1-NEXT:    switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
3883 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
3884 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
3885 // CHECK1-NEXT:    ]
3886 // CHECK1:       .omp.reduction.case1:
3887 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
3888 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[T_VAR1]], align 4
3889 // CHECK1-NEXT:    [[CMP11:%.*]] = icmp sgt i32 [[TMP19]], [[TMP20]]
3890 // CHECK1-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
3891 // CHECK1:       cond.true12:
3892 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP0]], align 4
3893 // CHECK1-NEXT:    br label [[COND_END14:%.*]]
3894 // CHECK1:       cond.false13:
3895 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[T_VAR1]], align 4
3896 // CHECK1-NEXT:    br label [[COND_END14]]
3897 // CHECK1:       cond.end14:
3898 // CHECK1-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP21]], [[COND_TRUE12]] ], [ [[TMP22]], [[COND_FALSE13]] ]
3899 // CHECK1-NEXT:    store i32 [[COND15]], ptr [[TMP0]], align 4
3900 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3901 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3902 // CHECK1:       .omp.reduction.case2:
3903 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[T_VAR1]], align 4
3904 // CHECK1-NEXT:    [[TMP24:%.*]] = atomicrmw max ptr [[TMP0]], i32 [[TMP23]] monotonic, align 4
3905 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
3906 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
3907 // CHECK1:       .omp.reduction.default:
3908 // CHECK1-NEXT:    ret void
3909 //
3910 //
3911 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined.omp.reduction.reduction_func
3912 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
3913 // CHECK1-NEXT:  entry:
3914 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
3915 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
3916 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
3917 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
3918 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
3919 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
3920 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
3921 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
3922 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
3923 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
3924 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
3925 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
3926 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
3927 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3928 // CHECK1:       cond.true:
3929 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
3930 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3931 // CHECK1:       cond.false:
3932 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
3933 // CHECK1-NEXT:    br label [[COND_END]]
3934 // CHECK1:       cond.end:
3935 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
3936 // CHECK1-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
3937 // CHECK1-NEXT:    ret void
3938 //
3939 //
3940 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83
3941 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3942 // CHECK1-NEXT:  entry:
3943 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
3944 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3945 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
3946 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3947 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
3948 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3949 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
3950 // CHECK1-NEXT:    ret void
3951 //
3952 //
3953 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined
3954 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
3955 // CHECK1-NEXT:  entry:
3956 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
3957 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
3958 // CHECK1-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
3959 // CHECK1-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
3960 // CHECK1-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
3961 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3962 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3963 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3964 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3965 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3966 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3967 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
3968 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
3969 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
3970 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
3971 // CHECK1-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
3972 // CHECK1-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
3973 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
3974 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
3975 // CHECK1-NEXT:    store i32 2147483647, ptr [[T_VAR1]], align 4
3976 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3977 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
3978 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3979 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3980 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
3981 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
3982 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3983 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3984 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
3985 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3986 // CHECK1:       cond.true:
3987 // CHECK1-NEXT:    br label [[COND_END:%.*]]
3988 // CHECK1:       cond.false:
3989 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3990 // CHECK1-NEXT:    br label [[COND_END]]
3991 // CHECK1:       cond.end:
3992 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3993 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3994 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3995 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
3996 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3997 // CHECK1:       omp.inner.for.cond:
3998 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3999 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4000 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4001 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4002 // CHECK1:       omp.inner.for.body:
4003 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4004 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4005 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4006 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
4007 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
4008 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
4009 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4010 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
4011 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
4012 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP10]], [[TMP12]]
4013 // CHECK1-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
4014 // CHECK1:       cond.true4:
4015 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[T_VAR1]], align 4
4016 // CHECK1-NEXT:    br label [[COND_END8:%.*]]
4017 // CHECK1:       cond.false5:
4018 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
4019 // CHECK1-NEXT:    [[IDXPROM6:%.*]] = sext i32 [[TMP14]] to i64
4020 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM6]]
4021 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[ARRAYIDX7]], align 4
4022 // CHECK1-NEXT:    br label [[COND_END8]]
4023 // CHECK1:       cond.end8:
4024 // CHECK1-NEXT:    [[COND9:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE4]] ], [ [[TMP15]], [[COND_FALSE5]] ]
4025 // CHECK1-NEXT:    store i32 [[COND9]], ptr [[T_VAR1]], align 4
4026 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4027 // CHECK1:       omp.body.continue:
4028 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4029 // CHECK1:       omp.inner.for.inc:
4030 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4031 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP16]], 1
4032 // CHECK1-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
4033 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
4034 // CHECK1:       omp.inner.for.end:
4035 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4036 // CHECK1:       omp.loop.exit:
4037 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
4038 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
4039 // CHECK1-NEXT:    store ptr [[T_VAR1]], ptr [[TMP17]], align 8
4040 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
4041 // CHECK1-NEXT:    switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
4042 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
4043 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
4044 // CHECK1-NEXT:    ]
4045 // CHECK1:       .omp.reduction.case1:
4046 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
4047 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[T_VAR1]], align 4
4048 // CHECK1-NEXT:    [[CMP11:%.*]] = icmp slt i32 [[TMP19]], [[TMP20]]
4049 // CHECK1-NEXT:    br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
4050 // CHECK1:       cond.true12:
4051 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP0]], align 4
4052 // CHECK1-NEXT:    br label [[COND_END14:%.*]]
4053 // CHECK1:       cond.false13:
4054 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[T_VAR1]], align 4
4055 // CHECK1-NEXT:    br label [[COND_END14]]
4056 // CHECK1:       cond.end14:
4057 // CHECK1-NEXT:    [[COND15:%.*]] = phi i32 [ [[TMP21]], [[COND_TRUE12]] ], [ [[TMP22]], [[COND_FALSE13]] ]
4058 // CHECK1-NEXT:    store i32 [[COND15]], ptr [[TMP0]], align 4
4059 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
4060 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4061 // CHECK1:       .omp.reduction.case2:
4062 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[T_VAR1]], align 4
4063 // CHECK1-NEXT:    [[TMP24:%.*]] = atomicrmw min ptr [[TMP0]], i32 [[TMP23]] monotonic, align 4
4064 // CHECK1-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
4065 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4066 // CHECK1:       .omp.reduction.default:
4067 // CHECK1-NEXT:    ret void
4068 //
4069 //
4070 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined.omp.reduction.reduction_func
4071 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
4072 // CHECK1-NEXT:  entry:
4073 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
4074 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
4075 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
4076 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
4077 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
4078 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
4079 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
4080 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
4081 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
4082 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
4083 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
4084 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
4085 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], [[TMP9]]
4086 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4087 // CHECK1:       cond.true:
4088 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
4089 // CHECK1-NEXT:    br label [[COND_END:%.*]]
4090 // CHECK1:       cond.false:
4091 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
4092 // CHECK1-NEXT:    br label [[COND_END]]
4093 // CHECK1:       cond.end:
4094 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
4095 // CHECK1-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
4096 // CHECK1-NEXT:    ret void
4097 //
4098 //
4099 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4100 // CHECK1-SAME: () #[[ATTR7:[0-9]+]] {
4101 // CHECK1-NEXT:  entry:
4102 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
4103 // CHECK1-NEXT:    ret void
4104 //
4105 //
4106 // CHECK3-LABEL: define {{[^@]+}}@main
4107 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
4108 // CHECK3-NEXT:  entry:
4109 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
4110 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
4111 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
4112 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
4113 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4114 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
4115 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x ptr], align 4
4116 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x ptr], align 4
4117 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x ptr], align 4
4118 // CHECK3-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
4119 // CHECK3-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4120 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x ptr], align 4
4121 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x ptr], align 4
4122 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x ptr], align 4
4123 // CHECK3-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
4124 // CHECK3-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4125 // CHECK3-NEXT:    [[AND_VAR:%.*]] = alloca i8, align 1
4126 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [1 x ptr], align 4
4127 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [1 x ptr], align 4
4128 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [1 x ptr], align 4
4129 // CHECK3-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
4130 // CHECK3-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4131 // CHECK3-NEXT:    [[OR_VAR:%.*]] = alloca i8, align 1
4132 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [1 x ptr], align 4
4133 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS23:%.*]] = alloca [1 x ptr], align 4
4134 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [1 x ptr], align 4
4135 // CHECK3-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
4136 // CHECK3-NEXT:    [[KERNEL_ARGS26:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4137 // CHECK3-NEXT:    [[BIT_VAR:%.*]] = alloca i32, align 4
4138 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [1 x ptr], align 4
4139 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [1 x ptr], align 4
4140 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [1 x ptr], align 4
4141 // CHECK3-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
4142 // CHECK3-NEXT:    [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4143 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [1 x ptr], align 4
4144 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [1 x ptr], align 4
4145 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [1 x ptr], align 4
4146 // CHECK3-NEXT:    [[_TMP39:%.*]] = alloca i32, align 4
4147 // CHECK3-NEXT:    [[KERNEL_ARGS40:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4148 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS43:%.*]] = alloca [1 x ptr], align 4
4149 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS44:%.*]] = alloca [1 x ptr], align 4
4150 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS45:%.*]] = alloca [1 x ptr], align 4
4151 // CHECK3-NEXT:    [[_TMP46:%.*]] = alloca i32, align 4
4152 // CHECK3-NEXT:    [[KERNEL_ARGS47:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4153 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS50:%.*]] = alloca [1 x ptr], align 4
4154 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS51:%.*]] = alloca [1 x ptr], align 4
4155 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS52:%.*]] = alloca [1 x ptr], align 4
4156 // CHECK3-NEXT:    [[_TMP53:%.*]] = alloca i32, align 4
4157 // CHECK3-NEXT:    [[KERNEL_ARGS54:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4158 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS57:%.*]] = alloca [1 x ptr], align 4
4159 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS58:%.*]] = alloca [1 x ptr], align 4
4160 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS59:%.*]] = alloca [1 x ptr], align 4
4161 // CHECK3-NEXT:    [[_TMP60:%.*]] = alloca i32, align 4
4162 // CHECK3-NEXT:    [[KERNEL_ARGS61:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4163 // CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
4164 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4165 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP0]], align 4
4166 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4167 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP1]], align 4
4168 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4169 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
4170 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4171 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4172 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
4173 // CHECK3-NEXT:    store i32 2, ptr [[TMP5]], align 4
4174 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
4175 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
4176 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
4177 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
4178 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
4179 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
4180 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
4181 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 4
4182 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
4183 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 4
4184 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
4185 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
4186 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
4187 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
4188 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
4189 // CHECK3-NEXT:    store i64 2, ptr [[TMP13]], align 8
4190 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
4191 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
4192 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
4193 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
4194 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
4195 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
4196 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
4197 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
4198 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.region_id, ptr [[KERNEL_ARGS]])
4199 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
4200 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4201 // CHECK3:       omp_offload.failed:
4202 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209(ptr @_ZZ4mainE5sivar) #[[ATTR2:[0-9]+]]
4203 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4204 // CHECK3:       omp_offload.cont:
4205 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
4206 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP20]], align 4
4207 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
4208 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP21]], align 4
4209 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
4210 // CHECK3-NEXT:    store ptr null, ptr [[TMP22]], align 4
4211 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
4212 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
4213 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
4214 // CHECK3-NEXT:    store i32 2, ptr [[TMP25]], align 4
4215 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
4216 // CHECK3-NEXT:    store i32 1, ptr [[TMP26]], align 4
4217 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
4218 // CHECK3-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 4
4219 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
4220 // CHECK3-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 4
4221 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
4222 // CHECK3-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP29]], align 4
4223 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
4224 // CHECK3-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP30]], align 4
4225 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
4226 // CHECK3-NEXT:    store ptr null, ptr [[TMP31]], align 4
4227 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
4228 // CHECK3-NEXT:    store ptr null, ptr [[TMP32]], align 4
4229 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
4230 // CHECK3-NEXT:    store i64 2, ptr [[TMP33]], align 8
4231 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
4232 // CHECK3-NEXT:    store i64 0, ptr [[TMP34]], align 8
4233 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
4234 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
4235 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
4236 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
4237 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
4238 // CHECK3-NEXT:    store i32 0, ptr [[TMP37]], align 4
4239 // CHECK3-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.region_id, ptr [[KERNEL_ARGS5]])
4240 // CHECK3-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
4241 // CHECK3-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
4242 // CHECK3:       omp_offload.failed6:
4243 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
4244 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
4245 // CHECK3:       omp_offload.cont7:
4246 // CHECK3-NEXT:    store i32 1, ptr @_ZZ4mainE5sivar, align 4
4247 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
4248 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP40]], align 4
4249 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
4250 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP41]], align 4
4251 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
4252 // CHECK3-NEXT:    store ptr null, ptr [[TMP42]], align 4
4253 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
4254 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
4255 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
4256 // CHECK3-NEXT:    store i32 2, ptr [[TMP45]], align 4
4257 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
4258 // CHECK3-NEXT:    store i32 1, ptr [[TMP46]], align 4
4259 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
4260 // CHECK3-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 4
4261 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
4262 // CHECK3-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 4
4263 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
4264 // CHECK3-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP49]], align 4
4265 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
4266 // CHECK3-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP50]], align 4
4267 // CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
4268 // CHECK3-NEXT:    store ptr null, ptr [[TMP51]], align 4
4269 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
4270 // CHECK3-NEXT:    store ptr null, ptr [[TMP52]], align 4
4271 // CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
4272 // CHECK3-NEXT:    store i64 2, ptr [[TMP53]], align 8
4273 // CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
4274 // CHECK3-NEXT:    store i64 0, ptr [[TMP54]], align 8
4275 // CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
4276 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
4277 // CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
4278 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
4279 // CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
4280 // CHECK3-NEXT:    store i32 0, ptr [[TMP57]], align 4
4281 // CHECK3-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.region_id, ptr [[KERNEL_ARGS12]])
4282 // CHECK3-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
4283 // CHECK3-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
4284 // CHECK3:       omp_offload.failed13:
4285 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
4286 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
4287 // CHECK3:       omp_offload.cont14:
4288 // CHECK3-NEXT:    store i8 1, ptr [[AND_VAR]], align 1
4289 // CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
4290 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[TMP60]], align 4
4291 // CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
4292 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[TMP61]], align 4
4293 // CHECK3-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS17]], i32 0, i32 0
4294 // CHECK3-NEXT:    store ptr null, ptr [[TMP62]], align 4
4295 // CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
4296 // CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
4297 // CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 0
4298 // CHECK3-NEXT:    store i32 2, ptr [[TMP65]], align 4
4299 // CHECK3-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 1
4300 // CHECK3-NEXT:    store i32 1, ptr [[TMP66]], align 4
4301 // CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 2
4302 // CHECK3-NEXT:    store ptr [[TMP63]], ptr [[TMP67]], align 4
4303 // CHECK3-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 3
4304 // CHECK3-NEXT:    store ptr [[TMP64]], ptr [[TMP68]], align 4
4305 // CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 4
4306 // CHECK3-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP69]], align 4
4307 // CHECK3-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 5
4308 // CHECK3-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP70]], align 4
4309 // CHECK3-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 6
4310 // CHECK3-NEXT:    store ptr null, ptr [[TMP71]], align 4
4311 // CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 7
4312 // CHECK3-NEXT:    store ptr null, ptr [[TMP72]], align 4
4313 // CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 8
4314 // CHECK3-NEXT:    store i64 2, ptr [[TMP73]], align 8
4315 // CHECK3-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 9
4316 // CHECK3-NEXT:    store i64 0, ptr [[TMP74]], align 8
4317 // CHECK3-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 10
4318 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP75]], align 4
4319 // CHECK3-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 11
4320 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP76]], align 4
4321 // CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 12
4322 // CHECK3-NEXT:    store i32 0, ptr [[TMP77]], align 4
4323 // CHECK3-NEXT:    [[TMP78:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.region_id, ptr [[KERNEL_ARGS19]])
4324 // CHECK3-NEXT:    [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
4325 // CHECK3-NEXT:    br i1 [[TMP79]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
4326 // CHECK3:       omp_offload.failed20:
4327 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226(ptr [[AND_VAR]]) #[[ATTR2]]
4328 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
4329 // CHECK3:       omp_offload.cont21:
4330 // CHECK3-NEXT:    store i8 0, ptr [[OR_VAR]], align 1
4331 // CHECK3-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
4332 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[TMP80]], align 4
4333 // CHECK3-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
4334 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[TMP81]], align 4
4335 // CHECK3-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS24]], i32 0, i32 0
4336 // CHECK3-NEXT:    store ptr null, ptr [[TMP82]], align 4
4337 // CHECK3-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
4338 // CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
4339 // CHECK3-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 0
4340 // CHECK3-NEXT:    store i32 2, ptr [[TMP85]], align 4
4341 // CHECK3-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 1
4342 // CHECK3-NEXT:    store i32 1, ptr [[TMP86]], align 4
4343 // CHECK3-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 2
4344 // CHECK3-NEXT:    store ptr [[TMP83]], ptr [[TMP87]], align 4
4345 // CHECK3-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 3
4346 // CHECK3-NEXT:    store ptr [[TMP84]], ptr [[TMP88]], align 4
4347 // CHECK3-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 4
4348 // CHECK3-NEXT:    store ptr @.offload_sizes.7, ptr [[TMP89]], align 4
4349 // CHECK3-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 5
4350 // CHECK3-NEXT:    store ptr @.offload_maptypes.8, ptr [[TMP90]], align 4
4351 // CHECK3-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 6
4352 // CHECK3-NEXT:    store ptr null, ptr [[TMP91]], align 4
4353 // CHECK3-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 7
4354 // CHECK3-NEXT:    store ptr null, ptr [[TMP92]], align 4
4355 // CHECK3-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 8
4356 // CHECK3-NEXT:    store i64 2, ptr [[TMP93]], align 8
4357 // CHECK3-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 9
4358 // CHECK3-NEXT:    store i64 0, ptr [[TMP94]], align 8
4359 // CHECK3-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 10
4360 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP95]], align 4
4361 // CHECK3-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 11
4362 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP96]], align 4
4363 // CHECK3-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 12
4364 // CHECK3-NEXT:    store i32 0, ptr [[TMP97]], align 4
4365 // CHECK3-NEXT:    [[TMP98:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.region_id, ptr [[KERNEL_ARGS26]])
4366 // CHECK3-NEXT:    [[TMP99:%.*]] = icmp ne i32 [[TMP98]], 0
4367 // CHECK3-NEXT:    br i1 [[TMP99]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
4368 // CHECK3:       omp_offload.failed27:
4369 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232(ptr [[OR_VAR]]) #[[ATTR2]]
4370 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
4371 // CHECK3:       omp_offload.cont28:
4372 // CHECK3-NEXT:    store i32 1, ptr [[BIT_VAR]], align 4
4373 // CHECK3-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
4374 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP100]], align 4
4375 // CHECK3-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
4376 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP101]], align 4
4377 // CHECK3-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS31]], i32 0, i32 0
4378 // CHECK3-NEXT:    store ptr null, ptr [[TMP102]], align 4
4379 // CHECK3-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
4380 // CHECK3-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
4381 // CHECK3-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 0
4382 // CHECK3-NEXT:    store i32 2, ptr [[TMP105]], align 4
4383 // CHECK3-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 1
4384 // CHECK3-NEXT:    store i32 1, ptr [[TMP106]], align 4
4385 // CHECK3-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 2
4386 // CHECK3-NEXT:    store ptr [[TMP103]], ptr [[TMP107]], align 4
4387 // CHECK3-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 3
4388 // CHECK3-NEXT:    store ptr [[TMP104]], ptr [[TMP108]], align 4
4389 // CHECK3-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 4
4390 // CHECK3-NEXT:    store ptr @.offload_sizes.9, ptr [[TMP109]], align 4
4391 // CHECK3-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 5
4392 // CHECK3-NEXT:    store ptr @.offload_maptypes.10, ptr [[TMP110]], align 4
4393 // CHECK3-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 6
4394 // CHECK3-NEXT:    store ptr null, ptr [[TMP111]], align 4
4395 // CHECK3-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 7
4396 // CHECK3-NEXT:    store ptr null, ptr [[TMP112]], align 4
4397 // CHECK3-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 8
4398 // CHECK3-NEXT:    store i64 2, ptr [[TMP113]], align 8
4399 // CHECK3-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 9
4400 // CHECK3-NEXT:    store i64 0, ptr [[TMP114]], align 8
4401 // CHECK3-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 10
4402 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP115]], align 4
4403 // CHECK3-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 11
4404 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP116]], align 4
4405 // CHECK3-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 12
4406 // CHECK3-NEXT:    store i32 0, ptr [[TMP117]], align 4
4407 // CHECK3-NEXT:    [[TMP118:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.region_id, ptr [[KERNEL_ARGS33]])
4408 // CHECK3-NEXT:    [[TMP119:%.*]] = icmp ne i32 [[TMP118]], 0
4409 // CHECK3-NEXT:    br i1 [[TMP119]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
4410 // CHECK3:       omp_offload.failed34:
4411 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238(ptr [[BIT_VAR]]) #[[ATTR2]]
4412 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT35]]
4413 // CHECK3:       omp_offload.cont35:
4414 // CHECK3-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
4415 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP120]], align 4
4416 // CHECK3-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
4417 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP121]], align 4
4418 // CHECK3-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS38]], i32 0, i32 0
4419 // CHECK3-NEXT:    store ptr null, ptr [[TMP122]], align 4
4420 // CHECK3-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
4421 // CHECK3-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
4422 // CHECK3-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 0
4423 // CHECK3-NEXT:    store i32 2, ptr [[TMP125]], align 4
4424 // CHECK3-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 1
4425 // CHECK3-NEXT:    store i32 1, ptr [[TMP126]], align 4
4426 // CHECK3-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 2
4427 // CHECK3-NEXT:    store ptr [[TMP123]], ptr [[TMP127]], align 4
4428 // CHECK3-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 3
4429 // CHECK3-NEXT:    store ptr [[TMP124]], ptr [[TMP128]], align 4
4430 // CHECK3-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 4
4431 // CHECK3-NEXT:    store ptr @.offload_sizes.11, ptr [[TMP129]], align 4
4432 // CHECK3-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 5
4433 // CHECK3-NEXT:    store ptr @.offload_maptypes.12, ptr [[TMP130]], align 4
4434 // CHECK3-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 6
4435 // CHECK3-NEXT:    store ptr null, ptr [[TMP131]], align 4
4436 // CHECK3-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 7
4437 // CHECK3-NEXT:    store ptr null, ptr [[TMP132]], align 4
4438 // CHECK3-NEXT:    [[TMP133:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 8
4439 // CHECK3-NEXT:    store i64 2, ptr [[TMP133]], align 8
4440 // CHECK3-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 9
4441 // CHECK3-NEXT:    store i64 0, ptr [[TMP134]], align 8
4442 // CHECK3-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 10
4443 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP135]], align 4
4444 // CHECK3-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 11
4445 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP136]], align 4
4446 // CHECK3-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 12
4447 // CHECK3-NEXT:    store i32 0, ptr [[TMP137]], align 4
4448 // CHECK3-NEXT:    [[TMP138:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.region_id, ptr [[KERNEL_ARGS40]])
4449 // CHECK3-NEXT:    [[TMP139:%.*]] = icmp ne i32 [[TMP138]], 0
4450 // CHECK3-NEXT:    br i1 [[TMP139]], label [[OMP_OFFLOAD_FAILED41:%.*]], label [[OMP_OFFLOAD_CONT42:%.*]]
4451 // CHECK3:       omp_offload.failed41:
4452 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243(ptr [[BIT_VAR]]) #[[ATTR2]]
4453 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT42]]
4454 // CHECK3:       omp_offload.cont42:
4455 // CHECK3-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
4456 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP140]], align 4
4457 // CHECK3-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
4458 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP141]], align 4
4459 // CHECK3-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS45]], i32 0, i32 0
4460 // CHECK3-NEXT:    store ptr null, ptr [[TMP142]], align 4
4461 // CHECK3-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
4462 // CHECK3-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
4463 // CHECK3-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 0
4464 // CHECK3-NEXT:    store i32 2, ptr [[TMP145]], align 4
4465 // CHECK3-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 1
4466 // CHECK3-NEXT:    store i32 1, ptr [[TMP146]], align 4
4467 // CHECK3-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 2
4468 // CHECK3-NEXT:    store ptr [[TMP143]], ptr [[TMP147]], align 4
4469 // CHECK3-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 3
4470 // CHECK3-NEXT:    store ptr [[TMP144]], ptr [[TMP148]], align 4
4471 // CHECK3-NEXT:    [[TMP149:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 4
4472 // CHECK3-NEXT:    store ptr @.offload_sizes.13, ptr [[TMP149]], align 4
4473 // CHECK3-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 5
4474 // CHECK3-NEXT:    store ptr @.offload_maptypes.14, ptr [[TMP150]], align 4
4475 // CHECK3-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 6
4476 // CHECK3-NEXT:    store ptr null, ptr [[TMP151]], align 4
4477 // CHECK3-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 7
4478 // CHECK3-NEXT:    store ptr null, ptr [[TMP152]], align 4
4479 // CHECK3-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 8
4480 // CHECK3-NEXT:    store i64 2, ptr [[TMP153]], align 8
4481 // CHECK3-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 9
4482 // CHECK3-NEXT:    store i64 0, ptr [[TMP154]], align 8
4483 // CHECK3-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 10
4484 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP155]], align 4
4485 // CHECK3-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 11
4486 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP156]], align 4
4487 // CHECK3-NEXT:    [[TMP157:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 12
4488 // CHECK3-NEXT:    store i32 0, ptr [[TMP157]], align 4
4489 // CHECK3-NEXT:    [[TMP158:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.region_id, ptr [[KERNEL_ARGS47]])
4490 // CHECK3-NEXT:    [[TMP159:%.*]] = icmp ne i32 [[TMP158]], 0
4491 // CHECK3-NEXT:    br i1 [[TMP159]], label [[OMP_OFFLOAD_FAILED48:%.*]], label [[OMP_OFFLOAD_CONT49:%.*]]
4492 // CHECK3:       omp_offload.failed48:
4493 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248(ptr [[BIT_VAR]]) #[[ATTR2]]
4494 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT49]]
4495 // CHECK3:       omp_offload.cont49:
4496 // CHECK3-NEXT:    store i32 0, ptr @_ZZ4mainE5sivar, align 4
4497 // CHECK3-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
4498 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP160]], align 4
4499 // CHECK3-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
4500 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP161]], align 4
4501 // CHECK3-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 0
4502 // CHECK3-NEXT:    store ptr null, ptr [[TMP162]], align 4
4503 // CHECK3-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
4504 // CHECK3-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
4505 // CHECK3-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 0
4506 // CHECK3-NEXT:    store i32 2, ptr [[TMP165]], align 4
4507 // CHECK3-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 1
4508 // CHECK3-NEXT:    store i32 1, ptr [[TMP166]], align 4
4509 // CHECK3-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 2
4510 // CHECK3-NEXT:    store ptr [[TMP163]], ptr [[TMP167]], align 4
4511 // CHECK3-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 3
4512 // CHECK3-NEXT:    store ptr [[TMP164]], ptr [[TMP168]], align 4
4513 // CHECK3-NEXT:    [[TMP169:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 4
4514 // CHECK3-NEXT:    store ptr @.offload_sizes.15, ptr [[TMP169]], align 4
4515 // CHECK3-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 5
4516 // CHECK3-NEXT:    store ptr @.offload_maptypes.16, ptr [[TMP170]], align 4
4517 // CHECK3-NEXT:    [[TMP171:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 6
4518 // CHECK3-NEXT:    store ptr null, ptr [[TMP171]], align 4
4519 // CHECK3-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 7
4520 // CHECK3-NEXT:    store ptr null, ptr [[TMP172]], align 4
4521 // CHECK3-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 8
4522 // CHECK3-NEXT:    store i64 2, ptr [[TMP173]], align 8
4523 // CHECK3-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 9
4524 // CHECK3-NEXT:    store i64 0, ptr [[TMP174]], align 8
4525 // CHECK3-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 10
4526 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP175]], align 4
4527 // CHECK3-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 11
4528 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP176]], align 4
4529 // CHECK3-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 12
4530 // CHECK3-NEXT:    store i32 0, ptr [[TMP177]], align 4
4531 // CHECK3-NEXT:    [[TMP178:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.region_id, ptr [[KERNEL_ARGS54]])
4532 // CHECK3-NEXT:    [[TMP179:%.*]] = icmp ne i32 [[TMP178]], 0
4533 // CHECK3-NEXT:    br i1 [[TMP179]], label [[OMP_OFFLOAD_FAILED55:%.*]], label [[OMP_OFFLOAD_CONT56:%.*]]
4534 // CHECK3:       omp_offload.failed55:
4535 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
4536 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT56]]
4537 // CHECK3:       omp_offload.cont56:
4538 // CHECK3-NEXT:    store i32 10, ptr @_ZZ4mainE5sivar, align 4
4539 // CHECK3-NEXT:    [[TMP180:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
4540 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP180]], align 4
4541 // CHECK3-NEXT:    [[TMP181:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
4542 // CHECK3-NEXT:    store ptr @_ZZ4mainE5sivar, ptr [[TMP181]], align 4
4543 // CHECK3-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS59]], i32 0, i32 0
4544 // CHECK3-NEXT:    store ptr null, ptr [[TMP182]], align 4
4545 // CHECK3-NEXT:    [[TMP183:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
4546 // CHECK3-NEXT:    [[TMP184:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
4547 // CHECK3-NEXT:    [[TMP185:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 0
4548 // CHECK3-NEXT:    store i32 2, ptr [[TMP185]], align 4
4549 // CHECK3-NEXT:    [[TMP186:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 1
4550 // CHECK3-NEXT:    store i32 1, ptr [[TMP186]], align 4
4551 // CHECK3-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 2
4552 // CHECK3-NEXT:    store ptr [[TMP183]], ptr [[TMP187]], align 4
4553 // CHECK3-NEXT:    [[TMP188:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 3
4554 // CHECK3-NEXT:    store ptr [[TMP184]], ptr [[TMP188]], align 4
4555 // CHECK3-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 4
4556 // CHECK3-NEXT:    store ptr @.offload_sizes.17, ptr [[TMP189]], align 4
4557 // CHECK3-NEXT:    [[TMP190:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 5
4558 // CHECK3-NEXT:    store ptr @.offload_maptypes.18, ptr [[TMP190]], align 4
4559 // CHECK3-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 6
4560 // CHECK3-NEXT:    store ptr null, ptr [[TMP191]], align 4
4561 // CHECK3-NEXT:    [[TMP192:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 7
4562 // CHECK3-NEXT:    store ptr null, ptr [[TMP192]], align 4
4563 // CHECK3-NEXT:    [[TMP193:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 8
4564 // CHECK3-NEXT:    store i64 2, ptr [[TMP193]], align 8
4565 // CHECK3-NEXT:    [[TMP194:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 9
4566 // CHECK3-NEXT:    store i64 0, ptr [[TMP194]], align 8
4567 // CHECK3-NEXT:    [[TMP195:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 10
4568 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP195]], align 4
4569 // CHECK3-NEXT:    [[TMP196:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 11
4570 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP196]], align 4
4571 // CHECK3-NEXT:    [[TMP197:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 12
4572 // CHECK3-NEXT:    store i32 0, ptr [[TMP197]], align 4
4573 // CHECK3-NEXT:    [[TMP198:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.region_id, ptr [[KERNEL_ARGS61]])
4574 // CHECK3-NEXT:    [[TMP199:%.*]] = icmp ne i32 [[TMP198]], 0
4575 // CHECK3-NEXT:    br i1 [[TMP199]], label [[OMP_OFFLOAD_FAILED62:%.*]], label [[OMP_OFFLOAD_CONT63:%.*]]
4576 // CHECK3:       omp_offload.failed62:
4577 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260(ptr @_ZZ4mainE5sivar) #[[ATTR2]]
4578 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT63]]
4579 // CHECK3:       omp_offload.cont63:
4580 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v()
4581 // CHECK3-NEXT:    ret i32 [[CALL]]
4582 //
4583 //
4584 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209
4585 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] {
4586 // CHECK3-NEXT:  entry:
4587 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
4588 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
4589 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
4590 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined, ptr [[TMP0]])
4591 // CHECK3-NEXT:    ret void
4592 //
4593 //
4594 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined
4595 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
4596 // CHECK3-NEXT:  entry:
4597 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
4598 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
4599 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
4600 // CHECK3-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
4601 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4602 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4603 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4604 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4605 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4606 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4607 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4608 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
4609 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
4610 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
4611 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
4612 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
4613 // CHECK3-NEXT:    store i32 0, ptr [[SIVAR1]], align 4
4614 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
4615 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
4616 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
4617 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
4618 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
4619 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
4620 // 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)
4621 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4622 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
4623 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4624 // CHECK3:       cond.true:
4625 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4626 // CHECK3:       cond.false:
4627 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4628 // CHECK3-NEXT:    br label [[COND_END]]
4629 // CHECK3:       cond.end:
4630 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4631 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
4632 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
4633 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
4634 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4635 // CHECK3:       omp.inner.for.cond:
4636 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4637 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4638 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4639 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4640 // CHECK3:       omp.inner.for.body:
4641 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4642 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
4643 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4644 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
4645 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
4646 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
4647 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
4648 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[SIVAR1]], align 4
4649 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4650 // CHECK3:       omp.body.continue:
4651 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4652 // CHECK3:       omp.inner.for.inc:
4653 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4654 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
4655 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
4656 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
4657 // CHECK3:       omp.inner.for.end:
4658 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4659 // CHECK3:       omp.loop.exit:
4660 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
4661 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
4662 // CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 4
4663 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
4664 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
4665 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
4666 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
4667 // CHECK3-NEXT:    ]
4668 // CHECK3:       .omp.reduction.case1:
4669 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
4670 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
4671 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4672 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
4673 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
4674 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4675 // CHECK3:       .omp.reduction.case2:
4676 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
4677 // CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
4678 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
4679 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4680 // CHECK3:       .omp.reduction.default:
4681 // CHECK3-NEXT:    ret void
4682 //
4683 //
4684 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l209.omp_outlined.omp.reduction.reduction_func
4685 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
4686 // CHECK3-NEXT:  entry:
4687 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
4688 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
4689 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
4690 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
4691 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
4692 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
4693 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
4694 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
4695 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
4696 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
4697 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
4698 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
4699 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
4700 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
4701 // CHECK3-NEXT:    ret void
4702 //
4703 //
4704 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214
4705 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
4706 // CHECK3-NEXT:  entry:
4707 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
4708 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
4709 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
4710 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined, ptr [[TMP0]])
4711 // CHECK3-NEXT:    ret void
4712 //
4713 //
4714 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined
4715 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
4716 // CHECK3-NEXT:  entry:
4717 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
4718 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
4719 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
4720 // CHECK3-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
4721 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4722 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4723 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4724 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4725 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4726 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4727 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4728 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
4729 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
4730 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
4731 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
4732 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
4733 // CHECK3-NEXT:    store i32 0, ptr [[SIVAR1]], align 4
4734 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
4735 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
4736 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
4737 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
4738 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
4739 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
4740 // 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)
4741 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4742 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
4743 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4744 // CHECK3:       cond.true:
4745 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4746 // CHECK3:       cond.false:
4747 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4748 // CHECK3-NEXT:    br label [[COND_END]]
4749 // CHECK3:       cond.end:
4750 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4751 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
4752 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
4753 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
4754 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4755 // CHECK3:       omp.inner.for.cond:
4756 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4757 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4758 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4759 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4760 // CHECK3:       omp.inner.for.body:
4761 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4762 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
4763 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4764 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
4765 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
4766 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
4767 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP10]], [[TMP9]]
4768 // CHECK3-NEXT:    store i32 [[SUB]], ptr [[SIVAR1]], align 4
4769 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4770 // CHECK3:       omp.body.continue:
4771 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4772 // CHECK3:       omp.inner.for.inc:
4773 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4774 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
4775 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
4776 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
4777 // CHECK3:       omp.inner.for.end:
4778 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4779 // CHECK3:       omp.loop.exit:
4780 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
4781 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
4782 // CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 4
4783 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
4784 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
4785 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
4786 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
4787 // CHECK3-NEXT:    ]
4788 // CHECK3:       .omp.reduction.case1:
4789 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
4790 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
4791 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4792 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[TMP0]], align 4
4793 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
4794 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4795 // CHECK3:       .omp.reduction.case2:
4796 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
4797 // CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
4798 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
4799 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4800 // CHECK3:       .omp.reduction.default:
4801 // CHECK3-NEXT:    ret void
4802 //
4803 //
4804 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l214.omp_outlined.omp.reduction.reduction_func
4805 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
4806 // CHECK3-NEXT:  entry:
4807 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
4808 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
4809 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
4810 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
4811 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
4812 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
4813 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
4814 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
4815 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
4816 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
4817 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
4818 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
4819 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
4820 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
4821 // CHECK3-NEXT:    ret void
4822 //
4823 //
4824 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220
4825 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
4826 // CHECK3-NEXT:  entry:
4827 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
4828 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
4829 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
4830 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined, ptr [[TMP0]])
4831 // CHECK3-NEXT:    ret void
4832 //
4833 //
4834 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined
4835 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
4836 // CHECK3-NEXT:  entry:
4837 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
4838 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
4839 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
4840 // CHECK3-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
4841 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4842 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4843 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4844 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4845 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4846 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4847 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4848 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
4849 // CHECK3-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i32, align 4
4850 // CHECK3-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
4851 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
4852 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
4853 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
4854 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
4855 // CHECK3-NEXT:    store i32 1, ptr [[SIVAR1]], align 4
4856 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
4857 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
4858 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
4859 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
4860 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
4861 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
4862 // 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)
4863 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4864 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
4865 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4866 // CHECK3:       cond.true:
4867 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4868 // CHECK3:       cond.false:
4869 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4870 // CHECK3-NEXT:    br label [[COND_END]]
4871 // CHECK3:       cond.end:
4872 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4873 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
4874 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
4875 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
4876 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4877 // CHECK3:       omp.inner.for.cond:
4878 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4879 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
4880 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4881 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4882 // CHECK3:       omp.inner.for.body:
4883 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4884 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
4885 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4886 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
4887 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[SIVAR1]], align 4
4888 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
4889 // CHECK3-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[TMP9]], [[TMP10]]
4890 // CHECK3-NEXT:    store i32 [[MUL3]], ptr [[SIVAR1]], align 4
4891 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4892 // CHECK3:       omp.body.continue:
4893 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4894 // CHECK3:       omp.inner.for.inc:
4895 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
4896 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
4897 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
4898 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
4899 // CHECK3:       omp.inner.for.end:
4900 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4901 // CHECK3:       omp.loop.exit:
4902 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
4903 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
4904 // CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP12]], align 4
4905 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
4906 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
4907 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
4908 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
4909 // CHECK3-NEXT:    ]
4910 // CHECK3:       .omp.reduction.case1:
4911 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
4912 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[SIVAR1]], align 4
4913 // CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP14]], [[TMP15]]
4914 // CHECK3-NEXT:    store i32 [[MUL5]], ptr [[TMP0]], align 4
4915 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
4916 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4917 // CHECK3:       .omp.reduction.case2:
4918 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
4919 // CHECK3-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, ptr [[TMP0]] monotonic, align 4
4920 // CHECK3-NEXT:    br label [[ATOMIC_CONT:%.*]]
4921 // CHECK3:       atomic_cont:
4922 // CHECK3-NEXT:    [[TMP17:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP22:%.*]], [[ATOMIC_CONT]] ]
4923 // CHECK3-NEXT:    store i32 [[TMP17]], ptr [[_TMP6]], align 4
4924 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[_TMP6]], align 4
4925 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[SIVAR1]], align 4
4926 // CHECK3-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[TMP18]], [[TMP19]]
4927 // CHECK3-NEXT:    store i32 [[MUL7]], ptr [[ATOMIC_TEMP]], align 4
4928 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
4929 // CHECK3-NEXT:    [[TMP21:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP17]], i32 [[TMP20]] monotonic monotonic, align 4
4930 // CHECK3-NEXT:    [[TMP22]] = extractvalue { i32, i1 } [[TMP21]], 0
4931 // CHECK3-NEXT:    [[TMP23:%.*]] = extractvalue { i32, i1 } [[TMP21]], 1
4932 // CHECK3-NEXT:    br i1 [[TMP23]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
4933 // CHECK3:       atomic_exit:
4934 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
4935 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
4936 // CHECK3:       .omp.reduction.default:
4937 // CHECK3-NEXT:    ret void
4938 //
4939 //
4940 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l220.omp_outlined.omp.reduction.reduction_func
4941 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
4942 // CHECK3-NEXT:  entry:
4943 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
4944 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
4945 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
4946 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
4947 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
4948 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
4949 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
4950 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
4951 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
4952 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
4953 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
4954 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
4955 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], [[TMP9]]
4956 // CHECK3-NEXT:    store i32 [[MUL]], ptr [[TMP7]], align 4
4957 // CHECK3-NEXT:    ret void
4958 //
4959 //
4960 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226
4961 // CHECK3-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]]) #[[ATTR1]] {
4962 // CHECK3-NEXT:  entry:
4963 // CHECK3-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 4
4964 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 4
4965 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 4
4966 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined, ptr [[TMP0]])
4967 // CHECK3-NEXT:    ret void
4968 //
4969 //
4970 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined
4971 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]]) #[[ATTR1]] {
4972 // CHECK3-NEXT:  entry:
4973 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
4974 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
4975 // CHECK3-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 4
4976 // CHECK3-NEXT:    [[AND_VAR1:%.*]] = alloca i8, align 1
4977 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4978 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4979 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4980 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4981 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4982 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4983 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4984 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
4985 // CHECK3-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
4986 // CHECK3-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
4987 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
4988 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
4989 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 4
4990 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 4
4991 // CHECK3-NEXT:    store i8 1, ptr [[AND_VAR1]], align 1
4992 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
4993 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
4994 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
4995 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
4996 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
4997 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
4998 // 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)
4999 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5000 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
5001 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5002 // CHECK3:       cond.true:
5003 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5004 // CHECK3:       cond.false:
5005 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5006 // CHECK3-NEXT:    br label [[COND_END]]
5007 // CHECK3:       cond.end:
5008 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5009 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
5010 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
5011 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
5012 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5013 // CHECK3:       omp.inner.for.cond:
5014 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5015 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5016 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5017 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5018 // CHECK3:       omp.inner.for.body:
5019 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5020 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5021 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5022 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
5023 // CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[AND_VAR1]], align 1
5024 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP9]] to i1
5025 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
5026 // CHECK3:       land.rhs:
5027 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
5028 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP10]], 0
5029 // CHECK3-NEXT:    br label [[LAND_END]]
5030 // CHECK3:       land.end:
5031 // CHECK3-NEXT:    [[TMP11:%.*]] = phi i1 [ false, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LAND_RHS]] ]
5032 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP11]] to i8
5033 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[AND_VAR1]], align 1
5034 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5035 // CHECK3:       omp.body.continue:
5036 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5037 // CHECK3:       omp.inner.for.inc:
5038 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5039 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
5040 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
5041 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5042 // CHECK3:       omp.inner.for.end:
5043 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5044 // CHECK3:       omp.loop.exit:
5045 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
5046 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
5047 // CHECK3-NEXT:    store ptr [[AND_VAR1]], ptr [[TMP13]], align 4
5048 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
5049 // CHECK3-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5050 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5051 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5052 // CHECK3-NEXT:    ]
5053 // CHECK3:       .omp.reduction.case1:
5054 // CHECK3-NEXT:    [[TMP15:%.*]] = load i8, ptr [[TMP0]], align 1
5055 // CHECK3-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP15]] to i1
5056 // CHECK3-NEXT:    br i1 [[TOBOOL5]], label [[LAND_RHS6:%.*]], label [[LAND_END8:%.*]]
5057 // CHECK3:       land.rhs6:
5058 // CHECK3-NEXT:    [[TMP16:%.*]] = load i8, ptr [[AND_VAR1]], align 1
5059 // CHECK3-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP16]] to i1
5060 // CHECK3-NEXT:    br label [[LAND_END8]]
5061 // CHECK3:       land.end8:
5062 // CHECK3-NEXT:    [[TMP17:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LAND_RHS6]] ]
5063 // CHECK3-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP17]] to i8
5064 // CHECK3-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
5065 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5066 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5067 // CHECK3:       .omp.reduction.case2:
5068 // CHECK3-NEXT:    [[TMP18:%.*]] = load i8, ptr [[AND_VAR1]], align 1
5069 // CHECK3-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP18]] to i1
5070 // CHECK3-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
5071 // CHECK3-NEXT:    br label [[ATOMIC_CONT:%.*]]
5072 // CHECK3:       atomic_cont:
5073 // CHECK3-NEXT:    [[TMP19:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP25:%.*]], [[LAND_END17:%.*]] ]
5074 // CHECK3-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP19]] to i1
5075 // CHECK3-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
5076 // CHECK3-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
5077 // CHECK3-NEXT:    [[TMP20:%.*]] = load i8, ptr [[_TMP12]], align 1
5078 // CHECK3-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP20]] to i1
5079 // CHECK3-NEXT:    br i1 [[TOBOOL14]], label [[LAND_RHS15:%.*]], label [[LAND_END17]]
5080 // CHECK3:       land.rhs15:
5081 // CHECK3-NEXT:    [[TMP21:%.*]] = load i8, ptr [[AND_VAR1]], align 1
5082 // CHECK3-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP21]] to i1
5083 // CHECK3-NEXT:    br label [[LAND_END17]]
5084 // CHECK3:       land.end17:
5085 // CHECK3-NEXT:    [[TMP22:%.*]] = phi i1 [ false, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LAND_RHS15]] ]
5086 // CHECK3-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP22]] to i8
5087 // CHECK3-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
5088 // CHECK3-NEXT:    [[TMP23:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
5089 // CHECK3-NEXT:    [[TMP24:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP19]], i8 [[TMP23]] monotonic monotonic, align 1
5090 // CHECK3-NEXT:    [[TMP25]] = extractvalue { i8, i1 } [[TMP24]], 0
5091 // CHECK3-NEXT:    [[TMP26:%.*]] = extractvalue { i8, i1 } [[TMP24]], 1
5092 // CHECK3-NEXT:    br i1 [[TMP26]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
5093 // CHECK3:       atomic_exit:
5094 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5095 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5096 // CHECK3:       .omp.reduction.default:
5097 // CHECK3-NEXT:    ret void
5098 //
5099 //
5100 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l226.omp_outlined.omp.reduction.reduction_func
5101 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
5102 // CHECK3-NEXT:  entry:
5103 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
5104 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
5105 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
5106 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
5107 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
5108 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
5109 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
5110 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
5111 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
5112 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
5113 // CHECK3-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
5114 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
5115 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
5116 // CHECK3:       land.rhs:
5117 // CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
5118 // CHECK3-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
5119 // CHECK3-NEXT:    br label [[LAND_END]]
5120 // CHECK3:       land.end:
5121 // CHECK3-NEXT:    [[TMP10:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LAND_RHS]] ]
5122 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
5123 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
5124 // CHECK3-NEXT:    ret void
5125 //
5126 //
5127 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232
5128 // CHECK3-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]]) #[[ATTR1]] {
5129 // CHECK3-NEXT:  entry:
5130 // CHECK3-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 4
5131 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 4
5132 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 4
5133 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined, ptr [[TMP0]])
5134 // CHECK3-NEXT:    ret void
5135 //
5136 //
5137 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined
5138 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]]) #[[ATTR1]] {
5139 // CHECK3-NEXT:  entry:
5140 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
5141 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
5142 // CHECK3-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 4
5143 // CHECK3-NEXT:    [[OR_VAR1:%.*]] = alloca i8, align 1
5144 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5145 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5146 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5147 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5148 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5149 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5150 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5151 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
5152 // CHECK3-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
5153 // CHECK3-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
5154 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
5155 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
5156 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 4
5157 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 4
5158 // CHECK3-NEXT:    store i8 0, ptr [[OR_VAR1]], align 1
5159 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
5160 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
5161 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
5162 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
5163 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
5164 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
5165 // 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)
5166 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5167 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
5168 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5169 // CHECK3:       cond.true:
5170 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5171 // CHECK3:       cond.false:
5172 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5173 // CHECK3-NEXT:    br label [[COND_END]]
5174 // CHECK3:       cond.end:
5175 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5176 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
5177 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
5178 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
5179 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5180 // CHECK3:       omp.inner.for.cond:
5181 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5182 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5183 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5184 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5185 // CHECK3:       omp.inner.for.body:
5186 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5187 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5188 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5189 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
5190 // CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[OR_VAR1]], align 1
5191 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP9]] to i1
5192 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
5193 // CHECK3:       lor.rhs:
5194 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
5195 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP10]], 0
5196 // CHECK3-NEXT:    br label [[LOR_END]]
5197 // CHECK3:       lor.end:
5198 // CHECK3-NEXT:    [[TMP11:%.*]] = phi i1 [ true, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LOR_RHS]] ]
5199 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP11]] to i8
5200 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[OR_VAR1]], align 1
5201 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5202 // CHECK3:       omp.body.continue:
5203 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5204 // CHECK3:       omp.inner.for.inc:
5205 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5206 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
5207 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
5208 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5209 // CHECK3:       omp.inner.for.end:
5210 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5211 // CHECK3:       omp.loop.exit:
5212 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
5213 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
5214 // CHECK3-NEXT:    store ptr [[OR_VAR1]], ptr [[TMP13]], align 4
5215 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
5216 // CHECK3-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5217 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5218 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5219 // CHECK3-NEXT:    ]
5220 // CHECK3:       .omp.reduction.case1:
5221 // CHECK3-NEXT:    [[TMP15:%.*]] = load i8, ptr [[TMP0]], align 1
5222 // CHECK3-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP15]] to i1
5223 // CHECK3-NEXT:    br i1 [[TOBOOL5]], label [[LOR_END8:%.*]], label [[LOR_RHS6:%.*]]
5224 // CHECK3:       lor.rhs6:
5225 // CHECK3-NEXT:    [[TMP16:%.*]] = load i8, ptr [[OR_VAR1]], align 1
5226 // CHECK3-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP16]] to i1
5227 // CHECK3-NEXT:    br label [[LOR_END8]]
5228 // CHECK3:       lor.end8:
5229 // CHECK3-NEXT:    [[TMP17:%.*]] = phi i1 [ true, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LOR_RHS6]] ]
5230 // CHECK3-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP17]] to i8
5231 // CHECK3-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
5232 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5233 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5234 // CHECK3:       .omp.reduction.case2:
5235 // CHECK3-NEXT:    [[TMP18:%.*]] = load i8, ptr [[OR_VAR1]], align 1
5236 // CHECK3-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP18]] to i1
5237 // CHECK3-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
5238 // CHECK3-NEXT:    br label [[ATOMIC_CONT:%.*]]
5239 // CHECK3:       atomic_cont:
5240 // CHECK3-NEXT:    [[TMP19:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP25:%.*]], [[LOR_END17:%.*]] ]
5241 // CHECK3-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP19]] to i1
5242 // CHECK3-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
5243 // CHECK3-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
5244 // CHECK3-NEXT:    [[TMP20:%.*]] = load i8, ptr [[_TMP12]], align 1
5245 // CHECK3-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP20]] to i1
5246 // CHECK3-NEXT:    br i1 [[TOBOOL14]], label [[LOR_END17]], label [[LOR_RHS15:%.*]]
5247 // CHECK3:       lor.rhs15:
5248 // CHECK3-NEXT:    [[TMP21:%.*]] = load i8, ptr [[OR_VAR1]], align 1
5249 // CHECK3-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP21]] to i1
5250 // CHECK3-NEXT:    br label [[LOR_END17]]
5251 // CHECK3:       lor.end17:
5252 // CHECK3-NEXT:    [[TMP22:%.*]] = phi i1 [ true, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LOR_RHS15]] ]
5253 // CHECK3-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP22]] to i8
5254 // CHECK3-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
5255 // CHECK3-NEXT:    [[TMP23:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
5256 // CHECK3-NEXT:    [[TMP24:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP19]], i8 [[TMP23]] monotonic monotonic, align 1
5257 // CHECK3-NEXT:    [[TMP25]] = extractvalue { i8, i1 } [[TMP24]], 0
5258 // CHECK3-NEXT:    [[TMP26:%.*]] = extractvalue { i8, i1 } [[TMP24]], 1
5259 // CHECK3-NEXT:    br i1 [[TMP26]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
5260 // CHECK3:       atomic_exit:
5261 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5262 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5263 // CHECK3:       .omp.reduction.default:
5264 // CHECK3-NEXT:    ret void
5265 //
5266 //
5267 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l232.omp_outlined.omp.reduction.reduction_func
5268 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
5269 // CHECK3-NEXT:  entry:
5270 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
5271 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
5272 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
5273 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
5274 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
5275 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
5276 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
5277 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
5278 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
5279 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
5280 // CHECK3-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
5281 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
5282 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
5283 // CHECK3:       lor.rhs:
5284 // CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
5285 // CHECK3-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
5286 // CHECK3-NEXT:    br label [[LOR_END]]
5287 // CHECK3:       lor.end:
5288 // CHECK3-NEXT:    [[TMP10:%.*]] = phi i1 [ true, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LOR_RHS]] ]
5289 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
5290 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
5291 // CHECK3-NEXT:    ret void
5292 //
5293 //
5294 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238
5295 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
5296 // CHECK3-NEXT:  entry:
5297 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
5298 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
5299 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
5300 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined, ptr [[TMP0]])
5301 // CHECK3-NEXT:    ret void
5302 //
5303 //
5304 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined
5305 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
5306 // CHECK3-NEXT:  entry:
5307 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
5308 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
5309 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
5310 // CHECK3-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
5311 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5312 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5313 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5314 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5315 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5316 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5317 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5318 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
5319 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
5320 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
5321 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
5322 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
5323 // CHECK3-NEXT:    store i32 -1, ptr [[BIT_VAR1]], align 4
5324 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
5325 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
5326 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
5327 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
5328 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
5329 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
5330 // 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)
5331 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5332 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
5333 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5334 // CHECK3:       cond.true:
5335 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5336 // CHECK3:       cond.false:
5337 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5338 // CHECK3-NEXT:    br label [[COND_END]]
5339 // CHECK3:       cond.end:
5340 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5341 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
5342 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
5343 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
5344 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5345 // CHECK3:       omp.inner.for.cond:
5346 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5347 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5348 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5349 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5350 // CHECK3:       omp.inner.for.body:
5351 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5352 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5353 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5354 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
5355 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5356 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
5357 // CHECK3-NEXT:    [[AND:%.*]] = and i32 [[TMP9]], [[TMP10]]
5358 // CHECK3-NEXT:    store i32 [[AND]], ptr [[BIT_VAR1]], align 4
5359 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5360 // CHECK3:       omp.body.continue:
5361 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5362 // CHECK3:       omp.inner.for.inc:
5363 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5364 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
5365 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
5366 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5367 // CHECK3:       omp.inner.for.end:
5368 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5369 // CHECK3:       omp.loop.exit:
5370 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
5371 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
5372 // CHECK3-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP12]], align 4
5373 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
5374 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5375 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5376 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5377 // CHECK3-NEXT:    ]
5378 // CHECK3:       .omp.reduction.case1:
5379 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
5380 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5381 // CHECK3-NEXT:    [[AND4:%.*]] = and i32 [[TMP14]], [[TMP15]]
5382 // CHECK3-NEXT:    store i32 [[AND4]], ptr [[TMP0]], align 4
5383 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5384 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5385 // CHECK3:       .omp.reduction.case2:
5386 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5387 // CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw and ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
5388 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5389 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5390 // CHECK3:       .omp.reduction.default:
5391 // CHECK3-NEXT:    ret void
5392 //
5393 //
5394 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l238.omp_outlined.omp.reduction.reduction_func
5395 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
5396 // CHECK3-NEXT:  entry:
5397 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
5398 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
5399 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
5400 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
5401 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
5402 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
5403 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
5404 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
5405 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
5406 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
5407 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
5408 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
5409 // CHECK3-NEXT:    [[AND:%.*]] = and i32 [[TMP8]], [[TMP9]]
5410 // CHECK3-NEXT:    store i32 [[AND]], ptr [[TMP7]], align 4
5411 // CHECK3-NEXT:    ret void
5412 //
5413 //
5414 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243
5415 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
5416 // CHECK3-NEXT:  entry:
5417 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
5418 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
5419 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
5420 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined, ptr [[TMP0]])
5421 // CHECK3-NEXT:    ret void
5422 //
5423 //
5424 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined
5425 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
5426 // CHECK3-NEXT:  entry:
5427 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
5428 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
5429 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
5430 // CHECK3-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
5431 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5432 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5433 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5434 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5435 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5436 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5437 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5438 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
5439 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
5440 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
5441 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
5442 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
5443 // CHECK3-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
5444 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
5445 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
5446 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
5447 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
5448 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
5449 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
5450 // 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)
5451 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5452 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
5453 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5454 // CHECK3:       cond.true:
5455 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5456 // CHECK3:       cond.false:
5457 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5458 // CHECK3-NEXT:    br label [[COND_END]]
5459 // CHECK3:       cond.end:
5460 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5461 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
5462 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
5463 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
5464 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5465 // CHECK3:       omp.inner.for.cond:
5466 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5467 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5468 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5469 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5470 // CHECK3:       omp.inner.for.body:
5471 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5472 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5473 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5474 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
5475 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5476 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
5477 // CHECK3-NEXT:    [[OR:%.*]] = or i32 [[TMP9]], [[TMP10]]
5478 // CHECK3-NEXT:    store i32 [[OR]], ptr [[BIT_VAR1]], align 4
5479 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5480 // CHECK3:       omp.body.continue:
5481 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5482 // CHECK3:       omp.inner.for.inc:
5483 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5484 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
5485 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
5486 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5487 // CHECK3:       omp.inner.for.end:
5488 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5489 // CHECK3:       omp.loop.exit:
5490 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
5491 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
5492 // CHECK3-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP12]], align 4
5493 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
5494 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5495 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5496 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5497 // CHECK3-NEXT:    ]
5498 // CHECK3:       .omp.reduction.case1:
5499 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
5500 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5501 // CHECK3-NEXT:    [[OR4:%.*]] = or i32 [[TMP14]], [[TMP15]]
5502 // CHECK3-NEXT:    store i32 [[OR4]], ptr [[TMP0]], align 4
5503 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5504 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5505 // CHECK3:       .omp.reduction.case2:
5506 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5507 // CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw or ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
5508 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5509 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5510 // CHECK3:       .omp.reduction.default:
5511 // CHECK3-NEXT:    ret void
5512 //
5513 //
5514 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l243.omp_outlined.omp.reduction.reduction_func
5515 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
5516 // CHECK3-NEXT:  entry:
5517 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
5518 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
5519 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
5520 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
5521 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
5522 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
5523 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
5524 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
5525 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
5526 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
5527 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
5528 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
5529 // CHECK3-NEXT:    [[OR:%.*]] = or i32 [[TMP8]], [[TMP9]]
5530 // CHECK3-NEXT:    store i32 [[OR]], ptr [[TMP7]], align 4
5531 // CHECK3-NEXT:    ret void
5532 //
5533 //
5534 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248
5535 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
5536 // CHECK3-NEXT:  entry:
5537 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
5538 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
5539 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
5540 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined, ptr [[TMP0]])
5541 // CHECK3-NEXT:    ret void
5542 //
5543 //
5544 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined
5545 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]]) #[[ATTR1]] {
5546 // CHECK3-NEXT:  entry:
5547 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
5548 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
5549 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
5550 // CHECK3-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
5551 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5552 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5553 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5554 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5555 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5556 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5557 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5558 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
5559 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
5560 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
5561 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
5562 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
5563 // CHECK3-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
5564 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
5565 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
5566 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
5567 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
5568 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
5569 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
5570 // 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)
5571 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5572 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
5573 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5574 // CHECK3:       cond.true:
5575 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5576 // CHECK3:       cond.false:
5577 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5578 // CHECK3-NEXT:    br label [[COND_END]]
5579 // CHECK3:       cond.end:
5580 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5581 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
5582 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
5583 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
5584 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5585 // CHECK3:       omp.inner.for.cond:
5586 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5587 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5588 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5589 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5590 // CHECK3:       omp.inner.for.body:
5591 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5592 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5593 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5594 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
5595 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5596 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
5597 // CHECK3-NEXT:    [[XOR:%.*]] = xor i32 [[TMP9]], [[TMP10]]
5598 // CHECK3-NEXT:    store i32 [[XOR]], ptr [[BIT_VAR1]], align 4
5599 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5600 // CHECK3:       omp.body.continue:
5601 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5602 // CHECK3:       omp.inner.for.inc:
5603 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5604 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
5605 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
5606 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5607 // CHECK3:       omp.inner.for.end:
5608 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5609 // CHECK3:       omp.loop.exit:
5610 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
5611 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
5612 // CHECK3-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP12]], align 4
5613 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
5614 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5615 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5616 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5617 // CHECK3-NEXT:    ]
5618 // CHECK3:       .omp.reduction.case1:
5619 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
5620 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5621 // CHECK3-NEXT:    [[XOR4:%.*]] = xor i32 [[TMP14]], [[TMP15]]
5622 // CHECK3-NEXT:    store i32 [[XOR4]], ptr [[TMP0]], align 4
5623 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5624 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5625 // CHECK3:       .omp.reduction.case2:
5626 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
5627 // CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw xor ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
5628 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5629 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5630 // CHECK3:       .omp.reduction.default:
5631 // CHECK3-NEXT:    ret void
5632 //
5633 //
5634 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l248.omp_outlined.omp.reduction.reduction_func
5635 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
5636 // CHECK3-NEXT:  entry:
5637 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
5638 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
5639 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
5640 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
5641 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
5642 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
5643 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
5644 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
5645 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
5646 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
5647 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
5648 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
5649 // CHECK3-NEXT:    [[XOR:%.*]] = xor i32 [[TMP8]], [[TMP9]]
5650 // CHECK3-NEXT:    store i32 [[XOR]], ptr [[TMP7]], align 4
5651 // CHECK3-NEXT:    ret void
5652 //
5653 //
5654 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254
5655 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
5656 // CHECK3-NEXT:  entry:
5657 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
5658 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
5659 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
5660 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined, ptr [[TMP0]])
5661 // CHECK3-NEXT:    ret void
5662 //
5663 //
5664 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined
5665 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
5666 // CHECK3-NEXT:  entry:
5667 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
5668 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
5669 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
5670 // CHECK3-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
5671 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5672 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5673 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5674 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5675 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5676 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5677 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5678 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
5679 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
5680 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
5681 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
5682 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
5683 // CHECK3-NEXT:    store i32 -2147483648, ptr [[SIVAR1]], align 4
5684 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
5685 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
5686 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
5687 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
5688 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
5689 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
5690 // 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)
5691 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5692 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
5693 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5694 // CHECK3:       cond.true:
5695 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5696 // CHECK3:       cond.false:
5697 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5698 // CHECK3-NEXT:    br label [[COND_END]]
5699 // CHECK3:       cond.end:
5700 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5701 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
5702 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
5703 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
5704 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5705 // CHECK3:       omp.inner.for.cond:
5706 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5707 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5708 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5709 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5710 // CHECK3:       omp.inner.for.body:
5711 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5712 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5713 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5714 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
5715 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[SIVAR1]], align 4
5716 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
5717 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sge i32 [[TMP9]], [[TMP10]]
5718 // CHECK3-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
5719 // CHECK3:       cond.true4:
5720 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[SIVAR1]], align 4
5721 // CHECK3-NEXT:    br label [[COND_END6:%.*]]
5722 // CHECK3:       cond.false5:
5723 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
5724 // CHECK3-NEXT:    br label [[COND_END6]]
5725 // CHECK3:       cond.end6:
5726 // CHECK3-NEXT:    [[COND7:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE4]] ], [ [[TMP12]], [[COND_FALSE5]] ]
5727 // CHECK3-NEXT:    store i32 [[COND7]], ptr [[SIVAR1]], align 4
5728 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5729 // CHECK3:       omp.body.continue:
5730 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5731 // CHECK3:       omp.inner.for.inc:
5732 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5733 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
5734 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
5735 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5736 // CHECK3:       omp.inner.for.end:
5737 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5738 // CHECK3:       omp.loop.exit:
5739 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
5740 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
5741 // CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP14]], align 4
5742 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
5743 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5744 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5745 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5746 // CHECK3-NEXT:    ]
5747 // CHECK3:       .omp.reduction.case1:
5748 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
5749 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
5750 // CHECK3-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP16]], [[TMP17]]
5751 // CHECK3-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
5752 // CHECK3:       cond.true10:
5753 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP0]], align 4
5754 // CHECK3-NEXT:    br label [[COND_END12:%.*]]
5755 // CHECK3:       cond.false11:
5756 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[SIVAR1]], align 4
5757 // CHECK3-NEXT:    br label [[COND_END12]]
5758 // CHECK3:       cond.end12:
5759 // CHECK3-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP18]], [[COND_TRUE10]] ], [ [[TMP19]], [[COND_FALSE11]] ]
5760 // CHECK3-NEXT:    store i32 [[COND13]], ptr [[TMP0]], align 4
5761 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5762 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5763 // CHECK3:       .omp.reduction.case2:
5764 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[SIVAR1]], align 4
5765 // CHECK3-NEXT:    [[TMP21:%.*]] = atomicrmw max ptr [[TMP0]], i32 [[TMP20]] monotonic, align 4
5766 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5767 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5768 // CHECK3:       .omp.reduction.default:
5769 // CHECK3-NEXT:    ret void
5770 //
5771 //
5772 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l254.omp_outlined.omp.reduction.reduction_func
5773 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
5774 // CHECK3-NEXT:  entry:
5775 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
5776 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
5777 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
5778 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
5779 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
5780 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
5781 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
5782 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
5783 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
5784 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
5785 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
5786 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
5787 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
5788 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5789 // CHECK3:       cond.true:
5790 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
5791 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5792 // CHECK3:       cond.false:
5793 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
5794 // CHECK3-NEXT:    br label [[COND_END]]
5795 // CHECK3:       cond.end:
5796 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
5797 // CHECK3-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
5798 // CHECK3-NEXT:    ret void
5799 //
5800 //
5801 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260
5802 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
5803 // CHECK3-NEXT:  entry:
5804 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
5805 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
5806 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
5807 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined, ptr [[TMP0]])
5808 // CHECK3-NEXT:    ret void
5809 //
5810 //
5811 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined
5812 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] {
5813 // CHECK3-NEXT:  entry:
5814 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
5815 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
5816 // CHECK3-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 4
5817 // CHECK3-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
5818 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5819 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5820 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5821 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5822 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5823 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5824 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5825 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
5826 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
5827 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
5828 // CHECK3-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 4
5829 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 4
5830 // CHECK3-NEXT:    store i32 2147483647, ptr [[SIVAR1]], align 4
5831 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
5832 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
5833 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
5834 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
5835 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
5836 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
5837 // 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)
5838 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5839 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
5840 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5841 // CHECK3:       cond.true:
5842 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5843 // CHECK3:       cond.false:
5844 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5845 // CHECK3-NEXT:    br label [[COND_END]]
5846 // CHECK3:       cond.end:
5847 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5848 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
5849 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
5850 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
5851 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5852 // CHECK3:       omp.inner.for.cond:
5853 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5854 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
5855 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5856 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5857 // CHECK3:       omp.inner.for.body:
5858 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5859 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5860 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5861 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
5862 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[SIVAR1]], align 4
5863 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
5864 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
5865 // CHECK3-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
5866 // CHECK3:       cond.true4:
5867 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[SIVAR1]], align 4
5868 // CHECK3-NEXT:    br label [[COND_END6:%.*]]
5869 // CHECK3:       cond.false5:
5870 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
5871 // CHECK3-NEXT:    br label [[COND_END6]]
5872 // CHECK3:       cond.end6:
5873 // CHECK3-NEXT:    [[COND7:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE4]] ], [ [[TMP12]], [[COND_FALSE5]] ]
5874 // CHECK3-NEXT:    store i32 [[COND7]], ptr [[SIVAR1]], align 4
5875 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5876 // CHECK3:       omp.body.continue:
5877 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5878 // CHECK3:       omp.inner.for.inc:
5879 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
5880 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
5881 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
5882 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5883 // CHECK3:       omp.inner.for.end:
5884 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5885 // CHECK3:       omp.loop.exit:
5886 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
5887 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
5888 // CHECK3-NEXT:    store ptr [[SIVAR1]], ptr [[TMP14]], align 4
5889 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
5890 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
5891 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
5892 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
5893 // CHECK3-NEXT:    ]
5894 // CHECK3:       .omp.reduction.case1:
5895 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
5896 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
5897 // CHECK3-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP16]], [[TMP17]]
5898 // CHECK3-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
5899 // CHECK3:       cond.true10:
5900 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP0]], align 4
5901 // CHECK3-NEXT:    br label [[COND_END12:%.*]]
5902 // CHECK3:       cond.false11:
5903 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[SIVAR1]], align 4
5904 // CHECK3-NEXT:    br label [[COND_END12]]
5905 // CHECK3:       cond.end12:
5906 // CHECK3-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP18]], [[COND_TRUE10]] ], [ [[TMP19]], [[COND_FALSE11]] ]
5907 // CHECK3-NEXT:    store i32 [[COND13]], ptr [[TMP0]], align 4
5908 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5909 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5910 // CHECK3:       .omp.reduction.case2:
5911 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[SIVAR1]], align 4
5912 // CHECK3-NEXT:    [[TMP21:%.*]] = atomicrmw min ptr [[TMP0]], i32 [[TMP20]] monotonic, align 4
5913 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
5914 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
5915 // CHECK3:       .omp.reduction.default:
5916 // CHECK3-NEXT:    ret void
5917 //
5918 //
5919 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l260.omp_outlined.omp.reduction.reduction_func
5920 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
5921 // CHECK3-NEXT:  entry:
5922 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
5923 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
5924 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
5925 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
5926 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
5927 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
5928 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
5929 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
5930 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
5931 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
5932 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
5933 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
5934 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], [[TMP9]]
5935 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5936 // CHECK3:       cond.true:
5937 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
5938 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5939 // CHECK3:       cond.false:
5940 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
5941 // CHECK3-NEXT:    br label [[COND_END]]
5942 // CHECK3:       cond.end:
5943 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
5944 // CHECK3-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
5945 // CHECK3-NEXT:    ret void
5946 //
5947 //
5948 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
5949 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] comdat {
5950 // CHECK3-NEXT:  entry:
5951 // CHECK3-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
5952 // CHECK3-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
5953 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
5954 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
5955 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
5956 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5957 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
5958 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [2 x ptr], align 4
5959 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [2 x ptr], align 4
5960 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [2 x ptr], align 4
5961 // CHECK3-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
5962 // CHECK3-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
5963 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x ptr], align 4
5964 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x ptr], align 4
5965 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x ptr], align 4
5966 // CHECK3-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
5967 // CHECK3-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
5968 // CHECK3-NEXT:    [[AND_VAR:%.*]] = alloca i8, align 1
5969 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS15:%.*]] = alloca [2 x ptr], align 4
5970 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS16:%.*]] = alloca [2 x ptr], align 4
5971 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS17:%.*]] = alloca [2 x ptr], align 4
5972 // CHECK3-NEXT:    [[_TMP18:%.*]] = alloca i32, align 4
5973 // CHECK3-NEXT:    [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
5974 // CHECK3-NEXT:    [[OR_VAR:%.*]] = alloca i8, align 1
5975 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [2 x ptr], align 4
5976 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS23:%.*]] = alloca [2 x ptr], align 4
5977 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [2 x ptr], align 4
5978 // CHECK3-NEXT:    [[_TMP25:%.*]] = alloca i32, align 4
5979 // CHECK3-NEXT:    [[KERNEL_ARGS26:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
5980 // CHECK3-NEXT:    [[BIT_VAR:%.*]] = alloca i32, align 4
5981 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [2 x ptr], align 4
5982 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [2 x ptr], align 4
5983 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [2 x ptr], align 4
5984 // CHECK3-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
5985 // CHECK3-NEXT:    [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
5986 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [2 x ptr], align 4
5987 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS37:%.*]] = alloca [2 x ptr], align 4
5988 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [2 x ptr], align 4
5989 // CHECK3-NEXT:    [[_TMP39:%.*]] = alloca i32, align 4
5990 // CHECK3-NEXT:    [[KERNEL_ARGS40:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
5991 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS43:%.*]] = alloca [2 x ptr], align 4
5992 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS44:%.*]] = alloca [2 x ptr], align 4
5993 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS45:%.*]] = alloca [2 x ptr], align 4
5994 // CHECK3-NEXT:    [[_TMP46:%.*]] = alloca i32, align 4
5995 // CHECK3-NEXT:    [[KERNEL_ARGS47:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
5996 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS50:%.*]] = alloca [2 x ptr], align 4
5997 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS51:%.*]] = alloca [2 x ptr], align 4
5998 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS52:%.*]] = alloca [2 x ptr], align 4
5999 // CHECK3-NEXT:    [[_TMP53:%.*]] = alloca i32, align 4
6000 // CHECK3-NEXT:    [[KERNEL_ARGS54:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
6001 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS57:%.*]] = alloca [2 x ptr], align 4
6002 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS58:%.*]] = alloca [2 x ptr], align 4
6003 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS59:%.*]] = alloca [2 x ptr], align 4
6004 // CHECK3-NEXT:    [[_TMP60:%.*]] = alloca i32, align 4
6005 // CHECK3-NEXT:    [[KERNEL_ARGS61:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
6006 // CHECK3-NEXT:    store i32 0, ptr [[T_VAR]], align 4
6007 // CHECK3-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i32 8, i1 false)
6008 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6009 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP0]], align 4
6010 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6011 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP1]], align 4
6012 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6013 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
6014 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6015 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6016 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
6017 // CHECK3-NEXT:    store i32 2, ptr [[TMP5]], align 4
6018 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
6019 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
6020 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
6021 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
6022 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
6023 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
6024 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
6025 // CHECK3-NEXT:    store ptr @.offload_sizes.19, ptr [[TMP9]], align 4
6026 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
6027 // CHECK3-NEXT:    store ptr @.offload_maptypes.20, ptr [[TMP10]], align 4
6028 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
6029 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
6030 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
6031 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
6032 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
6033 // CHECK3-NEXT:    store i64 2, ptr [[TMP13]], align 8
6034 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
6035 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
6036 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
6037 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
6038 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
6039 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
6040 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
6041 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
6042 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, ptr [[KERNEL_ARGS]])
6043 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
6044 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6045 // CHECK3:       omp_offload.failed:
6046 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(ptr [[T_VAR]]) #[[ATTR2]]
6047 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6048 // CHECK3:       omp_offload.cont:
6049 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
6050 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP20]], align 4
6051 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
6052 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP21]], align 4
6053 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
6054 // CHECK3-NEXT:    store ptr null, ptr [[TMP22]], align 4
6055 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 1
6056 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP23]], align 4
6057 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 1
6058 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP24]], align 4
6059 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 1
6060 // CHECK3-NEXT:    store ptr null, ptr [[TMP25]], align 4
6061 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
6062 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
6063 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
6064 // CHECK3-NEXT:    store i32 2, ptr [[TMP28]], align 4
6065 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
6066 // CHECK3-NEXT:    store i32 2, ptr [[TMP29]], align 4
6067 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
6068 // CHECK3-NEXT:    store ptr [[TMP26]], ptr [[TMP30]], align 4
6069 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
6070 // CHECK3-NEXT:    store ptr [[TMP27]], ptr [[TMP31]], align 4
6071 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
6072 // CHECK3-NEXT:    store ptr @.offload_sizes.21, ptr [[TMP32]], align 4
6073 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
6074 // CHECK3-NEXT:    store ptr @.offload_maptypes.22, ptr [[TMP33]], align 4
6075 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
6076 // CHECK3-NEXT:    store ptr null, ptr [[TMP34]], align 4
6077 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
6078 // CHECK3-NEXT:    store ptr null, ptr [[TMP35]], align 4
6079 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
6080 // CHECK3-NEXT:    store i64 2, ptr [[TMP36]], align 8
6081 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
6082 // CHECK3-NEXT:    store i64 0, ptr [[TMP37]], align 8
6083 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
6084 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP38]], align 4
6085 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
6086 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP39]], align 4
6087 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
6088 // CHECK3-NEXT:    store i32 0, ptr [[TMP40]], align 4
6089 // CHECK3-NEXT:    [[TMP41:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.region_id, ptr [[KERNEL_ARGS5]])
6090 // CHECK3-NEXT:    [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
6091 // CHECK3-NEXT:    br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
6092 // CHECK3:       omp_offload.failed6:
6093 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
6094 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
6095 // CHECK3:       omp_offload.cont7:
6096 // CHECK3-NEXT:    store i32 1, ptr [[T_VAR]], align 4
6097 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
6098 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP43]], align 4
6099 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
6100 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP44]], align 4
6101 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
6102 // CHECK3-NEXT:    store ptr null, ptr [[TMP45]], align 4
6103 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
6104 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP46]], align 4
6105 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
6106 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP47]], align 4
6107 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
6108 // CHECK3-NEXT:    store ptr null, ptr [[TMP48]], align 4
6109 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
6110 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
6111 // CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
6112 // CHECK3-NEXT:    store i32 2, ptr [[TMP51]], align 4
6113 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
6114 // CHECK3-NEXT:    store i32 2, ptr [[TMP52]], align 4
6115 // CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
6116 // CHECK3-NEXT:    store ptr [[TMP49]], ptr [[TMP53]], align 4
6117 // CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
6118 // CHECK3-NEXT:    store ptr [[TMP50]], ptr [[TMP54]], align 4
6119 // CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
6120 // CHECK3-NEXT:    store ptr @.offload_sizes.23, ptr [[TMP55]], align 4
6121 // CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
6122 // CHECK3-NEXT:    store ptr @.offload_maptypes.24, ptr [[TMP56]], align 4
6123 // CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
6124 // CHECK3-NEXT:    store ptr null, ptr [[TMP57]], align 4
6125 // CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
6126 // CHECK3-NEXT:    store ptr null, ptr [[TMP58]], align 4
6127 // CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
6128 // CHECK3-NEXT:    store i64 2, ptr [[TMP59]], align 8
6129 // CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
6130 // CHECK3-NEXT:    store i64 0, ptr [[TMP60]], align 8
6131 // CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
6132 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP61]], align 4
6133 // CHECK3-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
6134 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP62]], align 4
6135 // CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
6136 // CHECK3-NEXT:    store i32 0, ptr [[TMP63]], align 4
6137 // CHECK3-NEXT:    [[TMP64:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.region_id, ptr [[KERNEL_ARGS12]])
6138 // CHECK3-NEXT:    [[TMP65:%.*]] = icmp ne i32 [[TMP64]], 0
6139 // CHECK3-NEXT:    br i1 [[TMP65]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
6140 // CHECK3:       omp_offload.failed13:
6141 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
6142 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
6143 // CHECK3:       omp_offload.cont14:
6144 // CHECK3-NEXT:    store i8 1, ptr [[AND_VAR]], align 1
6145 // CHECK3-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
6146 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[TMP66]], align 4
6147 // CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
6148 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[TMP67]], align 4
6149 // CHECK3-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS17]], i32 0, i32 0
6150 // CHECK3-NEXT:    store ptr null, ptr [[TMP68]], align 4
6151 // CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 1
6152 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP69]], align 4
6153 // CHECK3-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 1
6154 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP70]], align 4
6155 // CHECK3-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS17]], i32 0, i32 1
6156 // CHECK3-NEXT:    store ptr null, ptr [[TMP71]], align 4
6157 // CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS15]], i32 0, i32 0
6158 // CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS16]], i32 0, i32 0
6159 // CHECK3-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 0
6160 // CHECK3-NEXT:    store i32 2, ptr [[TMP74]], align 4
6161 // CHECK3-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 1
6162 // CHECK3-NEXT:    store i32 2, ptr [[TMP75]], align 4
6163 // CHECK3-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 2
6164 // CHECK3-NEXT:    store ptr [[TMP72]], ptr [[TMP76]], align 4
6165 // CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 3
6166 // CHECK3-NEXT:    store ptr [[TMP73]], ptr [[TMP77]], align 4
6167 // CHECK3-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 4
6168 // CHECK3-NEXT:    store ptr @.offload_sizes.25, ptr [[TMP78]], align 4
6169 // CHECK3-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 5
6170 // CHECK3-NEXT:    store ptr @.offload_maptypes.26, ptr [[TMP79]], align 4
6171 // CHECK3-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 6
6172 // CHECK3-NEXT:    store ptr null, ptr [[TMP80]], align 4
6173 // CHECK3-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 7
6174 // CHECK3-NEXT:    store ptr null, ptr [[TMP81]], align 4
6175 // CHECK3-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 8
6176 // CHECK3-NEXT:    store i64 2, ptr [[TMP82]], align 8
6177 // CHECK3-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 9
6178 // CHECK3-NEXT:    store i64 0, ptr [[TMP83]], align 8
6179 // CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 10
6180 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP84]], align 4
6181 // CHECK3-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 11
6182 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP85]], align 4
6183 // CHECK3-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS19]], i32 0, i32 12
6184 // CHECK3-NEXT:    store i32 0, ptr [[TMP86]], align 4
6185 // CHECK3-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.region_id, ptr [[KERNEL_ARGS19]])
6186 // CHECK3-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
6187 // CHECK3-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
6188 // CHECK3:       omp_offload.failed20:
6189 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49(ptr [[AND_VAR]], ptr [[VEC]]) #[[ATTR2]]
6190 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
6191 // CHECK3:       omp_offload.cont21:
6192 // CHECK3-NEXT:    store i8 0, ptr [[OR_VAR]], align 1
6193 // CHECK3-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
6194 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[TMP89]], align 4
6195 // CHECK3-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
6196 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[TMP90]], align 4
6197 // CHECK3-NEXT:    [[TMP91:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS24]], i32 0, i32 0
6198 // CHECK3-NEXT:    store ptr null, ptr [[TMP91]], align 4
6199 // CHECK3-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 1
6200 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP92]], align 4
6201 // CHECK3-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 1
6202 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP93]], align 4
6203 // CHECK3-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS24]], i32 0, i32 1
6204 // CHECK3-NEXT:    store ptr null, ptr [[TMP94]], align 4
6205 // CHECK3-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
6206 // CHECK3-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
6207 // CHECK3-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 0
6208 // CHECK3-NEXT:    store i32 2, ptr [[TMP97]], align 4
6209 // CHECK3-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 1
6210 // CHECK3-NEXT:    store i32 2, ptr [[TMP98]], align 4
6211 // CHECK3-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 2
6212 // CHECK3-NEXT:    store ptr [[TMP95]], ptr [[TMP99]], align 4
6213 // CHECK3-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 3
6214 // CHECK3-NEXT:    store ptr [[TMP96]], ptr [[TMP100]], align 4
6215 // CHECK3-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 4
6216 // CHECK3-NEXT:    store ptr @.offload_sizes.27, ptr [[TMP101]], align 4
6217 // CHECK3-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 5
6218 // CHECK3-NEXT:    store ptr @.offload_maptypes.28, ptr [[TMP102]], align 4
6219 // CHECK3-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 6
6220 // CHECK3-NEXT:    store ptr null, ptr [[TMP103]], align 4
6221 // CHECK3-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 7
6222 // CHECK3-NEXT:    store ptr null, ptr [[TMP104]], align 4
6223 // CHECK3-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 8
6224 // CHECK3-NEXT:    store i64 2, ptr [[TMP105]], align 8
6225 // CHECK3-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 9
6226 // CHECK3-NEXT:    store i64 0, ptr [[TMP106]], align 8
6227 // CHECK3-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 10
6228 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP107]], align 4
6229 // CHECK3-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 11
6230 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP108]], align 4
6231 // CHECK3-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS26]], i32 0, i32 12
6232 // CHECK3-NEXT:    store i32 0, ptr [[TMP109]], align 4
6233 // CHECK3-NEXT:    [[TMP110:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.region_id, ptr [[KERNEL_ARGS26]])
6234 // CHECK3-NEXT:    [[TMP111:%.*]] = icmp ne i32 [[TMP110]], 0
6235 // CHECK3-NEXT:    br i1 [[TMP111]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
6236 // CHECK3:       omp_offload.failed27:
6237 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55(ptr [[OR_VAR]], ptr [[VEC]]) #[[ATTR2]]
6238 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
6239 // CHECK3:       omp_offload.cont28:
6240 // CHECK3-NEXT:    store i32 1, ptr [[BIT_VAR]], align 4
6241 // CHECK3-NEXT:    [[TMP112:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
6242 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP112]], align 4
6243 // CHECK3-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
6244 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP113]], align 4
6245 // CHECK3-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS31]], i32 0, i32 0
6246 // CHECK3-NEXT:    store ptr null, ptr [[TMP114]], align 4
6247 // CHECK3-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 1
6248 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP115]], align 4
6249 // CHECK3-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 1
6250 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP116]], align 4
6251 // CHECK3-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS31]], i32 0, i32 1
6252 // CHECK3-NEXT:    store ptr null, ptr [[TMP117]], align 4
6253 // CHECK3-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
6254 // CHECK3-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
6255 // CHECK3-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 0
6256 // CHECK3-NEXT:    store i32 2, ptr [[TMP120]], align 4
6257 // CHECK3-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 1
6258 // CHECK3-NEXT:    store i32 2, ptr [[TMP121]], align 4
6259 // CHECK3-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 2
6260 // CHECK3-NEXT:    store ptr [[TMP118]], ptr [[TMP122]], align 4
6261 // CHECK3-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 3
6262 // CHECK3-NEXT:    store ptr [[TMP119]], ptr [[TMP123]], align 4
6263 // CHECK3-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 4
6264 // CHECK3-NEXT:    store ptr @.offload_sizes.29, ptr [[TMP124]], align 4
6265 // CHECK3-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 5
6266 // CHECK3-NEXT:    store ptr @.offload_maptypes.30, ptr [[TMP125]], align 4
6267 // CHECK3-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 6
6268 // CHECK3-NEXT:    store ptr null, ptr [[TMP126]], align 4
6269 // CHECK3-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 7
6270 // CHECK3-NEXT:    store ptr null, ptr [[TMP127]], align 4
6271 // CHECK3-NEXT:    [[TMP128:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 8
6272 // CHECK3-NEXT:    store i64 2, ptr [[TMP128]], align 8
6273 // CHECK3-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 9
6274 // CHECK3-NEXT:    store i64 0, ptr [[TMP129]], align 8
6275 // CHECK3-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 10
6276 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP130]], align 4
6277 // CHECK3-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 11
6278 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP131]], align 4
6279 // CHECK3-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS33]], i32 0, i32 12
6280 // CHECK3-NEXT:    store i32 0, ptr [[TMP132]], align 4
6281 // CHECK3-NEXT:    [[TMP133:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.region_id, ptr [[KERNEL_ARGS33]])
6282 // CHECK3-NEXT:    [[TMP134:%.*]] = icmp ne i32 [[TMP133]], 0
6283 // CHECK3-NEXT:    br i1 [[TMP134]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
6284 // CHECK3:       omp_offload.failed34:
6285 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61(ptr [[BIT_VAR]], ptr [[VEC]]) #[[ATTR2]]
6286 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT35]]
6287 // CHECK3:       omp_offload.cont35:
6288 // CHECK3-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
6289 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP135]], align 4
6290 // CHECK3-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
6291 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP136]], align 4
6292 // CHECK3-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS38]], i32 0, i32 0
6293 // CHECK3-NEXT:    store ptr null, ptr [[TMP137]], align 4
6294 // CHECK3-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
6295 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP138]], align 4
6296 // CHECK3-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
6297 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP139]], align 4
6298 // CHECK3-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS38]], i32 0, i32 1
6299 // CHECK3-NEXT:    store ptr null, ptr [[TMP140]], align 4
6300 // CHECK3-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
6301 // CHECK3-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
6302 // CHECK3-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 0
6303 // CHECK3-NEXT:    store i32 2, ptr [[TMP143]], align 4
6304 // CHECK3-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 1
6305 // CHECK3-NEXT:    store i32 2, ptr [[TMP144]], align 4
6306 // CHECK3-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 2
6307 // CHECK3-NEXT:    store ptr [[TMP141]], ptr [[TMP145]], align 4
6308 // CHECK3-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 3
6309 // CHECK3-NEXT:    store ptr [[TMP142]], ptr [[TMP146]], align 4
6310 // CHECK3-NEXT:    [[TMP147:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 4
6311 // CHECK3-NEXT:    store ptr @.offload_sizes.31, ptr [[TMP147]], align 4
6312 // CHECK3-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 5
6313 // CHECK3-NEXT:    store ptr @.offload_maptypes.32, ptr [[TMP148]], align 4
6314 // CHECK3-NEXT:    [[TMP149:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 6
6315 // CHECK3-NEXT:    store ptr null, ptr [[TMP149]], align 4
6316 // CHECK3-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 7
6317 // CHECK3-NEXT:    store ptr null, ptr [[TMP150]], align 4
6318 // CHECK3-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 8
6319 // CHECK3-NEXT:    store i64 2, ptr [[TMP151]], align 8
6320 // CHECK3-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 9
6321 // CHECK3-NEXT:    store i64 0, ptr [[TMP152]], align 8
6322 // CHECK3-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 10
6323 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP153]], align 4
6324 // CHECK3-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 11
6325 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP154]], align 4
6326 // CHECK3-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS40]], i32 0, i32 12
6327 // CHECK3-NEXT:    store i32 0, ptr [[TMP155]], align 4
6328 // CHECK3-NEXT:    [[TMP156:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.region_id, ptr [[KERNEL_ARGS40]])
6329 // CHECK3-NEXT:    [[TMP157:%.*]] = icmp ne i32 [[TMP156]], 0
6330 // CHECK3-NEXT:    br i1 [[TMP157]], label [[OMP_OFFLOAD_FAILED41:%.*]], label [[OMP_OFFLOAD_CONT42:%.*]]
6331 // CHECK3:       omp_offload.failed41:
6332 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66(ptr [[BIT_VAR]], ptr [[VEC]]) #[[ATTR2]]
6333 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT42]]
6334 // CHECK3:       omp_offload.cont42:
6335 // CHECK3-NEXT:    [[TMP158:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
6336 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP158]], align 4
6337 // CHECK3-NEXT:    [[TMP159:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
6338 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[TMP159]], align 4
6339 // CHECK3-NEXT:    [[TMP160:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS45]], i32 0, i32 0
6340 // CHECK3-NEXT:    store ptr null, ptr [[TMP160]], align 4
6341 // CHECK3-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 1
6342 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP161]], align 4
6343 // CHECK3-NEXT:    [[TMP162:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 1
6344 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP162]], align 4
6345 // CHECK3-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS45]], i32 0, i32 1
6346 // CHECK3-NEXT:    store ptr null, ptr [[TMP163]], align 4
6347 // CHECK3-NEXT:    [[TMP164:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS43]], i32 0, i32 0
6348 // CHECK3-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS44]], i32 0, i32 0
6349 // CHECK3-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 0
6350 // CHECK3-NEXT:    store i32 2, ptr [[TMP166]], align 4
6351 // CHECK3-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 1
6352 // CHECK3-NEXT:    store i32 2, ptr [[TMP167]], align 4
6353 // CHECK3-NEXT:    [[TMP168:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 2
6354 // CHECK3-NEXT:    store ptr [[TMP164]], ptr [[TMP168]], align 4
6355 // CHECK3-NEXT:    [[TMP169:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 3
6356 // CHECK3-NEXT:    store ptr [[TMP165]], ptr [[TMP169]], align 4
6357 // CHECK3-NEXT:    [[TMP170:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 4
6358 // CHECK3-NEXT:    store ptr @.offload_sizes.33, ptr [[TMP170]], align 4
6359 // CHECK3-NEXT:    [[TMP171:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 5
6360 // CHECK3-NEXT:    store ptr @.offload_maptypes.34, ptr [[TMP171]], align 4
6361 // CHECK3-NEXT:    [[TMP172:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 6
6362 // CHECK3-NEXT:    store ptr null, ptr [[TMP172]], align 4
6363 // CHECK3-NEXT:    [[TMP173:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 7
6364 // CHECK3-NEXT:    store ptr null, ptr [[TMP173]], align 4
6365 // CHECK3-NEXT:    [[TMP174:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 8
6366 // CHECK3-NEXT:    store i64 2, ptr [[TMP174]], align 8
6367 // CHECK3-NEXT:    [[TMP175:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 9
6368 // CHECK3-NEXT:    store i64 0, ptr [[TMP175]], align 8
6369 // CHECK3-NEXT:    [[TMP176:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 10
6370 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP176]], align 4
6371 // CHECK3-NEXT:    [[TMP177:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 11
6372 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP177]], align 4
6373 // CHECK3-NEXT:    [[TMP178:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS47]], i32 0, i32 12
6374 // CHECK3-NEXT:    store i32 0, ptr [[TMP178]], align 4
6375 // CHECK3-NEXT:    [[TMP179:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.region_id, ptr [[KERNEL_ARGS47]])
6376 // CHECK3-NEXT:    [[TMP180:%.*]] = icmp ne i32 [[TMP179]], 0
6377 // CHECK3-NEXT:    br i1 [[TMP180]], label [[OMP_OFFLOAD_FAILED48:%.*]], label [[OMP_OFFLOAD_CONT49:%.*]]
6378 // CHECK3:       omp_offload.failed48:
6379 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71(ptr [[BIT_VAR]], ptr [[VEC]]) #[[ATTR2]]
6380 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT49]]
6381 // CHECK3:       omp_offload.cont49:
6382 // CHECK3-NEXT:    store i32 0, ptr [[T_VAR]], align 4
6383 // CHECK3-NEXT:    [[TMP181:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
6384 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP181]], align 4
6385 // CHECK3-NEXT:    [[TMP182:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
6386 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP182]], align 4
6387 // CHECK3-NEXT:    [[TMP183:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 0
6388 // CHECK3-NEXT:    store ptr null, ptr [[TMP183]], align 4
6389 // CHECK3-NEXT:    [[TMP184:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 1
6390 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP184]], align 4
6391 // CHECK3-NEXT:    [[TMP185:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 1
6392 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP185]], align 4
6393 // CHECK3-NEXT:    [[TMP186:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS52]], i32 0, i32 1
6394 // CHECK3-NEXT:    store ptr null, ptr [[TMP186]], align 4
6395 // CHECK3-NEXT:    [[TMP187:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS50]], i32 0, i32 0
6396 // CHECK3-NEXT:    [[TMP188:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS51]], i32 0, i32 0
6397 // CHECK3-NEXT:    [[TMP189:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 0
6398 // CHECK3-NEXT:    store i32 2, ptr [[TMP189]], align 4
6399 // CHECK3-NEXT:    [[TMP190:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 1
6400 // CHECK3-NEXT:    store i32 2, ptr [[TMP190]], align 4
6401 // CHECK3-NEXT:    [[TMP191:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 2
6402 // CHECK3-NEXT:    store ptr [[TMP187]], ptr [[TMP191]], align 4
6403 // CHECK3-NEXT:    [[TMP192:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 3
6404 // CHECK3-NEXT:    store ptr [[TMP188]], ptr [[TMP192]], align 4
6405 // CHECK3-NEXT:    [[TMP193:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 4
6406 // CHECK3-NEXT:    store ptr @.offload_sizes.35, ptr [[TMP193]], align 4
6407 // CHECK3-NEXT:    [[TMP194:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 5
6408 // CHECK3-NEXT:    store ptr @.offload_maptypes.36, ptr [[TMP194]], align 4
6409 // CHECK3-NEXT:    [[TMP195:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 6
6410 // CHECK3-NEXT:    store ptr null, ptr [[TMP195]], align 4
6411 // CHECK3-NEXT:    [[TMP196:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 7
6412 // CHECK3-NEXT:    store ptr null, ptr [[TMP196]], align 4
6413 // CHECK3-NEXT:    [[TMP197:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 8
6414 // CHECK3-NEXT:    store i64 2, ptr [[TMP197]], align 8
6415 // CHECK3-NEXT:    [[TMP198:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 9
6416 // CHECK3-NEXT:    store i64 0, ptr [[TMP198]], align 8
6417 // CHECK3-NEXT:    [[TMP199:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 10
6418 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP199]], align 4
6419 // CHECK3-NEXT:    [[TMP200:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 11
6420 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP200]], align 4
6421 // CHECK3-NEXT:    [[TMP201:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS54]], i32 0, i32 12
6422 // CHECK3-NEXT:    store i32 0, ptr [[TMP201]], align 4
6423 // CHECK3-NEXT:    [[TMP202:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.region_id, ptr [[KERNEL_ARGS54]])
6424 // CHECK3-NEXT:    [[TMP203:%.*]] = icmp ne i32 [[TMP202]], 0
6425 // CHECK3-NEXT:    br i1 [[TMP203]], label [[OMP_OFFLOAD_FAILED55:%.*]], label [[OMP_OFFLOAD_CONT56:%.*]]
6426 // CHECK3:       omp_offload.failed55:
6427 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
6428 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT56]]
6429 // CHECK3:       omp_offload.cont56:
6430 // CHECK3-NEXT:    store i32 10, ptr [[T_VAR]], align 4
6431 // CHECK3-NEXT:    [[TMP204:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
6432 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP204]], align 4
6433 // CHECK3-NEXT:    [[TMP205:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
6434 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[TMP205]], align 4
6435 // CHECK3-NEXT:    [[TMP206:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS59]], i32 0, i32 0
6436 // CHECK3-NEXT:    store ptr null, ptr [[TMP206]], align 4
6437 // CHECK3-NEXT:    [[TMP207:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 1
6438 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP207]], align 4
6439 // CHECK3-NEXT:    [[TMP208:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 1
6440 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[TMP208]], align 4
6441 // CHECK3-NEXT:    [[TMP209:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS59]], i32 0, i32 1
6442 // CHECK3-NEXT:    store ptr null, ptr [[TMP209]], align 4
6443 // CHECK3-NEXT:    [[TMP210:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS57]], i32 0, i32 0
6444 // CHECK3-NEXT:    [[TMP211:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS58]], i32 0, i32 0
6445 // CHECK3-NEXT:    [[TMP212:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 0
6446 // CHECK3-NEXT:    store i32 2, ptr [[TMP212]], align 4
6447 // CHECK3-NEXT:    [[TMP213:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 1
6448 // CHECK3-NEXT:    store i32 2, ptr [[TMP213]], align 4
6449 // CHECK3-NEXT:    [[TMP214:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 2
6450 // CHECK3-NEXT:    store ptr [[TMP210]], ptr [[TMP214]], align 4
6451 // CHECK3-NEXT:    [[TMP215:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 3
6452 // CHECK3-NEXT:    store ptr [[TMP211]], ptr [[TMP215]], align 4
6453 // CHECK3-NEXT:    [[TMP216:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 4
6454 // CHECK3-NEXT:    store ptr @.offload_sizes.37, ptr [[TMP216]], align 4
6455 // CHECK3-NEXT:    [[TMP217:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 5
6456 // CHECK3-NEXT:    store ptr @.offload_maptypes.38, ptr [[TMP217]], align 4
6457 // CHECK3-NEXT:    [[TMP218:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 6
6458 // CHECK3-NEXT:    store ptr null, ptr [[TMP218]], align 4
6459 // CHECK3-NEXT:    [[TMP219:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 7
6460 // CHECK3-NEXT:    store ptr null, ptr [[TMP219]], align 4
6461 // CHECK3-NEXT:    [[TMP220:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 8
6462 // CHECK3-NEXT:    store i64 2, ptr [[TMP220]], align 8
6463 // CHECK3-NEXT:    [[TMP221:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 9
6464 // CHECK3-NEXT:    store i64 0, ptr [[TMP221]], align 8
6465 // CHECK3-NEXT:    [[TMP222:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 10
6466 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP222]], align 4
6467 // CHECK3-NEXT:    [[TMP223:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 11
6468 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP223]], align 4
6469 // CHECK3-NEXT:    [[TMP224:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS61]], i32 0, i32 12
6470 // CHECK3-NEXT:    store i32 0, ptr [[TMP224]], align 4
6471 // CHECK3-NEXT:    [[TMP225:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.region_id, ptr [[KERNEL_ARGS61]])
6472 // CHECK3-NEXT:    [[TMP226:%.*]] = icmp ne i32 [[TMP225]], 0
6473 // CHECK3-NEXT:    br i1 [[TMP226]], label [[OMP_OFFLOAD_FAILED62:%.*]], label [[OMP_OFFLOAD_CONT63:%.*]]
6474 // CHECK3:       omp_offload.failed62:
6475 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83(ptr [[T_VAR]], ptr [[VEC]]) #[[ATTR2]]
6476 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT63]]
6477 // CHECK3:       omp_offload.cont63:
6478 // CHECK3-NEXT:    ret i32 0
6479 //
6480 //
6481 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32
6482 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] {
6483 // CHECK3-NEXT:  entry:
6484 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
6485 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
6486 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
6487 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined, ptr [[TMP0]])
6488 // CHECK3-NEXT:    ret void
6489 //
6490 //
6491 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined
6492 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] {
6493 // CHECK3-NEXT:  entry:
6494 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
6495 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
6496 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
6497 // CHECK3-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
6498 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6499 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6500 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6501 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6502 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6503 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6504 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
6505 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
6506 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
6507 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
6508 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
6509 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
6510 // CHECK3-NEXT:    store i32 0, ptr [[T_VAR1]], align 4
6511 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
6512 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
6513 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
6514 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
6515 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
6516 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
6517 // 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)
6518 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6519 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
6520 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6521 // CHECK3:       cond.true:
6522 // CHECK3-NEXT:    br label [[COND_END:%.*]]
6523 // CHECK3:       cond.false:
6524 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6525 // CHECK3-NEXT:    br label [[COND_END]]
6526 // CHECK3:       cond.end:
6527 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6528 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
6529 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
6530 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
6531 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6532 // CHECK3:       omp.inner.for.cond:
6533 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6534 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6535 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6536 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6537 // CHECK3:       omp.inner.for.body:
6538 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6539 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
6540 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6541 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
6542 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
6543 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
6544 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
6545 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[T_VAR1]], align 4
6546 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6547 // CHECK3:       omp.body.continue:
6548 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6549 // CHECK3:       omp.inner.for.inc:
6550 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6551 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
6552 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
6553 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
6554 // CHECK3:       omp.inner.for.end:
6555 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6556 // CHECK3:       omp.loop.exit:
6557 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
6558 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
6559 // CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP12]], align 4
6560 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
6561 // CHECK3-NEXT:    switch i32 [[TMP13]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
6562 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
6563 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
6564 // CHECK3-NEXT:    ]
6565 // CHECK3:       .omp.reduction.case1:
6566 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[TMP0]], align 4
6567 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR1]], align 4
6568 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
6569 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
6570 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
6571 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
6572 // CHECK3:       .omp.reduction.case2:
6573 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[T_VAR1]], align 4
6574 // CHECK3-NEXT:    [[TMP17:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP16]] monotonic, align 4
6575 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
6576 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
6577 // CHECK3:       .omp.reduction.default:
6578 // CHECK3-NEXT:    ret void
6579 //
6580 //
6581 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.omp_outlined.omp.reduction.reduction_func
6582 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
6583 // CHECK3-NEXT:  entry:
6584 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
6585 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
6586 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
6587 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
6588 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
6589 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
6590 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
6591 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
6592 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
6593 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
6594 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
6595 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
6596 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
6597 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
6598 // CHECK3-NEXT:    ret void
6599 //
6600 //
6601 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37
6602 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
6603 // CHECK3-NEXT:  entry:
6604 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
6605 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
6606 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
6607 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
6608 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
6609 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
6610 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
6611 // CHECK3-NEXT:    ret void
6612 //
6613 //
6614 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined
6615 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
6616 // CHECK3-NEXT:  entry:
6617 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
6618 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
6619 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
6620 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
6621 // CHECK3-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
6622 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6623 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6624 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6625 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6626 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6627 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6628 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
6629 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
6630 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
6631 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
6632 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
6633 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
6634 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
6635 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
6636 // CHECK3-NEXT:    store i32 0, ptr [[T_VAR1]], align 4
6637 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
6638 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
6639 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
6640 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
6641 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
6642 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
6643 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
6644 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6645 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
6646 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6647 // CHECK3:       cond.true:
6648 // CHECK3-NEXT:    br label [[COND_END:%.*]]
6649 // CHECK3:       cond.false:
6650 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6651 // CHECK3-NEXT:    br label [[COND_END]]
6652 // CHECK3:       cond.end:
6653 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6654 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
6655 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
6656 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
6657 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6658 // CHECK3:       omp.inner.for.cond:
6659 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6660 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6661 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6662 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6663 // CHECK3:       omp.inner.for.body:
6664 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6665 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6666 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6667 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
6668 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
6669 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP10]]
6670 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
6671 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[T_VAR1]], align 4
6672 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP12]], [[TMP11]]
6673 // CHECK3-NEXT:    store i32 [[SUB]], ptr [[T_VAR1]], align 4
6674 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6675 // CHECK3:       omp.body.continue:
6676 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6677 // CHECK3:       omp.inner.for.inc:
6678 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6679 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
6680 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
6681 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
6682 // CHECK3:       omp.inner.for.end:
6683 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6684 // CHECK3:       omp.loop.exit:
6685 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
6686 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
6687 // CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP14]], align 4
6688 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
6689 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
6690 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
6691 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
6692 // CHECK3-NEXT:    ]
6693 // CHECK3:       .omp.reduction.case1:
6694 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
6695 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR1]], align 4
6696 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
6697 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[TMP0]], align 4
6698 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
6699 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
6700 // CHECK3:       .omp.reduction.case2:
6701 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR1]], align 4
6702 // CHECK3-NEXT:    [[TMP19:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
6703 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
6704 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
6705 // CHECK3:       .omp.reduction.default:
6706 // CHECK3-NEXT:    ret void
6707 //
6708 //
6709 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l37.omp_outlined.omp.reduction.reduction_func
6710 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
6711 // CHECK3-NEXT:  entry:
6712 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
6713 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
6714 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
6715 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
6716 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
6717 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
6718 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
6719 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
6720 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
6721 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
6722 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
6723 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
6724 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
6725 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
6726 // CHECK3-NEXT:    ret void
6727 //
6728 //
6729 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43
6730 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
6731 // CHECK3-NEXT:  entry:
6732 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
6733 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
6734 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
6735 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
6736 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
6737 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
6738 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
6739 // CHECK3-NEXT:    ret void
6740 //
6741 //
6742 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined
6743 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
6744 // CHECK3-NEXT:  entry:
6745 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
6746 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
6747 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
6748 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
6749 // CHECK3-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
6750 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6751 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6752 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6753 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6754 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6755 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6756 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
6757 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
6758 // CHECK3-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i32, align 4
6759 // CHECK3-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
6760 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
6761 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
6762 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
6763 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
6764 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
6765 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
6766 // CHECK3-NEXT:    store i32 1, ptr [[T_VAR1]], align 4
6767 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
6768 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
6769 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
6770 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
6771 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
6772 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
6773 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
6774 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6775 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
6776 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6777 // CHECK3:       cond.true:
6778 // CHECK3-NEXT:    br label [[COND_END:%.*]]
6779 // CHECK3:       cond.false:
6780 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6781 // CHECK3-NEXT:    br label [[COND_END]]
6782 // CHECK3:       cond.end:
6783 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6784 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
6785 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
6786 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
6787 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6788 // CHECK3:       omp.inner.for.cond:
6789 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6790 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6791 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6792 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6793 // CHECK3:       omp.inner.for.body:
6794 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6795 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6796 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6797 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
6798 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
6799 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
6800 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
6801 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
6802 // CHECK3-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[TMP10]], [[TMP12]]
6803 // CHECK3-NEXT:    store i32 [[MUL3]], ptr [[T_VAR1]], align 4
6804 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6805 // CHECK3:       omp.body.continue:
6806 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6807 // CHECK3:       omp.inner.for.inc:
6808 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6809 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
6810 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
6811 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
6812 // CHECK3:       omp.inner.for.end:
6813 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6814 // CHECK3:       omp.loop.exit:
6815 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
6816 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
6817 // CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP14]], align 4
6818 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
6819 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
6820 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
6821 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
6822 // CHECK3-NEXT:    ]
6823 // CHECK3:       .omp.reduction.case1:
6824 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
6825 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR1]], align 4
6826 // CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP16]], [[TMP17]]
6827 // CHECK3-NEXT:    store i32 [[MUL5]], ptr [[TMP0]], align 4
6828 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
6829 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
6830 // CHECK3:       .omp.reduction.case2:
6831 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR1]], align 4
6832 // CHECK3-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, ptr [[TMP0]] monotonic, align 4
6833 // CHECK3-NEXT:    br label [[ATOMIC_CONT:%.*]]
6834 // CHECK3:       atomic_cont:
6835 // CHECK3-NEXT:    [[TMP19:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP24:%.*]], [[ATOMIC_CONT]] ]
6836 // CHECK3-NEXT:    store i32 [[TMP19]], ptr [[_TMP6]], align 4
6837 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[_TMP6]], align 4
6838 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[T_VAR1]], align 4
6839 // CHECK3-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[TMP20]], [[TMP21]]
6840 // CHECK3-NEXT:    store i32 [[MUL7]], ptr [[ATOMIC_TEMP]], align 4
6841 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
6842 // CHECK3-NEXT:    [[TMP23:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP19]], i32 [[TMP22]] monotonic monotonic, align 4
6843 // CHECK3-NEXT:    [[TMP24]] = extractvalue { i32, i1 } [[TMP23]], 0
6844 // CHECK3-NEXT:    [[TMP25:%.*]] = extractvalue { i32, i1 } [[TMP23]], 1
6845 // CHECK3-NEXT:    br i1 [[TMP25]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
6846 // CHECK3:       atomic_exit:
6847 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
6848 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
6849 // CHECK3:       .omp.reduction.default:
6850 // CHECK3-NEXT:    ret void
6851 //
6852 //
6853 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l43.omp_outlined.omp.reduction.reduction_func
6854 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
6855 // CHECK3-NEXT:  entry:
6856 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
6857 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
6858 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
6859 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
6860 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
6861 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
6862 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
6863 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
6864 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
6865 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
6866 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
6867 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
6868 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], [[TMP9]]
6869 // CHECK3-NEXT:    store i32 [[MUL]], ptr [[TMP7]], align 4
6870 // CHECK3-NEXT:    ret void
6871 //
6872 //
6873 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49
6874 // CHECK3-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
6875 // CHECK3-NEXT:  entry:
6876 // CHECK3-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 4
6877 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
6878 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 4
6879 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
6880 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 4
6881 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
6882 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
6883 // CHECK3-NEXT:    ret void
6884 //
6885 //
6886 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined
6887 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
6888 // CHECK3-NEXT:  entry:
6889 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
6890 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
6891 // CHECK3-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 4
6892 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
6893 // CHECK3-NEXT:    [[AND_VAR1:%.*]] = alloca i8, align 1
6894 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6895 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6896 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6897 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6898 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6899 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6900 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
6901 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
6902 // CHECK3-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
6903 // CHECK3-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
6904 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
6905 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
6906 // CHECK3-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 4
6907 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
6908 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 4
6909 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
6910 // CHECK3-NEXT:    store i8 1, ptr [[AND_VAR1]], align 1
6911 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
6912 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
6913 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
6914 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
6915 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
6916 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
6917 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
6918 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6919 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
6920 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6921 // CHECK3:       cond.true:
6922 // CHECK3-NEXT:    br label [[COND_END:%.*]]
6923 // CHECK3:       cond.false:
6924 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6925 // CHECK3-NEXT:    br label [[COND_END]]
6926 // CHECK3:       cond.end:
6927 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6928 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
6929 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
6930 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
6931 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6932 // CHECK3:       omp.inner.for.cond:
6933 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6934 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
6935 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6936 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6937 // CHECK3:       omp.inner.for.body:
6938 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6939 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6940 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6941 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
6942 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8, ptr [[AND_VAR1]], align 1
6943 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP10]] to i1
6944 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
6945 // CHECK3:       land.rhs:
6946 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
6947 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
6948 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
6949 // CHECK3-NEXT:    [[REM:%.*]] = srem i32 [[TMP12]], 2
6950 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[REM]], 0
6951 // CHECK3-NEXT:    br label [[LAND_END]]
6952 // CHECK3:       land.end:
6953 // CHECK3-NEXT:    [[TMP13:%.*]] = phi i1 [ false, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LAND_RHS]] ]
6954 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP13]] to i8
6955 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[AND_VAR1]], align 1
6956 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6957 // CHECK3:       omp.body.continue:
6958 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6959 // CHECK3:       omp.inner.for.inc:
6960 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
6961 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
6962 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
6963 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
6964 // CHECK3:       omp.inner.for.end:
6965 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6966 // CHECK3:       omp.loop.exit:
6967 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
6968 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
6969 // CHECK3-NEXT:    store ptr [[AND_VAR1]], ptr [[TMP15]], align 4
6970 // CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
6971 // CHECK3-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
6972 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
6973 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
6974 // CHECK3-NEXT:    ]
6975 // CHECK3:       .omp.reduction.case1:
6976 // CHECK3-NEXT:    [[TMP17:%.*]] = load i8, ptr [[TMP0]], align 1
6977 // CHECK3-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP17]] to i1
6978 // CHECK3-NEXT:    br i1 [[TOBOOL5]], label [[LAND_RHS6:%.*]], label [[LAND_END8:%.*]]
6979 // CHECK3:       land.rhs6:
6980 // CHECK3-NEXT:    [[TMP18:%.*]] = load i8, ptr [[AND_VAR1]], align 1
6981 // CHECK3-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP18]] to i1
6982 // CHECK3-NEXT:    br label [[LAND_END8]]
6983 // CHECK3:       land.end8:
6984 // CHECK3-NEXT:    [[TMP19:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LAND_RHS6]] ]
6985 // CHECK3-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP19]] to i8
6986 // CHECK3-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
6987 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
6988 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
6989 // CHECK3:       .omp.reduction.case2:
6990 // CHECK3-NEXT:    [[TMP20:%.*]] = load i8, ptr [[AND_VAR1]], align 1
6991 // CHECK3-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP20]] to i1
6992 // CHECK3-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
6993 // CHECK3-NEXT:    br label [[ATOMIC_CONT:%.*]]
6994 // CHECK3:       atomic_cont:
6995 // CHECK3-NEXT:    [[TMP21:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP27:%.*]], [[LAND_END17:%.*]] ]
6996 // CHECK3-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP21]] to i1
6997 // CHECK3-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
6998 // CHECK3-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
6999 // CHECK3-NEXT:    [[TMP22:%.*]] = load i8, ptr [[_TMP12]], align 1
7000 // CHECK3-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP22]] to i1
7001 // CHECK3-NEXT:    br i1 [[TOBOOL14]], label [[LAND_RHS15:%.*]], label [[LAND_END17]]
7002 // CHECK3:       land.rhs15:
7003 // CHECK3-NEXT:    [[TMP23:%.*]] = load i8, ptr [[AND_VAR1]], align 1
7004 // CHECK3-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP23]] to i1
7005 // CHECK3-NEXT:    br label [[LAND_END17]]
7006 // CHECK3:       land.end17:
7007 // CHECK3-NEXT:    [[TMP24:%.*]] = phi i1 [ false, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LAND_RHS15]] ]
7008 // CHECK3-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP24]] to i8
7009 // CHECK3-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
7010 // CHECK3-NEXT:    [[TMP25:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
7011 // CHECK3-NEXT:    [[TMP26:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP21]], i8 [[TMP25]] monotonic monotonic, align 1
7012 // CHECK3-NEXT:    [[TMP27]] = extractvalue { i8, i1 } [[TMP26]], 0
7013 // CHECK3-NEXT:    [[TMP28:%.*]] = extractvalue { i8, i1 } [[TMP26]], 1
7014 // CHECK3-NEXT:    br i1 [[TMP28]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
7015 // CHECK3:       atomic_exit:
7016 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7017 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7018 // CHECK3:       .omp.reduction.default:
7019 // CHECK3-NEXT:    ret void
7020 //
7021 //
7022 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l49.omp_outlined.omp.reduction.reduction_func
7023 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
7024 // CHECK3-NEXT:  entry:
7025 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
7026 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
7027 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
7028 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
7029 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
7030 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
7031 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
7032 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
7033 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
7034 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
7035 // CHECK3-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
7036 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
7037 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
7038 // CHECK3:       land.rhs:
7039 // CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
7040 // CHECK3-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
7041 // CHECK3-NEXT:    br label [[LAND_END]]
7042 // CHECK3:       land.end:
7043 // CHECK3-NEXT:    [[TMP10:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LAND_RHS]] ]
7044 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
7045 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
7046 // CHECK3-NEXT:    ret void
7047 //
7048 //
7049 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55
7050 // CHECK3-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7051 // CHECK3-NEXT:  entry:
7052 // CHECK3-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 4
7053 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7054 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 4
7055 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7056 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 4
7057 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7058 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
7059 // CHECK3-NEXT:    ret void
7060 //
7061 //
7062 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined
7063 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7064 // CHECK3-NEXT:  entry:
7065 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
7066 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
7067 // CHECK3-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 4
7068 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7069 // CHECK3-NEXT:    [[OR_VAR1:%.*]] = alloca i8, align 1
7070 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7071 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7072 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7073 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7074 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7075 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7076 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
7077 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
7078 // CHECK3-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
7079 // CHECK3-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
7080 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
7081 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
7082 // CHECK3-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 4
7083 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7084 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 4
7085 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7086 // CHECK3-NEXT:    store i8 0, ptr [[OR_VAR1]], align 1
7087 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
7088 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
7089 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
7090 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
7091 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
7092 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
7093 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
7094 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7095 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
7096 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7097 // CHECK3:       cond.true:
7098 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7099 // CHECK3:       cond.false:
7100 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7101 // CHECK3-NEXT:    br label [[COND_END]]
7102 // CHECK3:       cond.end:
7103 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7104 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
7105 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
7106 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
7107 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7108 // CHECK3:       omp.inner.for.cond:
7109 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7110 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7111 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7112 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7113 // CHECK3:       omp.inner.for.body:
7114 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7115 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7116 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7117 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
7118 // CHECK3-NEXT:    [[TMP10:%.*]] = load i8, ptr [[OR_VAR1]], align 1
7119 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP10]] to i1
7120 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
7121 // CHECK3:       lor.rhs:
7122 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
7123 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
7124 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
7125 // CHECK3-NEXT:    [[REM:%.*]] = srem i32 [[TMP12]], 2
7126 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[REM]], 0
7127 // CHECK3-NEXT:    br label [[LOR_END]]
7128 // CHECK3:       lor.end:
7129 // CHECK3-NEXT:    [[TMP13:%.*]] = phi i1 [ true, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LOR_RHS]] ]
7130 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP13]] to i8
7131 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[OR_VAR1]], align 1
7132 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7133 // CHECK3:       omp.body.continue:
7134 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7135 // CHECK3:       omp.inner.for.inc:
7136 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7137 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP14]], 1
7138 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
7139 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
7140 // CHECK3:       omp.inner.for.end:
7141 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7142 // CHECK3:       omp.loop.exit:
7143 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
7144 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
7145 // CHECK3-NEXT:    store ptr [[OR_VAR1]], ptr [[TMP15]], align 4
7146 // CHECK3-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
7147 // CHECK3-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
7148 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
7149 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
7150 // CHECK3-NEXT:    ]
7151 // CHECK3:       .omp.reduction.case1:
7152 // CHECK3-NEXT:    [[TMP17:%.*]] = load i8, ptr [[TMP0]], align 1
7153 // CHECK3-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP17]] to i1
7154 // CHECK3-NEXT:    br i1 [[TOBOOL5]], label [[LOR_END8:%.*]], label [[LOR_RHS6:%.*]]
7155 // CHECK3:       lor.rhs6:
7156 // CHECK3-NEXT:    [[TMP18:%.*]] = load i8, ptr [[OR_VAR1]], align 1
7157 // CHECK3-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP18]] to i1
7158 // CHECK3-NEXT:    br label [[LOR_END8]]
7159 // CHECK3:       lor.end8:
7160 // CHECK3-NEXT:    [[TMP19:%.*]] = phi i1 [ true, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LOR_RHS6]] ]
7161 // CHECK3-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP19]] to i8
7162 // CHECK3-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
7163 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7164 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7165 // CHECK3:       .omp.reduction.case2:
7166 // CHECK3-NEXT:    [[TMP20:%.*]] = load i8, ptr [[OR_VAR1]], align 1
7167 // CHECK3-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP20]] to i1
7168 // CHECK3-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
7169 // CHECK3-NEXT:    br label [[ATOMIC_CONT:%.*]]
7170 // CHECK3:       atomic_cont:
7171 // CHECK3-NEXT:    [[TMP21:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP27:%.*]], [[LOR_END17:%.*]] ]
7172 // CHECK3-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP21]] to i1
7173 // CHECK3-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
7174 // CHECK3-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
7175 // CHECK3-NEXT:    [[TMP22:%.*]] = load i8, ptr [[_TMP12]], align 1
7176 // CHECK3-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP22]] to i1
7177 // CHECK3-NEXT:    br i1 [[TOBOOL14]], label [[LOR_END17]], label [[LOR_RHS15:%.*]]
7178 // CHECK3:       lor.rhs15:
7179 // CHECK3-NEXT:    [[TMP23:%.*]] = load i8, ptr [[OR_VAR1]], align 1
7180 // CHECK3-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP23]] to i1
7181 // CHECK3-NEXT:    br label [[LOR_END17]]
7182 // CHECK3:       lor.end17:
7183 // CHECK3-NEXT:    [[TMP24:%.*]] = phi i1 [ true, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LOR_RHS15]] ]
7184 // CHECK3-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP24]] to i8
7185 // CHECK3-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
7186 // CHECK3-NEXT:    [[TMP25:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
7187 // CHECK3-NEXT:    [[TMP26:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP21]], i8 [[TMP25]] monotonic monotonic, align 1
7188 // CHECK3-NEXT:    [[TMP27]] = extractvalue { i8, i1 } [[TMP26]], 0
7189 // CHECK3-NEXT:    [[TMP28:%.*]] = extractvalue { i8, i1 } [[TMP26]], 1
7190 // CHECK3-NEXT:    br i1 [[TMP28]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
7191 // CHECK3:       atomic_exit:
7192 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7193 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7194 // CHECK3:       .omp.reduction.default:
7195 // CHECK3-NEXT:    ret void
7196 //
7197 //
7198 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l55.omp_outlined.omp.reduction.reduction_func
7199 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
7200 // CHECK3-NEXT:  entry:
7201 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
7202 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
7203 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
7204 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
7205 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
7206 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
7207 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
7208 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
7209 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
7210 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
7211 // CHECK3-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
7212 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
7213 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
7214 // CHECK3:       lor.rhs:
7215 // CHECK3-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
7216 // CHECK3-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
7217 // CHECK3-NEXT:    br label [[LOR_END]]
7218 // CHECK3:       lor.end:
7219 // CHECK3-NEXT:    [[TMP10:%.*]] = phi i1 [ true, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LOR_RHS]] ]
7220 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
7221 // CHECK3-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
7222 // CHECK3-NEXT:    ret void
7223 //
7224 //
7225 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61
7226 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7227 // CHECK3-NEXT:  entry:
7228 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
7229 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7230 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
7231 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7232 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
7233 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7234 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
7235 // CHECK3-NEXT:    ret void
7236 //
7237 //
7238 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined
7239 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7240 // CHECK3-NEXT:  entry:
7241 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
7242 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
7243 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
7244 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7245 // CHECK3-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
7246 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7247 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7248 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7249 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7250 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7251 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7252 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
7253 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
7254 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
7255 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
7256 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
7257 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7258 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
7259 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7260 // CHECK3-NEXT:    store i32 -1, ptr [[BIT_VAR1]], align 4
7261 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
7262 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
7263 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
7264 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
7265 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
7266 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
7267 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
7268 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7269 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
7270 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7271 // CHECK3:       cond.true:
7272 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7273 // CHECK3:       cond.false:
7274 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7275 // CHECK3-NEXT:    br label [[COND_END]]
7276 // CHECK3:       cond.end:
7277 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7278 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
7279 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
7280 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
7281 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7282 // CHECK3:       omp.inner.for.cond:
7283 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7284 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7285 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7286 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7287 // CHECK3:       omp.inner.for.body:
7288 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7289 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7290 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7291 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
7292 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7293 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
7294 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
7295 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
7296 // CHECK3-NEXT:    [[AND:%.*]] = and i32 [[TMP10]], [[TMP12]]
7297 // CHECK3-NEXT:    store i32 [[AND]], ptr [[BIT_VAR1]], align 4
7298 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7299 // CHECK3:       omp.body.continue:
7300 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7301 // CHECK3:       omp.inner.for.inc:
7302 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7303 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
7304 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
7305 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
7306 // CHECK3:       omp.inner.for.end:
7307 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7308 // CHECK3:       omp.loop.exit:
7309 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
7310 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
7311 // CHECK3-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP14]], align 4
7312 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
7313 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
7314 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
7315 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
7316 // CHECK3-NEXT:    ]
7317 // CHECK3:       .omp.reduction.case1:
7318 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
7319 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7320 // CHECK3-NEXT:    [[AND4:%.*]] = and i32 [[TMP16]], [[TMP17]]
7321 // CHECK3-NEXT:    store i32 [[AND4]], ptr [[TMP0]], align 4
7322 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7323 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7324 // CHECK3:       .omp.reduction.case2:
7325 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7326 // CHECK3-NEXT:    [[TMP19:%.*]] = atomicrmw and ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
7327 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7328 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7329 // CHECK3:       .omp.reduction.default:
7330 // CHECK3-NEXT:    ret void
7331 //
7332 //
7333 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l61.omp_outlined.omp.reduction.reduction_func
7334 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
7335 // CHECK3-NEXT:  entry:
7336 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
7337 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
7338 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
7339 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
7340 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
7341 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
7342 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
7343 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
7344 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
7345 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
7346 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
7347 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
7348 // CHECK3-NEXT:    [[AND:%.*]] = and i32 [[TMP8]], [[TMP9]]
7349 // CHECK3-NEXT:    store i32 [[AND]], ptr [[TMP7]], align 4
7350 // CHECK3-NEXT:    ret void
7351 //
7352 //
7353 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66
7354 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7355 // CHECK3-NEXT:  entry:
7356 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
7357 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7358 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
7359 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7360 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
7361 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7362 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
7363 // CHECK3-NEXT:    ret void
7364 //
7365 //
7366 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined
7367 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7368 // CHECK3-NEXT:  entry:
7369 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
7370 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
7371 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
7372 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7373 // CHECK3-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
7374 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7375 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7376 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7377 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7378 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7379 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7380 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
7381 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
7382 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
7383 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
7384 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
7385 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7386 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
7387 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7388 // CHECK3-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
7389 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
7390 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
7391 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
7392 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
7393 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
7394 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
7395 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
7396 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7397 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
7398 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7399 // CHECK3:       cond.true:
7400 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7401 // CHECK3:       cond.false:
7402 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7403 // CHECK3-NEXT:    br label [[COND_END]]
7404 // CHECK3:       cond.end:
7405 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7406 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
7407 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
7408 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
7409 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7410 // CHECK3:       omp.inner.for.cond:
7411 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7412 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7413 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7414 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7415 // CHECK3:       omp.inner.for.body:
7416 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7417 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7418 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7419 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
7420 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7421 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
7422 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
7423 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
7424 // CHECK3-NEXT:    [[OR:%.*]] = or i32 [[TMP10]], [[TMP12]]
7425 // CHECK3-NEXT:    store i32 [[OR]], ptr [[BIT_VAR1]], align 4
7426 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7427 // CHECK3:       omp.body.continue:
7428 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7429 // CHECK3:       omp.inner.for.inc:
7430 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7431 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
7432 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
7433 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
7434 // CHECK3:       omp.inner.for.end:
7435 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7436 // CHECK3:       omp.loop.exit:
7437 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
7438 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
7439 // CHECK3-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP14]], align 4
7440 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
7441 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
7442 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
7443 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
7444 // CHECK3-NEXT:    ]
7445 // CHECK3:       .omp.reduction.case1:
7446 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
7447 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7448 // CHECK3-NEXT:    [[OR4:%.*]] = or i32 [[TMP16]], [[TMP17]]
7449 // CHECK3-NEXT:    store i32 [[OR4]], ptr [[TMP0]], align 4
7450 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7451 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7452 // CHECK3:       .omp.reduction.case2:
7453 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7454 // CHECK3-NEXT:    [[TMP19:%.*]] = atomicrmw or ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
7455 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7456 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7457 // CHECK3:       .omp.reduction.default:
7458 // CHECK3-NEXT:    ret void
7459 //
7460 //
7461 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l66.omp_outlined.omp.reduction.reduction_func
7462 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
7463 // CHECK3-NEXT:  entry:
7464 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
7465 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
7466 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
7467 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
7468 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
7469 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
7470 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
7471 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
7472 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
7473 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
7474 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
7475 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
7476 // CHECK3-NEXT:    [[OR:%.*]] = or i32 [[TMP8]], [[TMP9]]
7477 // CHECK3-NEXT:    store i32 [[OR]], ptr [[TMP7]], align 4
7478 // CHECK3-NEXT:    ret void
7479 //
7480 //
7481 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71
7482 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7483 // CHECK3-NEXT:  entry:
7484 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
7485 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7486 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
7487 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7488 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
7489 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7490 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
7491 // CHECK3-NEXT:    ret void
7492 //
7493 //
7494 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined
7495 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7496 // CHECK3-NEXT:  entry:
7497 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
7498 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
7499 // CHECK3-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 4
7500 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7501 // CHECK3-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
7502 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7503 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7504 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7505 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7506 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7507 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7508 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
7509 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
7510 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
7511 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
7512 // CHECK3-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 4
7513 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7514 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 4
7515 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7516 // CHECK3-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
7517 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
7518 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
7519 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
7520 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
7521 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
7522 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
7523 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
7524 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7525 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
7526 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7527 // CHECK3:       cond.true:
7528 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7529 // CHECK3:       cond.false:
7530 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7531 // CHECK3-NEXT:    br label [[COND_END]]
7532 // CHECK3:       cond.end:
7533 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7534 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
7535 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
7536 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
7537 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7538 // CHECK3:       omp.inner.for.cond:
7539 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7540 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7541 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7542 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7543 // CHECK3:       omp.inner.for.body:
7544 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7545 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7546 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7547 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
7548 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7549 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
7550 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
7551 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
7552 // CHECK3-NEXT:    [[XOR:%.*]] = xor i32 [[TMP10]], [[TMP12]]
7553 // CHECK3-NEXT:    store i32 [[XOR]], ptr [[BIT_VAR1]], align 4
7554 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7555 // CHECK3:       omp.body.continue:
7556 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7557 // CHECK3:       omp.inner.for.inc:
7558 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7559 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1
7560 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
7561 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
7562 // CHECK3:       omp.inner.for.end:
7563 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7564 // CHECK3:       omp.loop.exit:
7565 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
7566 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
7567 // CHECK3-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP14]], align 4
7568 // CHECK3-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
7569 // CHECK3-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
7570 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
7571 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
7572 // CHECK3-NEXT:    ]
7573 // CHECK3:       .omp.reduction.case1:
7574 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP0]], align 4
7575 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7576 // CHECK3-NEXT:    [[XOR4:%.*]] = xor i32 [[TMP16]], [[TMP17]]
7577 // CHECK3-NEXT:    store i32 [[XOR4]], ptr [[TMP0]], align 4
7578 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7579 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7580 // CHECK3:       .omp.reduction.case2:
7581 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
7582 // CHECK3-NEXT:    [[TMP19:%.*]] = atomicrmw xor ptr [[TMP0]], i32 [[TMP18]] monotonic, align 4
7583 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7584 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7585 // CHECK3:       .omp.reduction.default:
7586 // CHECK3-NEXT:    ret void
7587 //
7588 //
7589 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l71.omp_outlined.omp.reduction.reduction_func
7590 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
7591 // CHECK3-NEXT:  entry:
7592 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
7593 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
7594 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
7595 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
7596 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
7597 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
7598 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
7599 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
7600 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
7601 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
7602 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
7603 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
7604 // CHECK3-NEXT:    [[XOR:%.*]] = xor i32 [[TMP8]], [[TMP9]]
7605 // CHECK3-NEXT:    store i32 [[XOR]], ptr [[TMP7]], align 4
7606 // CHECK3-NEXT:    ret void
7607 //
7608 //
7609 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77
7610 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7611 // CHECK3-NEXT:  entry:
7612 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
7613 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7614 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
7615 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7616 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
7617 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7618 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
7619 // CHECK3-NEXT:    ret void
7620 //
7621 //
7622 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined
7623 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7624 // CHECK3-NEXT:  entry:
7625 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
7626 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
7627 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
7628 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7629 // CHECK3-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
7630 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7631 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7632 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7633 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7634 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7635 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7636 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
7637 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
7638 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
7639 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
7640 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
7641 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7642 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
7643 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7644 // CHECK3-NEXT:    store i32 -2147483648, ptr [[T_VAR1]], align 4
7645 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
7646 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
7647 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
7648 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
7649 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
7650 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
7651 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
7652 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7653 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
7654 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7655 // CHECK3:       cond.true:
7656 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7657 // CHECK3:       cond.false:
7658 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7659 // CHECK3-NEXT:    br label [[COND_END]]
7660 // CHECK3:       cond.end:
7661 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7662 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
7663 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
7664 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
7665 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7666 // CHECK3:       omp.inner.for.cond:
7667 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7668 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7669 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7670 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7671 // CHECK3:       omp.inner.for.body:
7672 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7673 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7674 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7675 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
7676 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
7677 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
7678 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
7679 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
7680 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sge i32 [[TMP10]], [[TMP12]]
7681 // CHECK3-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
7682 // CHECK3:       cond.true4:
7683 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[T_VAR1]], align 4
7684 // CHECK3-NEXT:    br label [[COND_END7:%.*]]
7685 // CHECK3:       cond.false5:
7686 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
7687 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP14]]
7688 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[ARRAYIDX6]], align 4
7689 // CHECK3-NEXT:    br label [[COND_END7]]
7690 // CHECK3:       cond.end7:
7691 // CHECK3-NEXT:    [[COND8:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE4]] ], [ [[TMP15]], [[COND_FALSE5]] ]
7692 // CHECK3-NEXT:    store i32 [[COND8]], ptr [[T_VAR1]], align 4
7693 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7694 // CHECK3:       omp.body.continue:
7695 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7696 // CHECK3:       omp.inner.for.inc:
7697 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7698 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP16]], 1
7699 // CHECK3-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
7700 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
7701 // CHECK3:       omp.inner.for.end:
7702 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7703 // CHECK3:       omp.loop.exit:
7704 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
7705 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
7706 // CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP17]], align 4
7707 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
7708 // CHECK3-NEXT:    switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
7709 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
7710 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
7711 // CHECK3-NEXT:    ]
7712 // CHECK3:       .omp.reduction.case1:
7713 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
7714 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[T_VAR1]], align 4
7715 // CHECK3-NEXT:    [[CMP10:%.*]] = icmp sgt i32 [[TMP19]], [[TMP20]]
7716 // CHECK3-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
7717 // CHECK3:       cond.true11:
7718 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP0]], align 4
7719 // CHECK3-NEXT:    br label [[COND_END13:%.*]]
7720 // CHECK3:       cond.false12:
7721 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[T_VAR1]], align 4
7722 // CHECK3-NEXT:    br label [[COND_END13]]
7723 // CHECK3:       cond.end13:
7724 // CHECK3-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP21]], [[COND_TRUE11]] ], [ [[TMP22]], [[COND_FALSE12]] ]
7725 // CHECK3-NEXT:    store i32 [[COND14]], ptr [[TMP0]], align 4
7726 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7727 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7728 // CHECK3:       .omp.reduction.case2:
7729 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[T_VAR1]], align 4
7730 // CHECK3-NEXT:    [[TMP24:%.*]] = atomicrmw max ptr [[TMP0]], i32 [[TMP23]] monotonic, align 4
7731 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7732 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7733 // CHECK3:       .omp.reduction.default:
7734 // CHECK3-NEXT:    ret void
7735 //
7736 //
7737 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l77.omp_outlined.omp.reduction.reduction_func
7738 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
7739 // CHECK3-NEXT:  entry:
7740 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
7741 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
7742 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
7743 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
7744 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
7745 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
7746 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
7747 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
7748 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
7749 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
7750 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
7751 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
7752 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
7753 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7754 // CHECK3:       cond.true:
7755 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
7756 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7757 // CHECK3:       cond.false:
7758 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
7759 // CHECK3-NEXT:    br label [[COND_END]]
7760 // CHECK3:       cond.end:
7761 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
7762 // CHECK3-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
7763 // CHECK3-NEXT:    ret void
7764 //
7765 //
7766 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83
7767 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7768 // CHECK3-NEXT:  entry:
7769 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
7770 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7771 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
7772 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7773 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
7774 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7775 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined, ptr [[TMP0]], ptr [[TMP1]])
7776 // CHECK3-NEXT:    ret void
7777 //
7778 //
7779 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined
7780 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]]) #[[ATTR1]] {
7781 // CHECK3-NEXT:  entry:
7782 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
7783 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
7784 // CHECK3-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
7785 // CHECK3-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
7786 // CHECK3-NEXT:    [[T_VAR1:%.*]] = alloca i32, align 4
7787 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7788 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7789 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7790 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7791 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7792 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7793 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
7794 // CHECK3-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 4
7795 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
7796 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
7797 // CHECK3-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
7798 // CHECK3-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
7799 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
7800 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
7801 // CHECK3-NEXT:    store i32 2147483647, ptr [[T_VAR1]], align 4
7802 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
7803 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
7804 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
7805 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
7806 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
7807 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
7808 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP3]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
7809 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7810 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
7811 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7812 // CHECK3:       cond.true:
7813 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7814 // CHECK3:       cond.false:
7815 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7816 // CHECK3-NEXT:    br label [[COND_END]]
7817 // CHECK3:       cond.end:
7818 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7819 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
7820 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
7821 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
7822 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7823 // CHECK3:       omp.inner.for.cond:
7824 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7825 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7826 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7827 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7828 // CHECK3:       omp.inner.for.body:
7829 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7830 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7831 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7832 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
7833 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[T_VAR1]], align 4
7834 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
7835 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP11]]
7836 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
7837 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP10]], [[TMP12]]
7838 // CHECK3-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
7839 // CHECK3:       cond.true4:
7840 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[T_VAR1]], align 4
7841 // CHECK3-NEXT:    br label [[COND_END7:%.*]]
7842 // CHECK3:       cond.false5:
7843 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
7844 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x i32], ptr [[TMP1]], i32 0, i32 [[TMP14]]
7845 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[ARRAYIDX6]], align 4
7846 // CHECK3-NEXT:    br label [[COND_END7]]
7847 // CHECK3:       cond.end7:
7848 // CHECK3-NEXT:    [[COND8:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE4]] ], [ [[TMP15]], [[COND_FALSE5]] ]
7849 // CHECK3-NEXT:    store i32 [[COND8]], ptr [[T_VAR1]], align 4
7850 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7851 // CHECK3:       omp.body.continue:
7852 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7853 // CHECK3:       omp.inner.for.inc:
7854 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
7855 // CHECK3-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP16]], 1
7856 // CHECK3-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
7857 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
7858 // CHECK3:       omp.inner.for.end:
7859 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7860 // CHECK3:       omp.loop.exit:
7861 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
7862 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0
7863 // CHECK3-NEXT:    store ptr [[T_VAR1]], ptr [[TMP17]], align 4
7864 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP3]], i32 1, i32 4, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
7865 // CHECK3-NEXT:    switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
7866 // CHECK3-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
7867 // CHECK3-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
7868 // CHECK3-NEXT:    ]
7869 // CHECK3:       .omp.reduction.case1:
7870 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
7871 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[T_VAR1]], align 4
7872 // CHECK3-NEXT:    [[CMP10:%.*]] = icmp slt i32 [[TMP19]], [[TMP20]]
7873 // CHECK3-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
7874 // CHECK3:       cond.true11:
7875 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP0]], align 4
7876 // CHECK3-NEXT:    br label [[COND_END13:%.*]]
7877 // CHECK3:       cond.false12:
7878 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[T_VAR1]], align 4
7879 // CHECK3-NEXT:    br label [[COND_END13]]
7880 // CHECK3:       cond.end13:
7881 // CHECK3-NEXT:    [[COND14:%.*]] = phi i32 [ [[TMP21]], [[COND_TRUE11]] ], [ [[TMP22]], [[COND_FALSE12]] ]
7882 // CHECK3-NEXT:    store i32 [[COND14]], ptr [[TMP0]], align 4
7883 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7884 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7885 // CHECK3:       .omp.reduction.case2:
7886 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[T_VAR1]], align 4
7887 // CHECK3-NEXT:    [[TMP24:%.*]] = atomicrmw min ptr [[TMP0]], i32 [[TMP23]] monotonic, align 4
7888 // CHECK3-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP3]], ptr @.gomp_critical_user_.reduction.var)
7889 // CHECK3-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
7890 // CHECK3:       .omp.reduction.default:
7891 // CHECK3-NEXT:    ret void
7892 //
7893 //
7894 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l83.omp_outlined.omp.reduction.reduction_func
7895 // CHECK3-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
7896 // CHECK3-NEXT:  entry:
7897 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 4
7898 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 4
7899 // CHECK3-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 4
7900 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 4
7901 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 4
7902 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 4
7903 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i32 0, i32 0
7904 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 4
7905 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i32 0, i32 0
7906 // CHECK3-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 4
7907 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
7908 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
7909 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], [[TMP9]]
7910 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7911 // CHECK3:       cond.true:
7912 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
7913 // CHECK3-NEXT:    br label [[COND_END:%.*]]
7914 // CHECK3:       cond.false:
7915 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
7916 // CHECK3-NEXT:    br label [[COND_END]]
7917 // CHECK3:       cond.end:
7918 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
7919 // CHECK3-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
7920 // CHECK3-NEXT:    ret void
7921 //
7922 //
7923 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
7924 // CHECK3-SAME: () #[[ATTR7:[0-9]+]] {
7925 // CHECK3-NEXT:  entry:
7926 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
7927 // CHECK3-NEXT:    ret void
7928 //
7929 //
7930 // CHECK9-LABEL: define {{[^@]+}}@main
7931 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
7932 // CHECK9-NEXT:  entry:
7933 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
7934 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1
7935 // CHECK9-NEXT:    [[REF_TMP1:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 1
7936 // CHECK9-NEXT:    [[REF_TMP2:%.*]] = alloca [[CLASS_ANON_2:%.*]], align 1
7937 // CHECK9-NEXT:    [[REF_TMP3:%.*]] = alloca [[CLASS_ANON_4:%.*]], align 1
7938 // CHECK9-NEXT:    [[REF_TMP4:%.*]] = alloca [[CLASS_ANON_6:%.*]], align 1
7939 // CHECK9-NEXT:    [[REF_TMP5:%.*]] = alloca [[CLASS_ANON_8:%.*]], align 1
7940 // CHECK9-NEXT:    [[REF_TMP6:%.*]] = alloca [[CLASS_ANON_10:%.*]], align 1
7941 // CHECK9-NEXT:    [[REF_TMP7:%.*]] = alloca [[CLASS_ANON_12:%.*]], align 1
7942 // CHECK9-NEXT:    [[REF_TMP8:%.*]] = alloca [[CLASS_ANON_14:%.*]], align 1
7943 // CHECK9-NEXT:    [[REF_TMP9:%.*]] = alloca [[CLASS_ANON_16:%.*]], align 1
7944 // CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
7945 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP]])
7946 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_1clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP1]])
7947 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_2clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP2]])
7948 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_3clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP3]])
7949 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_4clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP4]])
7950 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_5clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP5]])
7951 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_6clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP6]])
7952 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_7clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP7]])
7953 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_8clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP8]])
7954 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_9clEv"(ptr noundef nonnull align 1 dereferenceable(1) [[REF_TMP9]])
7955 // CHECK9-NEXT:    ret i32 0
7956 //
7957 //
7958 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96
7959 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2:[0-9]+]] {
7960 // CHECK9-NEXT:  entry:
7961 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
7962 // CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
7963 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
7964 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined, ptr [[TMP0]])
7965 // CHECK9-NEXT:    ret void
7966 //
7967 //
7968 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined
7969 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] {
7970 // CHECK9-NEXT:  entry:
7971 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
7972 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
7973 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
7974 // CHECK9-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
7975 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7976 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7977 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7978 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7979 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7980 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7981 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7982 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_18:%.*]], align 8
7983 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
7984 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
7985 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
7986 // CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
7987 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
7988 // CHECK9-NEXT:    store i32 0, ptr [[SIVAR1]], align 4
7989 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
7990 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
7991 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
7992 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
7993 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
7994 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
7995 // CHECK9-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)
7996 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
7997 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
7998 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7999 // CHECK9:       cond.true:
8000 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8001 // CHECK9:       cond.false:
8002 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8003 // CHECK9-NEXT:    br label [[COND_END]]
8004 // CHECK9:       cond.end:
8005 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8006 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
8007 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
8008 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
8009 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8010 // CHECK9:       omp.inner.for.cond:
8011 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8012 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8013 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8014 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8015 // CHECK9:       omp.inner.for.body:
8016 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8017 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8018 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8019 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
8020 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
8021 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
8022 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], [[TMP9]]
8023 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[SIVAR1]], align 4
8024 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_18]], ptr [[REF_TMP]], i32 0, i32 0
8025 // CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP11]], align 8
8026 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
8027 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8028 // CHECK9:       omp.body.continue:
8029 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8030 // CHECK9:       omp.inner.for.inc:
8031 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8032 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
8033 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
8034 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
8035 // CHECK9:       omp.inner.for.end:
8036 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8037 // CHECK9:       omp.loop.exit:
8038 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
8039 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
8040 // CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP13]], align 8
8041 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
8042 // CHECK9-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
8043 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
8044 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
8045 // CHECK9-NEXT:    ]
8046 // CHECK9:       .omp.reduction.case1:
8047 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP0]], align 4
8048 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
8049 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
8050 // CHECK9-NEXT:    store i32 [[ADD5]], ptr [[TMP0]], align 4
8051 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8052 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8053 // CHECK9:       .omp.reduction.case2:
8054 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
8055 // CHECK9-NEXT:    [[TMP18:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP17]] monotonic, align 4
8056 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8057 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8058 // CHECK9:       .omp.reduction.default:
8059 // CHECK9-NEXT:    ret void
8060 //
8061 //
8062 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l96.omp_outlined.omp.reduction.reduction_func
8063 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
8064 // CHECK9-NEXT:  entry:
8065 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
8066 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
8067 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
8068 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
8069 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
8070 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
8071 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
8072 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
8073 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
8074 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
8075 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
8076 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
8077 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
8078 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
8079 // CHECK9-NEXT:    ret void
8080 //
8081 //
8082 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l112
8083 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] {
8084 // CHECK9-NEXT:  entry:
8085 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
8086 // CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8087 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
8088 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l112.omp_outlined, ptr [[TMP0]])
8089 // CHECK9-NEXT:    ret void
8090 //
8091 //
8092 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l112.omp_outlined
8093 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] {
8094 // CHECK9-NEXT:  entry:
8095 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
8096 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
8097 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
8098 // CHECK9-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
8099 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8100 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8101 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8102 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8103 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8104 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8105 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8106 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_19:%.*]], align 8
8107 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
8108 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
8109 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
8110 // CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8111 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
8112 // CHECK9-NEXT:    store i32 0, ptr [[SIVAR1]], align 4
8113 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
8114 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
8115 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
8116 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
8117 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
8118 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
8119 // 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)
8120 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8121 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
8122 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8123 // CHECK9:       cond.true:
8124 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8125 // CHECK9:       cond.false:
8126 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8127 // CHECK9-NEXT:    br label [[COND_END]]
8128 // CHECK9:       cond.end:
8129 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8130 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
8131 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
8132 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
8133 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8134 // CHECK9:       omp.inner.for.cond:
8135 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8136 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8137 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8138 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8139 // CHECK9:       omp.inner.for.body:
8140 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8141 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8142 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8143 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
8144 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
8145 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
8146 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP10]], [[TMP9]]
8147 // CHECK9-NEXT:    store i32 [[SUB]], ptr [[SIVAR1]], align 4
8148 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_19]], ptr [[REF_TMP]], i32 0, i32 0
8149 // CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP11]], align 8
8150 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_1clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
8151 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8152 // CHECK9:       omp.body.continue:
8153 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8154 // CHECK9:       omp.inner.for.inc:
8155 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8156 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
8157 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
8158 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
8159 // CHECK9:       omp.inner.for.end:
8160 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8161 // CHECK9:       omp.loop.exit:
8162 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
8163 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
8164 // CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP13]], align 8
8165 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l112.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
8166 // CHECK9-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
8167 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
8168 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
8169 // CHECK9-NEXT:    ]
8170 // CHECK9:       .omp.reduction.case1:
8171 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP0]], align 4
8172 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
8173 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
8174 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[TMP0]], align 4
8175 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8176 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8177 // CHECK9:       .omp.reduction.case2:
8178 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
8179 // CHECK9-NEXT:    [[TMP18:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP17]] monotonic, align 4
8180 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8181 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8182 // CHECK9:       .omp.reduction.default:
8183 // CHECK9-NEXT:    ret void
8184 //
8185 //
8186 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l112.omp_outlined.omp.reduction.reduction_func
8187 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
8188 // CHECK9-NEXT:  entry:
8189 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
8190 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
8191 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
8192 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
8193 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
8194 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
8195 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
8196 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
8197 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
8198 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
8199 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
8200 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
8201 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
8202 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
8203 // CHECK9-NEXT:    ret void
8204 //
8205 //
8206 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l122
8207 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] {
8208 // CHECK9-NEXT:  entry:
8209 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
8210 // CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8211 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
8212 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l122.omp_outlined, ptr [[TMP0]])
8213 // CHECK9-NEXT:    ret void
8214 //
8215 //
8216 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l122.omp_outlined
8217 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] {
8218 // CHECK9-NEXT:  entry:
8219 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
8220 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
8221 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca ptr, align 8
8222 // CHECK9-NEXT:    [[SIVAR1:%.*]] = alloca i32, align 4
8223 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8224 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8225 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8226 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8227 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8228 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8229 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8230 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_20:%.*]], align 8
8231 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
8232 // CHECK9-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i32, align 4
8233 // CHECK9-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
8234 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
8235 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
8236 // CHECK9-NEXT:    store ptr [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8237 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[SIVAR_ADDR]], align 8
8238 // CHECK9-NEXT:    store i32 1, ptr [[SIVAR1]], align 4
8239 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
8240 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
8241 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
8242 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
8243 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
8244 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
8245 // 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)
8246 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8247 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
8248 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8249 // CHECK9:       cond.true:
8250 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8251 // CHECK9:       cond.false:
8252 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8253 // CHECK9-NEXT:    br label [[COND_END]]
8254 // CHECK9:       cond.end:
8255 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8256 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
8257 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
8258 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
8259 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8260 // CHECK9:       omp.inner.for.cond:
8261 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8262 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8263 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8264 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8265 // CHECK9:       omp.inner.for.body:
8266 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8267 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8268 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8269 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
8270 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
8271 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[SIVAR1]], align 4
8272 // CHECK9-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[TMP10]], [[TMP9]]
8273 // CHECK9-NEXT:    store i32 [[MUL3]], ptr [[SIVAR1]], align 4
8274 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_20]], ptr [[REF_TMP]], i32 0, i32 0
8275 // CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP11]], align 8
8276 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_2clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
8277 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8278 // CHECK9:       omp.body.continue:
8279 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8280 // CHECK9:       omp.inner.for.inc:
8281 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8282 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
8283 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
8284 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
8285 // CHECK9:       omp.inner.for.end:
8286 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8287 // CHECK9:       omp.loop.exit:
8288 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
8289 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
8290 // CHECK9-NEXT:    store ptr [[SIVAR1]], ptr [[TMP13]], align 8
8291 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l122.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
8292 // CHECK9-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
8293 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
8294 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
8295 // CHECK9-NEXT:    ]
8296 // CHECK9:       .omp.reduction.case1:
8297 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP0]], align 4
8298 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[SIVAR1]], align 4
8299 // CHECK9-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[TMP15]], [[TMP16]]
8300 // CHECK9-NEXT:    store i32 [[MUL5]], ptr [[TMP0]], align 4
8301 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8302 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8303 // CHECK9:       .omp.reduction.case2:
8304 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[SIVAR1]], align 4
8305 // CHECK9-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i32, ptr [[TMP0]] monotonic, align 4
8306 // CHECK9-NEXT:    br label [[ATOMIC_CONT:%.*]]
8307 // CHECK9:       atomic_cont:
8308 // CHECK9-NEXT:    [[TMP18:%.*]] = phi i32 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP23:%.*]], [[ATOMIC_CONT]] ]
8309 // CHECK9-NEXT:    store i32 [[TMP18]], ptr [[_TMP6]], align 4
8310 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[_TMP6]], align 4
8311 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[SIVAR1]], align 4
8312 // CHECK9-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[TMP19]], [[TMP20]]
8313 // CHECK9-NEXT:    store i32 [[MUL7]], ptr [[ATOMIC_TEMP]], align 4
8314 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ATOMIC_TEMP]], align 4
8315 // CHECK9-NEXT:    [[TMP22:%.*]] = cmpxchg ptr [[TMP0]], i32 [[TMP18]], i32 [[TMP21]] monotonic monotonic, align 4
8316 // CHECK9-NEXT:    [[TMP23]] = extractvalue { i32, i1 } [[TMP22]], 0
8317 // CHECK9-NEXT:    [[TMP24:%.*]] = extractvalue { i32, i1 } [[TMP22]], 1
8318 // CHECK9-NEXT:    br i1 [[TMP24]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
8319 // CHECK9:       atomic_exit:
8320 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8321 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8322 // CHECK9:       .omp.reduction.default:
8323 // CHECK9-NEXT:    ret void
8324 //
8325 //
8326 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l122.omp_outlined.omp.reduction.reduction_func
8327 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
8328 // CHECK9-NEXT:  entry:
8329 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
8330 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
8331 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
8332 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
8333 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
8334 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
8335 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
8336 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
8337 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
8338 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
8339 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
8340 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
8341 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], [[TMP9]]
8342 // CHECK9-NEXT:    store i32 [[MUL]], ptr [[TMP7]], align 4
8343 // CHECK9-NEXT:    ret void
8344 //
8345 //
8346 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l133
8347 // CHECK9-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8348 // CHECK9-NEXT:  entry:
8349 // CHECK9-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 8
8350 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8351 // CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
8352 // CHECK9-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 8
8353 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8354 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 8
8355 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
8356 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[SIVAR_CASTED]], align 4
8357 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
8358 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l133.omp_outlined, ptr [[TMP0]], i64 [[TMP2]])
8359 // CHECK9-NEXT:    ret void
8360 //
8361 //
8362 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l133.omp_outlined
8363 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[AND_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8364 // CHECK9-NEXT:  entry:
8365 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
8366 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
8367 // CHECK9-NEXT:    [[AND_VAR_ADDR:%.*]] = alloca ptr, align 8
8368 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8369 // CHECK9-NEXT:    [[AND_VAR1:%.*]] = alloca i8, align 1
8370 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8371 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8372 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8373 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8374 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8375 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8376 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8377 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_21:%.*]], align 8
8378 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
8379 // CHECK9-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
8380 // CHECK9-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
8381 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
8382 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
8383 // CHECK9-NEXT:    store ptr [[AND_VAR]], ptr [[AND_VAR_ADDR]], align 8
8384 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8385 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[AND_VAR_ADDR]], align 8
8386 // CHECK9-NEXT:    store i8 1, ptr [[AND_VAR1]], align 1
8387 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
8388 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
8389 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
8390 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
8391 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
8392 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
8393 // 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)
8394 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8395 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
8396 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8397 // CHECK9:       cond.true:
8398 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8399 // CHECK9:       cond.false:
8400 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8401 // CHECK9-NEXT:    br label [[COND_END]]
8402 // CHECK9:       cond.end:
8403 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8404 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
8405 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
8406 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
8407 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8408 // CHECK9:       omp.inner.for.cond:
8409 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8410 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8411 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8412 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8413 // CHECK9:       omp.inner.for.body:
8414 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8415 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8416 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8417 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
8418 // CHECK9-NEXT:    [[TMP9:%.*]] = load i8, ptr [[AND_VAR1]], align 1
8419 // CHECK9-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP9]] to i1
8420 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
8421 // CHECK9:       land.rhs:
8422 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
8423 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP10]], 0
8424 // CHECK9-NEXT:    br label [[LAND_END]]
8425 // CHECK9:       land.end:
8426 // CHECK9-NEXT:    [[TMP11:%.*]] = phi i1 [ false, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LAND_RHS]] ]
8427 // CHECK9-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP11]] to i8
8428 // CHECK9-NEXT:    store i8 [[FROMBOOL]], ptr [[AND_VAR1]], align 1
8429 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_21]], ptr [[REF_TMP]], i32 0, i32 0
8430 // CHECK9-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP12]], align 8
8431 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_3clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
8432 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8433 // CHECK9:       omp.body.continue:
8434 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8435 // CHECK9:       omp.inner.for.inc:
8436 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8437 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
8438 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
8439 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
8440 // CHECK9:       omp.inner.for.end:
8441 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8442 // CHECK9:       omp.loop.exit:
8443 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
8444 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
8445 // CHECK9-NEXT:    store ptr [[AND_VAR1]], ptr [[TMP14]], align 8
8446 // CHECK9-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l133.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
8447 // CHECK9-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
8448 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
8449 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
8450 // CHECK9-NEXT:    ]
8451 // CHECK9:       .omp.reduction.case1:
8452 // CHECK9-NEXT:    [[TMP16:%.*]] = load i8, ptr [[TMP0]], align 1
8453 // CHECK9-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
8454 // CHECK9-NEXT:    br i1 [[TOBOOL5]], label [[LAND_RHS6:%.*]], label [[LAND_END8:%.*]]
8455 // CHECK9:       land.rhs6:
8456 // CHECK9-NEXT:    [[TMP17:%.*]] = load i8, ptr [[AND_VAR1]], align 1
8457 // CHECK9-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP17]] to i1
8458 // CHECK9-NEXT:    br label [[LAND_END8]]
8459 // CHECK9:       land.end8:
8460 // CHECK9-NEXT:    [[TMP18:%.*]] = phi i1 [ false, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LAND_RHS6]] ]
8461 // CHECK9-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP18]] to i8
8462 // CHECK9-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
8463 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8464 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8465 // CHECK9:       .omp.reduction.case2:
8466 // CHECK9-NEXT:    [[TMP19:%.*]] = load i8, ptr [[AND_VAR1]], align 1
8467 // CHECK9-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP19]] to i1
8468 // CHECK9-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
8469 // CHECK9-NEXT:    br label [[ATOMIC_CONT:%.*]]
8470 // CHECK9:       atomic_cont:
8471 // CHECK9-NEXT:    [[TMP20:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP26:%.*]], [[LAND_END17:%.*]] ]
8472 // CHECK9-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP20]] to i1
8473 // CHECK9-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
8474 // CHECK9-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
8475 // CHECK9-NEXT:    [[TMP21:%.*]] = load i8, ptr [[_TMP12]], align 1
8476 // CHECK9-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP21]] to i1
8477 // CHECK9-NEXT:    br i1 [[TOBOOL14]], label [[LAND_RHS15:%.*]], label [[LAND_END17]]
8478 // CHECK9:       land.rhs15:
8479 // CHECK9-NEXT:    [[TMP22:%.*]] = load i8, ptr [[AND_VAR1]], align 1
8480 // CHECK9-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP22]] to i1
8481 // CHECK9-NEXT:    br label [[LAND_END17]]
8482 // CHECK9:       land.end17:
8483 // CHECK9-NEXT:    [[TMP23:%.*]] = phi i1 [ false, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LAND_RHS15]] ]
8484 // CHECK9-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP23]] to i8
8485 // CHECK9-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
8486 // CHECK9-NEXT:    [[TMP24:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
8487 // CHECK9-NEXT:    [[TMP25:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP20]], i8 [[TMP24]] monotonic monotonic, align 1
8488 // CHECK9-NEXT:    [[TMP26]] = extractvalue { i8, i1 } [[TMP25]], 0
8489 // CHECK9-NEXT:    [[TMP27:%.*]] = extractvalue { i8, i1 } [[TMP25]], 1
8490 // CHECK9-NEXT:    br i1 [[TMP27]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
8491 // CHECK9:       atomic_exit:
8492 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8493 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8494 // CHECK9:       .omp.reduction.default:
8495 // CHECK9-NEXT:    ret void
8496 //
8497 //
8498 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l133.omp_outlined.omp.reduction.reduction_func
8499 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
8500 // CHECK9-NEXT:  entry:
8501 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
8502 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
8503 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
8504 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
8505 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
8506 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
8507 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
8508 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
8509 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
8510 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
8511 // CHECK9-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
8512 // CHECK9-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
8513 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
8514 // CHECK9:       land.rhs:
8515 // CHECK9-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
8516 // CHECK9-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
8517 // CHECK9-NEXT:    br label [[LAND_END]]
8518 // CHECK9:       land.end:
8519 // CHECK9-NEXT:    [[TMP10:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LAND_RHS]] ]
8520 // CHECK9-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
8521 // CHECK9-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
8522 // CHECK9-NEXT:    ret void
8523 //
8524 //
8525 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l144
8526 // CHECK9-SAME: (ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8527 // CHECK9-NEXT:  entry:
8528 // CHECK9-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 8
8529 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8530 // CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
8531 // CHECK9-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 8
8532 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8533 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 8
8534 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
8535 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[SIVAR_CASTED]], align 4
8536 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
8537 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l144.omp_outlined, ptr [[TMP0]], i64 [[TMP2]])
8538 // CHECK9-NEXT:    ret void
8539 //
8540 //
8541 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l144.omp_outlined
8542 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 1 dereferenceable(1) [[OR_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8543 // CHECK9-NEXT:  entry:
8544 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
8545 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
8546 // CHECK9-NEXT:    [[OR_VAR_ADDR:%.*]] = alloca ptr, align 8
8547 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8548 // CHECK9-NEXT:    [[OR_VAR1:%.*]] = alloca i8, align 1
8549 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8550 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8551 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8552 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8553 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8554 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8555 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8556 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_22:%.*]], align 8
8557 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
8558 // CHECK9-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
8559 // CHECK9-NEXT:    [[_TMP12:%.*]] = alloca i8, align 1
8560 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
8561 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
8562 // CHECK9-NEXT:    store ptr [[OR_VAR]], ptr [[OR_VAR_ADDR]], align 8
8563 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8564 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[OR_VAR_ADDR]], align 8
8565 // CHECK9-NEXT:    store i8 0, ptr [[OR_VAR1]], align 1
8566 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
8567 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
8568 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
8569 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
8570 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
8571 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
8572 // 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)
8573 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8574 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
8575 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8576 // CHECK9:       cond.true:
8577 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8578 // CHECK9:       cond.false:
8579 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8580 // CHECK9-NEXT:    br label [[COND_END]]
8581 // CHECK9:       cond.end:
8582 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8583 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
8584 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
8585 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
8586 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8587 // CHECK9:       omp.inner.for.cond:
8588 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8589 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8590 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8591 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8592 // CHECK9:       omp.inner.for.body:
8593 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8594 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8595 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8596 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
8597 // CHECK9-NEXT:    [[TMP9:%.*]] = load i8, ptr [[OR_VAR1]], align 1
8598 // CHECK9-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP9]] to i1
8599 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
8600 // CHECK9:       lor.rhs:
8601 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
8602 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp eq i32 [[TMP10]], 0
8603 // CHECK9-NEXT:    br label [[LOR_END]]
8604 // CHECK9:       lor.end:
8605 // CHECK9-NEXT:    [[TMP11:%.*]] = phi i1 [ true, [[OMP_INNER_FOR_BODY]] ], [ [[CMP3]], [[LOR_RHS]] ]
8606 // CHECK9-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP11]] to i8
8607 // CHECK9-NEXT:    store i8 [[FROMBOOL]], ptr [[OR_VAR1]], align 1
8608 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[CLASS_ANON_22]], ptr [[REF_TMP]], i32 0, i32 0
8609 // CHECK9-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP12]], align 8
8610 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_4clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
8611 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8612 // CHECK9:       omp.body.continue:
8613 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8614 // CHECK9:       omp.inner.for.inc:
8615 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8616 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1
8617 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
8618 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
8619 // CHECK9:       omp.inner.for.end:
8620 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8621 // CHECK9:       omp.loop.exit:
8622 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
8623 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
8624 // CHECK9-NEXT:    store ptr [[OR_VAR1]], ptr [[TMP14]], align 8
8625 // CHECK9-NEXT:    [[TMP15:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l144.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
8626 // CHECK9-NEXT:    switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
8627 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
8628 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
8629 // CHECK9-NEXT:    ]
8630 // CHECK9:       .omp.reduction.case1:
8631 // CHECK9-NEXT:    [[TMP16:%.*]] = load i8, ptr [[TMP0]], align 1
8632 // CHECK9-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
8633 // CHECK9-NEXT:    br i1 [[TOBOOL5]], label [[LOR_END8:%.*]], label [[LOR_RHS6:%.*]]
8634 // CHECK9:       lor.rhs6:
8635 // CHECK9-NEXT:    [[TMP17:%.*]] = load i8, ptr [[OR_VAR1]], align 1
8636 // CHECK9-NEXT:    [[TOBOOL7:%.*]] = trunc i8 [[TMP17]] to i1
8637 // CHECK9-NEXT:    br label [[LOR_END8]]
8638 // CHECK9:       lor.end8:
8639 // CHECK9-NEXT:    [[TMP18:%.*]] = phi i1 [ true, [[DOTOMP_REDUCTION_CASE1]] ], [ [[TOBOOL7]], [[LOR_RHS6]] ]
8640 // CHECK9-NEXT:    [[FROMBOOL9:%.*]] = zext i1 [[TMP18]] to i8
8641 // CHECK9-NEXT:    store i8 [[FROMBOOL9]], ptr [[TMP0]], align 1
8642 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8643 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8644 // CHECK9:       .omp.reduction.case2:
8645 // CHECK9-NEXT:    [[TMP19:%.*]] = load i8, ptr [[OR_VAR1]], align 1
8646 // CHECK9-NEXT:    [[TOBOOL10:%.*]] = trunc i8 [[TMP19]] to i1
8647 // CHECK9-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[TMP0]] monotonic, align 1
8648 // CHECK9-NEXT:    br label [[ATOMIC_CONT:%.*]]
8649 // CHECK9:       atomic_cont:
8650 // CHECK9-NEXT:    [[TMP20:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[TMP26:%.*]], [[LOR_END17:%.*]] ]
8651 // CHECK9-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP20]] to i1
8652 // CHECK9-NEXT:    [[FROMBOOL13:%.*]] = zext i1 [[TOBOOL11]] to i8
8653 // CHECK9-NEXT:    store i8 [[FROMBOOL13]], ptr [[_TMP12]], align 1
8654 // CHECK9-NEXT:    [[TMP21:%.*]] = load i8, ptr [[_TMP12]], align 1
8655 // CHECK9-NEXT:    [[TOBOOL14:%.*]] = trunc i8 [[TMP21]] to i1
8656 // CHECK9-NEXT:    br i1 [[TOBOOL14]], label [[LOR_END17]], label [[LOR_RHS15:%.*]]
8657 // CHECK9:       lor.rhs15:
8658 // CHECK9-NEXT:    [[TMP22:%.*]] = load i8, ptr [[OR_VAR1]], align 1
8659 // CHECK9-NEXT:    [[TOBOOL16:%.*]] = trunc i8 [[TMP22]] to i1
8660 // CHECK9-NEXT:    br label [[LOR_END17]]
8661 // CHECK9:       lor.end17:
8662 // CHECK9-NEXT:    [[TMP23:%.*]] = phi i1 [ true, [[ATOMIC_CONT]] ], [ [[TOBOOL16]], [[LOR_RHS15]] ]
8663 // CHECK9-NEXT:    [[FROMBOOL18:%.*]] = zext i1 [[TMP23]] to i8
8664 // CHECK9-NEXT:    store i8 [[FROMBOOL18]], ptr [[ATOMIC_TEMP]], align 1
8665 // CHECK9-NEXT:    [[TMP24:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
8666 // CHECK9-NEXT:    [[TMP25:%.*]] = cmpxchg ptr [[TMP0]], i8 [[TMP20]], i8 [[TMP24]] monotonic monotonic, align 1
8667 // CHECK9-NEXT:    [[TMP26]] = extractvalue { i8, i1 } [[TMP25]], 0
8668 // CHECK9-NEXT:    [[TMP27:%.*]] = extractvalue { i8, i1 } [[TMP25]], 1
8669 // CHECK9-NEXT:    br i1 [[TMP27]], label [[ATOMIC_EXIT:%.*]], label [[ATOMIC_CONT]]
8670 // CHECK9:       atomic_exit:
8671 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8672 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8673 // CHECK9:       .omp.reduction.default:
8674 // CHECK9-NEXT:    ret void
8675 //
8676 //
8677 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l144.omp_outlined.omp.reduction.reduction_func
8678 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
8679 // CHECK9-NEXT:  entry:
8680 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
8681 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
8682 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
8683 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
8684 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
8685 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
8686 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
8687 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
8688 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
8689 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
8690 // CHECK9-NEXT:    [[TMP8:%.*]] = load i8, ptr [[TMP7]], align 1
8691 // CHECK9-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
8692 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[LOR_END:%.*]], label [[LOR_RHS:%.*]]
8693 // CHECK9:       lor.rhs:
8694 // CHECK9-NEXT:    [[TMP9:%.*]] = load i8, ptr [[TMP5]], align 1
8695 // CHECK9-NEXT:    [[TOBOOL2:%.*]] = trunc i8 [[TMP9]] to i1
8696 // CHECK9-NEXT:    br label [[LOR_END]]
8697 // CHECK9:       lor.end:
8698 // CHECK9-NEXT:    [[TMP10:%.*]] = phi i1 [ true, [[ENTRY:%.*]] ], [ [[TOBOOL2]], [[LOR_RHS]] ]
8699 // CHECK9-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TMP10]] to i8
8700 // CHECK9-NEXT:    store i8 [[FROMBOOL]], ptr [[TMP7]], align 1
8701 // CHECK9-NEXT:    ret void
8702 //
8703 //
8704 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l155
8705 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8706 // CHECK9-NEXT:  entry:
8707 // CHECK9-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
8708 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8709 // CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
8710 // CHECK9-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
8711 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8712 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
8713 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
8714 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[SIVAR_CASTED]], align 4
8715 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
8716 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l155.omp_outlined, ptr [[TMP0]], i64 [[TMP2]])
8717 // CHECK9-NEXT:    ret void
8718 //
8719 //
8720 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l155.omp_outlined
8721 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8722 // CHECK9-NEXT:  entry:
8723 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
8724 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
8725 // CHECK9-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
8726 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8727 // CHECK9-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
8728 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8729 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8730 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8731 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8732 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8733 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8734 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8735 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_23:%.*]], align 8
8736 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
8737 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
8738 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
8739 // CHECK9-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
8740 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8741 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
8742 // CHECK9-NEXT:    store i32 -1, ptr [[BIT_VAR1]], align 4
8743 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
8744 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
8745 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
8746 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
8747 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
8748 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
8749 // 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)
8750 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8751 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
8752 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8753 // CHECK9:       cond.true:
8754 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8755 // CHECK9:       cond.false:
8756 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8757 // CHECK9-NEXT:    br label [[COND_END]]
8758 // CHECK9:       cond.end:
8759 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8760 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
8761 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
8762 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
8763 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8764 // CHECK9:       omp.inner.for.cond:
8765 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8766 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8767 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8768 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8769 // CHECK9:       omp.inner.for.body:
8770 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8771 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8772 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8773 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
8774 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
8775 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
8776 // CHECK9-NEXT:    [[AND:%.*]] = and i32 [[TMP9]], [[TMP10]]
8777 // CHECK9-NEXT:    store i32 [[AND]], ptr [[BIT_VAR1]], align 4
8778 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_23]], ptr [[REF_TMP]], i32 0, i32 0
8779 // CHECK9-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP11]], align 8
8780 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_5clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
8781 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8782 // CHECK9:       omp.body.continue:
8783 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8784 // CHECK9:       omp.inner.for.inc:
8785 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8786 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
8787 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
8788 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
8789 // CHECK9:       omp.inner.for.end:
8790 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8791 // CHECK9:       omp.loop.exit:
8792 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
8793 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
8794 // CHECK9-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP13]], align 8
8795 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l155.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
8796 // CHECK9-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
8797 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
8798 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
8799 // CHECK9-NEXT:    ]
8800 // CHECK9:       .omp.reduction.case1:
8801 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP0]], align 4
8802 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
8803 // CHECK9-NEXT:    [[AND4:%.*]] = and i32 [[TMP15]], [[TMP16]]
8804 // CHECK9-NEXT:    store i32 [[AND4]], ptr [[TMP0]], align 4
8805 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8806 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8807 // CHECK9:       .omp.reduction.case2:
8808 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
8809 // CHECK9-NEXT:    [[TMP18:%.*]] = atomicrmw and ptr [[TMP0]], i32 [[TMP17]] monotonic, align 4
8810 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8811 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8812 // CHECK9:       .omp.reduction.default:
8813 // CHECK9-NEXT:    ret void
8814 //
8815 //
8816 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l155.omp_outlined.omp.reduction.reduction_func
8817 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
8818 // CHECK9-NEXT:  entry:
8819 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
8820 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
8821 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
8822 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
8823 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
8824 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
8825 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
8826 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
8827 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
8828 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
8829 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
8830 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
8831 // CHECK9-NEXT:    [[AND:%.*]] = and i32 [[TMP8]], [[TMP9]]
8832 // CHECK9-NEXT:    store i32 [[AND]], ptr [[TMP7]], align 4
8833 // CHECK9-NEXT:    ret void
8834 //
8835 //
8836 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l166
8837 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8838 // CHECK9-NEXT:  entry:
8839 // CHECK9-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
8840 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8841 // CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
8842 // CHECK9-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
8843 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8844 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
8845 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
8846 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[SIVAR_CASTED]], align 4
8847 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
8848 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l166.omp_outlined, ptr [[TMP0]], i64 [[TMP2]])
8849 // CHECK9-NEXT:    ret void
8850 //
8851 //
8852 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l166.omp_outlined
8853 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8854 // CHECK9-NEXT:  entry:
8855 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
8856 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
8857 // CHECK9-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
8858 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8859 // CHECK9-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
8860 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8861 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8862 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8863 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8864 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8865 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8866 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8867 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_24:%.*]], align 8
8868 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
8869 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
8870 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
8871 // CHECK9-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
8872 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8873 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
8874 // CHECK9-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
8875 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
8876 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
8877 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
8878 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
8879 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
8880 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
8881 // 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)
8882 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8883 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
8884 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8885 // CHECK9:       cond.true:
8886 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8887 // CHECK9:       cond.false:
8888 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8889 // CHECK9-NEXT:    br label [[COND_END]]
8890 // CHECK9:       cond.end:
8891 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8892 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
8893 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
8894 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
8895 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8896 // CHECK9:       omp.inner.for.cond:
8897 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8898 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
8899 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8900 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8901 // CHECK9:       omp.inner.for.body:
8902 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8903 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8904 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8905 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
8906 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
8907 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
8908 // CHECK9-NEXT:    [[OR:%.*]] = or i32 [[TMP9]], [[TMP10]]
8909 // CHECK9-NEXT:    store i32 [[OR]], ptr [[BIT_VAR1]], align 4
8910 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_24]], ptr [[REF_TMP]], i32 0, i32 0
8911 // CHECK9-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP11]], align 8
8912 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_6clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
8913 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8914 // CHECK9:       omp.body.continue:
8915 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8916 // CHECK9:       omp.inner.for.inc:
8917 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
8918 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
8919 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
8920 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
8921 // CHECK9:       omp.inner.for.end:
8922 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8923 // CHECK9:       omp.loop.exit:
8924 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
8925 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
8926 // CHECK9-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP13]], align 8
8927 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l166.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
8928 // CHECK9-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
8929 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
8930 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
8931 // CHECK9-NEXT:    ]
8932 // CHECK9:       .omp.reduction.case1:
8933 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP0]], align 4
8934 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
8935 // CHECK9-NEXT:    [[OR4:%.*]] = or i32 [[TMP15]], [[TMP16]]
8936 // CHECK9-NEXT:    store i32 [[OR4]], ptr [[TMP0]], align 4
8937 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8938 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8939 // CHECK9:       .omp.reduction.case2:
8940 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
8941 // CHECK9-NEXT:    [[TMP18:%.*]] = atomicrmw or ptr [[TMP0]], i32 [[TMP17]] monotonic, align 4
8942 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
8943 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
8944 // CHECK9:       .omp.reduction.default:
8945 // CHECK9-NEXT:    ret void
8946 //
8947 //
8948 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l166.omp_outlined.omp.reduction.reduction_func
8949 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
8950 // CHECK9-NEXT:  entry:
8951 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
8952 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
8953 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
8954 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
8955 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
8956 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
8957 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
8958 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
8959 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
8960 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
8961 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
8962 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
8963 // CHECK9-NEXT:    [[OR:%.*]] = or i32 [[TMP8]], [[TMP9]]
8964 // CHECK9-NEXT:    store i32 [[OR]], ptr [[TMP7]], align 4
8965 // CHECK9-NEXT:    ret void
8966 //
8967 //
8968 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l177
8969 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8970 // CHECK9-NEXT:  entry:
8971 // CHECK9-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
8972 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8973 // CHECK9-NEXT:    [[SIVAR_CASTED:%.*]] = alloca i64, align 8
8974 // CHECK9-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
8975 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
8976 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
8977 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, ptr [[SIVAR_ADDR]], align 4
8978 // CHECK9-NEXT:    store i32 [[TMP1]], ptr [[SIVAR_CASTED]], align 4
8979 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[SIVAR_CASTED]], align 8
8980 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l177.omp_outlined, ptr [[TMP0]], i64 [[TMP2]])
8981 // CHECK9-NEXT:    ret void
8982 //
8983 //
8984 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l177.omp_outlined
8985 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[BIT_VAR:%.*]], i64 noundef [[SIVAR:%.*]]) #[[ATTR2]] {
8986 // CHECK9-NEXT:  entry:
8987 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
8988 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
8989 // CHECK9-NEXT:    [[BIT_VAR_ADDR:%.*]] = alloca ptr, align 8
8990 // CHECK9-NEXT:    [[SIVAR_ADDR:%.*]] = alloca i64, align 8
8991 // CHECK9-NEXT:    [[BIT_VAR1:%.*]] = alloca i32, align 4
8992 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8993 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8994 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8995 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8996 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8997 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8998 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8999 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_25:%.*]], align 8
9000 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
9001 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
9002 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
9003 // CHECK9-NEXT:    store ptr [[BIT_VAR]], ptr [[BIT_VAR_ADDR]], align 8
9004 // CHECK9-NEXT:    store i64 [[SIVAR]], ptr [[SIVAR_ADDR]], align 8
9005 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[BIT_VAR_ADDR]], align 8
9006 // CHECK9-NEXT:    store i32 0, ptr [[BIT_VAR1]], align 4
9007 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
9008 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
9009 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
9010 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
9011 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
9012 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
9013 // 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)
9014 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9015 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
9016 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9017 // CHECK9:       cond.true:
9018 // CHECK9-NEXT:    br label [[COND_END:%.*]]
9019 // CHECK9:       cond.false:
9020 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9021 // CHECK9-NEXT:    br label [[COND_END]]
9022 // CHECK9:       cond.end:
9023 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9024 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
9025 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
9026 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
9027 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9028 // CHECK9:       omp.inner.for.cond:
9029 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9030 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9031 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9032 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9033 // CHECK9:       omp.inner.for.body:
9034 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9035 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
9036 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9037 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
9038 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
9039 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
9040 // CHECK9-NEXT:    [[XOR:%.*]] = xor i32 [[TMP9]], [[TMP10]]
9041 // CHECK9-NEXT:    store i32 [[XOR]], ptr [[BIT_VAR1]], align 4
9042 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[CLASS_ANON_25]], ptr [[REF_TMP]], i32 0, i32 0
9043 // CHECK9-NEXT:    store ptr [[SIVAR_ADDR]], ptr [[TMP11]], align 8
9044 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_7clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
9045 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9046 // CHECK9:       omp.body.continue:
9047 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9048 // CHECK9:       omp.inner.for.inc:
9049 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9050 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
9051 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
9052 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
9053 // CHECK9:       omp.inner.for.end:
9054 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9055 // CHECK9:       omp.loop.exit:
9056 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
9057 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
9058 // CHECK9-NEXT:    store ptr [[BIT_VAR1]], ptr [[TMP13]], align 8
9059 // CHECK9-NEXT:    [[TMP14:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l177.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
9060 // CHECK9-NEXT:    switch i32 [[TMP14]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
9061 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
9062 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
9063 // CHECK9-NEXT:    ]
9064 // CHECK9:       .omp.reduction.case1:
9065 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP0]], align 4
9066 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
9067 // CHECK9-NEXT:    [[XOR4:%.*]] = xor i32 [[TMP15]], [[TMP16]]
9068 // CHECK9-NEXT:    store i32 [[XOR4]], ptr [[TMP0]], align 4
9069 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
9070 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
9071 // CHECK9:       .omp.reduction.case2:
9072 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[BIT_VAR1]], align 4
9073 // CHECK9-NEXT:    [[TMP18:%.*]] = atomicrmw xor ptr [[TMP0]], i32 [[TMP17]] monotonic, align 4
9074 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
9075 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
9076 // CHECK9:       .omp.reduction.default:
9077 // CHECK9-NEXT:    ret void
9078 //
9079 //
9080 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l177.omp_outlined.omp.reduction.reduction_func
9081 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
9082 // CHECK9-NEXT:  entry:
9083 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
9084 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
9085 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
9086 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
9087 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
9088 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
9089 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
9090 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
9091 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
9092 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
9093 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
9094 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
9095 // CHECK9-NEXT:    [[XOR:%.*]] = xor i32 [[TMP8]], [[TMP9]]
9096 // CHECK9-NEXT:    store i32 [[XOR]], ptr [[TMP7]], align 4
9097 // CHECK9-NEXT:    ret void
9098 //
9099 //
9100 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l188
9101 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MAX_VAR:%.*]]) #[[ATTR2]] {
9102 // CHECK9-NEXT:  entry:
9103 // CHECK9-NEXT:    [[MAX_VAR_ADDR:%.*]] = alloca ptr, align 8
9104 // CHECK9-NEXT:    store ptr [[MAX_VAR]], ptr [[MAX_VAR_ADDR]], align 8
9105 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_VAR_ADDR]], align 8
9106 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l188.omp_outlined, ptr [[TMP0]])
9107 // CHECK9-NEXT:    ret void
9108 //
9109 //
9110 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l188.omp_outlined
9111 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[MAX_VAR:%.*]]) #[[ATTR2]] {
9112 // CHECK9-NEXT:  entry:
9113 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
9114 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
9115 // CHECK9-NEXT:    [[MAX_VAR_ADDR:%.*]] = alloca ptr, align 8
9116 // CHECK9-NEXT:    [[MAX_VAR1:%.*]] = alloca i32, align 4
9117 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9118 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9119 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9120 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9121 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9122 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9123 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
9124 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_26:%.*]], align 8
9125 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
9126 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
9127 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
9128 // CHECK9-NEXT:    store ptr [[MAX_VAR]], ptr [[MAX_VAR_ADDR]], align 8
9129 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MAX_VAR_ADDR]], align 8
9130 // CHECK9-NEXT:    store i32 -2147483648, ptr [[MAX_VAR1]], align 4
9131 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
9132 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
9133 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
9134 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
9135 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
9136 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
9137 // 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)
9138 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9139 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
9140 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9141 // CHECK9:       cond.true:
9142 // CHECK9-NEXT:    br label [[COND_END:%.*]]
9143 // CHECK9:       cond.false:
9144 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9145 // CHECK9-NEXT:    br label [[COND_END]]
9146 // CHECK9:       cond.end:
9147 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9148 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
9149 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
9150 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
9151 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9152 // CHECK9:       omp.inner.for.cond:
9153 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9154 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9155 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9156 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9157 // CHECK9:       omp.inner.for.body:
9158 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9159 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
9160 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9161 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
9162 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[MAX_VAR1]], align 4
9163 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
9164 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp sge i32 [[TMP9]], [[TMP10]]
9165 // CHECK9-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
9166 // CHECK9:       cond.true4:
9167 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[MAX_VAR1]], align 4
9168 // CHECK9-NEXT:    br label [[COND_END6:%.*]]
9169 // CHECK9:       cond.false5:
9170 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
9171 // CHECK9-NEXT:    br label [[COND_END6]]
9172 // CHECK9:       cond.end6:
9173 // CHECK9-NEXT:    [[COND7:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE4]] ], [ [[TMP12]], [[COND_FALSE5]] ]
9174 // CHECK9-NEXT:    store i32 [[COND7]], ptr [[MAX_VAR1]], align 4
9175 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_26]], ptr [[REF_TMP]], i32 0, i32 0
9176 // CHECK9-NEXT:    store ptr [[MAX_VAR1]], ptr [[TMP13]], align 8
9177 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_8clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
9178 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9179 // CHECK9:       omp.body.continue:
9180 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9181 // CHECK9:       omp.inner.for.inc:
9182 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9183 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP14]], 1
9184 // CHECK9-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
9185 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
9186 // CHECK9:       omp.inner.for.end:
9187 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9188 // CHECK9:       omp.loop.exit:
9189 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
9190 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
9191 // CHECK9-NEXT:    store ptr [[MAX_VAR1]], ptr [[TMP15]], align 8
9192 // CHECK9-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l188.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
9193 // CHECK9-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
9194 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
9195 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
9196 // CHECK9-NEXT:    ]
9197 // CHECK9:       .omp.reduction.case1:
9198 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP0]], align 4
9199 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[MAX_VAR1]], align 4
9200 // CHECK9-NEXT:    [[CMP9:%.*]] = icmp sgt i32 [[TMP17]], [[TMP18]]
9201 // CHECK9-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
9202 // CHECK9:       cond.true10:
9203 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
9204 // CHECK9-NEXT:    br label [[COND_END12:%.*]]
9205 // CHECK9:       cond.false11:
9206 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[MAX_VAR1]], align 4
9207 // CHECK9-NEXT:    br label [[COND_END12]]
9208 // CHECK9:       cond.end12:
9209 // CHECK9-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP19]], [[COND_TRUE10]] ], [ [[TMP20]], [[COND_FALSE11]] ]
9210 // CHECK9-NEXT:    store i32 [[COND13]], ptr [[TMP0]], align 4
9211 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
9212 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
9213 // CHECK9:       .omp.reduction.case2:
9214 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[MAX_VAR1]], align 4
9215 // CHECK9-NEXT:    [[TMP22:%.*]] = atomicrmw max ptr [[TMP0]], i32 [[TMP21]] monotonic, align 4
9216 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
9217 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
9218 // CHECK9:       .omp.reduction.default:
9219 // CHECK9-NEXT:    ret void
9220 //
9221 //
9222 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l188.omp_outlined.omp.reduction.reduction_func
9223 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
9224 // CHECK9-NEXT:  entry:
9225 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
9226 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
9227 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
9228 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
9229 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
9230 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
9231 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
9232 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
9233 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
9234 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
9235 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
9236 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
9237 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
9238 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9239 // CHECK9:       cond.true:
9240 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
9241 // CHECK9-NEXT:    br label [[COND_END:%.*]]
9242 // CHECK9:       cond.false:
9243 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
9244 // CHECK9-NEXT:    br label [[COND_END]]
9245 // CHECK9:       cond.end:
9246 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
9247 // CHECK9-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
9248 // CHECK9-NEXT:    ret void
9249 //
9250 //
9251 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l199
9252 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[MIN_VAR:%.*]]) #[[ATTR2]] {
9253 // CHECK9-NEXT:  entry:
9254 // CHECK9-NEXT:    [[MIN_VAR_ADDR:%.*]] = alloca ptr, align 8
9255 // CHECK9-NEXT:    store ptr [[MIN_VAR]], ptr [[MIN_VAR_ADDR]], align 8
9256 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_VAR_ADDR]], align 8
9257 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l199.omp_outlined, ptr [[TMP0]])
9258 // CHECK9-NEXT:    ret void
9259 //
9260 //
9261 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l199.omp_outlined
9262 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[MIN_VAR:%.*]]) #[[ATTR2]] {
9263 // CHECK9-NEXT:  entry:
9264 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
9265 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
9266 // CHECK9-NEXT:    [[MIN_VAR_ADDR:%.*]] = alloca ptr, align 8
9267 // CHECK9-NEXT:    [[MIN_VAR1:%.*]] = alloca i32, align 4
9268 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9269 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9270 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9271 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9272 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9273 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9274 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
9275 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_27:%.*]], align 8
9276 // CHECK9-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
9277 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
9278 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
9279 // CHECK9-NEXT:    store ptr [[MIN_VAR]], ptr [[MIN_VAR_ADDR]], align 8
9280 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[MIN_VAR_ADDR]], align 8
9281 // CHECK9-NEXT:    store i32 2147483647, ptr [[MIN_VAR1]], align 4
9282 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
9283 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
9284 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
9285 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
9286 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
9287 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
9288 // 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)
9289 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9290 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1
9291 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9292 // CHECK9:       cond.true:
9293 // CHECK9-NEXT:    br label [[COND_END:%.*]]
9294 // CHECK9:       cond.false:
9295 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9296 // CHECK9-NEXT:    br label [[COND_END]]
9297 // CHECK9:       cond.end:
9298 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9299 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
9300 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
9301 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
9302 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9303 // CHECK9:       omp.inner.for.cond:
9304 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9305 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
9306 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9307 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9308 // CHECK9:       omp.inner.for.body:
9309 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9310 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
9311 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9312 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
9313 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[MIN_VAR1]], align 4
9314 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[I]], align 4
9315 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
9316 // CHECK9-NEXT:    br i1 [[CMP3]], label [[COND_TRUE4:%.*]], label [[COND_FALSE5:%.*]]
9317 // CHECK9:       cond.true4:
9318 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[MIN_VAR1]], align 4
9319 // CHECK9-NEXT:    br label [[COND_END6:%.*]]
9320 // CHECK9:       cond.false5:
9321 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[I]], align 4
9322 // CHECK9-NEXT:    br label [[COND_END6]]
9323 // CHECK9:       cond.end6:
9324 // CHECK9-NEXT:    [[COND7:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE4]] ], [ [[TMP12]], [[COND_FALSE5]] ]
9325 // CHECK9-NEXT:    store i32 [[COND7]], ptr [[MIN_VAR1]], align 4
9326 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_27]], ptr [[REF_TMP]], i32 0, i32 0
9327 // CHECK9-NEXT:    store ptr [[MIN_VAR1]], ptr [[TMP13]], align 8
9328 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_9clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
9329 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9330 // CHECK9:       omp.body.continue:
9331 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9332 // CHECK9:       omp.inner.for.inc:
9333 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
9334 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP14]], 1
9335 // CHECK9-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
9336 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
9337 // CHECK9:       omp.inner.for.end:
9338 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9339 // CHECK9:       omp.loop.exit:
9340 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
9341 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
9342 // CHECK9-NEXT:    store ptr [[MIN_VAR1]], ptr [[TMP15]], align 8
9343 // CHECK9-NEXT:    [[TMP16:%.*]] = call i32 @__kmpc_reduce(ptr @[[GLOB2]], i32 [[TMP2]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l199.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
9344 // CHECK9-NEXT:    switch i32 [[TMP16]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
9345 // CHECK9-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
9346 // CHECK9-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
9347 // CHECK9-NEXT:    ]
9348 // CHECK9:       .omp.reduction.case1:
9349 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP0]], align 4
9350 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[MIN_VAR1]], align 4
9351 // CHECK9-NEXT:    [[CMP9:%.*]] = icmp slt i32 [[TMP17]], [[TMP18]]
9352 // CHECK9-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
9353 // CHECK9:       cond.true10:
9354 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP0]], align 4
9355 // CHECK9-NEXT:    br label [[COND_END12:%.*]]
9356 // CHECK9:       cond.false11:
9357 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[MIN_VAR1]], align 4
9358 // CHECK9-NEXT:    br label [[COND_END12]]
9359 // CHECK9:       cond.end12:
9360 // CHECK9-NEXT:    [[COND13:%.*]] = phi i32 [ [[TMP19]], [[COND_TRUE10]] ], [ [[TMP20]], [[COND_FALSE11]] ]
9361 // CHECK9-NEXT:    store i32 [[COND13]], ptr [[TMP0]], align 4
9362 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
9363 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
9364 // CHECK9:       .omp.reduction.case2:
9365 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[MIN_VAR1]], align 4
9366 // CHECK9-NEXT:    [[TMP22:%.*]] = atomicrmw min ptr [[TMP0]], i32 [[TMP21]] monotonic, align 4
9367 // CHECK9-NEXT:    call void @__kmpc_end_reduce(ptr @[[GLOB2]], i32 [[TMP2]], ptr @.gomp_critical_user_.reduction.var)
9368 // CHECK9-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
9369 // CHECK9:       .omp.reduction.default:
9370 // CHECK9-NEXT:    ret void
9371 //
9372 //
9373 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l199.omp_outlined.omp.reduction.reduction_func
9374 // CHECK9-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
9375 // CHECK9-NEXT:  entry:
9376 // CHECK9-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
9377 // CHECK9-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
9378 // CHECK9-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
9379 // CHECK9-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
9380 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
9381 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
9382 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
9383 // CHECK9-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
9384 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
9385 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
9386 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
9387 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
9388 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP8]], [[TMP9]]
9389 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9390 // CHECK9:       cond.true:
9391 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP7]], align 4
9392 // CHECK9-NEXT:    br label [[COND_END:%.*]]
9393 // CHECK9:       cond.false:
9394 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 4
9395 // CHECK9-NEXT:    br label [[COND_END]]
9396 // CHECK9:       cond.end:
9397 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
9398 // CHECK9-NEXT:    store i32 [[COND]], ptr [[TMP7]], align 4
9399 // CHECK9-NEXT:    ret void
9400 //
9401 //
9402 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
9403 // CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
9404 // CHECK9-NEXT:  entry:
9405 // CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
9406 // CHECK9-NEXT:    ret void
9407 //