clang/test/OpenMP/teams_distribute_dist_schedule_codegen.cpp

   1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
   2 // expected-no-diagnostics
   3 #ifndef HEADER
   4 #define HEADER
   5
   6 // Test host codegen.
   7 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
   8 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
   9 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
  10 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
  11 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  12 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
  13
  14 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  15 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  16 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  17 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  18 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  19 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  20 #ifdef CK1
  21
  22 template <typename T, int X, long long Y>
  23 struct SS{
  24   T a[X];
  25   float b;
  26   int foo(void) {
  27
  28     #pragma omp target
  29     #pragma omp teams distribute
  30     for(int i = 0; i < X; i++) {
  31       a[i] = (T)0;
  32     }
  33     #pragma omp target
  34     #pragma omp teams distribute dist_schedule(static)
  35     for(int i = 0; i < X; i++) {
  36       a[i] = (T)0;
  37     }
  38     #pragma omp target
  39     #pragma omp teams distribute dist_schedule(static, X/2)
  40     for(int i = 0; i < X; i++) {
  41       a[i] = (T)0;
  42     }
  43
  44
  45
  46
  47
  48
  49     return a[0];
  50   }
  51 };
  52
  53 int teams_template_struct(void) {
  54   SS<int, 123, 456> V;
  55   return V.foo();
  56
  57 }
  58 #endif // CK1
  59
  60 // Test host codegen.
  61 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
  62 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  63 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
  64 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
  65 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  66 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
  67
  68 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  69 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  70 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  71 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  72 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  73 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  74 #ifdef CK2
  75
  76 template <typename T, int n>
  77 int tmain(T argc) {
  78   T a[n];
  79 #pragma omp target
  80 #pragma omp teams distribute
  81   for(int i = 0; i < n; i++) {
  82     a[i] = (T)0;
  83   }
  84 #pragma omp target
  85 #pragma omp teams distribute dist_schedule(static)
  86   for(int i = 0; i < n; i++) {
  87     a[i] = (T)0;
  88   }
  89 #pragma omp target
  90 #pragma omp teams distribute dist_schedule(static, n)
  91   for(int i = 0; i < n; i++) {
  92     a[i] = (T)0;
  93   }
  94   return 0;
  95 }
  96
  97 int main (int argc, char **argv) {
  98   int n = 100;
  99   int a[n];
 100 #pragma omp target
 101 #pragma omp teams distribute
 102   for(int i = 0; i < n; i++) {
 103     a[i] = 0;
 104   }
 105 #pragma omp target
 106 #pragma omp teams distribute dist_schedule(static)
 107   for(int i = 0; i < n; i++) {
 108     a[i] = 0;
 109   }
 110 #pragma omp target
 111 #pragma omp teams distribute dist_schedule(static, n)
 112   for(int i = 0; i < n; i++) {
 113     a[i] = 0;
 114   }
 115   return tmain<int, 10>(argc);
 116 }
 117
 118
 119
 120
 121
 122
 123
 124
 125
 126
 127
 128
 129
 130
 131
 132 #endif // CK2
 133 #endif // #ifndef HEADER
 134 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 135 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 136 // CHECK1-NEXT:  entry:
 137 // CHECK1-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 138 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(496) [[V]])
 139 // CHECK1-NEXT:    ret i32 [[CALL]]
 140 //
 141 //
 142 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 143 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat {
 144 // CHECK1-NEXT:  entry:
 145 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 146 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 147 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 148 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 149 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 150 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 151 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 8
 152 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 8
 153 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 8
 154 // CHECK1-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
 155 // CHECK1-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 156 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x ptr], align 8
 157 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x ptr], align 8
 158 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x ptr], align 8
 159 // CHECK1-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
 160 // CHECK1-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 161 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 162 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 163 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 164 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 165 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 8
 166 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 167 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
 168 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 169 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 170 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 171 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 172 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 173 // CHECK1-NEXT:    store i32 2, ptr [[TMP5]], align 4
 174 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 175 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
 176 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 177 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 178 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 179 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 180 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 181 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 8
 182 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 183 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 8
 184 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 185 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 186 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 187 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 188 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 189 // CHECK1-NEXT:    store i64 123, ptr [[TMP13]], align 8
 190 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 191 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
 192 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 193 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 194 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 195 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 196 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 197 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
 198 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 199 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 200 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 201 // CHECK1:       omp_offload.failed:
 202 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR3:[0-9]+]]
 203 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 204 // CHECK1:       omp_offload.cont:
 205 // CHECK1-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 206 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 207 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP20]], align 8
 208 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 209 // CHECK1-NEXT:    store ptr [[A2]], ptr [[TMP21]], align 8
 210 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
 211 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 212 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 213 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 214 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 0
 215 // CHECK1-NEXT:    store i32 2, ptr [[TMP25]], align 4
 216 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 1
 217 // CHECK1-NEXT:    store i32 1, ptr [[TMP26]], align 4
 218 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 2
 219 // CHECK1-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
 220 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 3
 221 // CHECK1-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
 222 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 4
 223 // CHECK1-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP29]], align 8
 224 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 5
 225 // CHECK1-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP30]], align 8
 226 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 6
 227 // CHECK1-NEXT:    store ptr null, ptr [[TMP31]], align 8
 228 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 7
 229 // CHECK1-NEXT:    store ptr null, ptr [[TMP32]], align 8
 230 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 8
 231 // CHECK1-NEXT:    store i64 123, ptr [[TMP33]], align 8
 232 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 9
 233 // CHECK1-NEXT:    store i64 0, ptr [[TMP34]], align 8
 234 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 10
 235 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
 236 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 11
 237 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
 238 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 12
 239 // CHECK1-NEXT:    store i32 0, ptr [[TMP37]], align 4
 240 // CHECK1-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.region_id, ptr [[KERNEL_ARGS7]])
 241 // CHECK1-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 242 // CHECK1-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
 243 // CHECK1:       omp_offload.failed8:
 244 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(ptr [[THIS1]]) #[[ATTR3]]
 245 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
 246 // CHECK1:       omp_offload.cont9:
 247 // CHECK1-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 248 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 249 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP40]], align 8
 250 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 251 // CHECK1-NEXT:    store ptr [[A10]], ptr [[TMP41]], align 8
 252 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
 253 // CHECK1-NEXT:    store ptr null, ptr [[TMP42]], align 8
 254 // CHECK1-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 255 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 256 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
 257 // CHECK1-NEXT:    store i32 2, ptr [[TMP45]], align 4
 258 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
 259 // CHECK1-NEXT:    store i32 1, ptr [[TMP46]], align 4
 260 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
 261 // CHECK1-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 8
 262 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
 263 // CHECK1-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 8
 264 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
 265 // CHECK1-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP49]], align 8
 266 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
 267 // CHECK1-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP50]], align 8
 268 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
 269 // CHECK1-NEXT:    store ptr null, ptr [[TMP51]], align 8
 270 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
 271 // CHECK1-NEXT:    store ptr null, ptr [[TMP52]], align 8
 272 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
 273 // CHECK1-NEXT:    store i64 123, ptr [[TMP53]], align 8
 274 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
 275 // CHECK1-NEXT:    store i64 0, ptr [[TMP54]], align 8
 276 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
 277 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
 278 // CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
 279 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
 280 // CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 281 // CHECK1-NEXT:    store i32 0, ptr [[TMP57]], align 4
 282 // CHECK1-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.region_id, ptr [[KERNEL_ARGS15]])
 283 // CHECK1-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
 284 // CHECK1-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 285 // CHECK1:       omp_offload.failed16:
 286 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(ptr [[THIS1]]) #[[ATTR3]]
 287 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
 288 // CHECK1:       omp_offload.cont17:
 289 // CHECK1-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 290 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A18]], i64 0, i64 0
 291 // CHECK1-NEXT:    [[TMP60:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 292 // CHECK1-NEXT:    ret i32 [[TMP60]]
 293 //
 294 //
 295 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 296 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 297 // CHECK1-NEXT:  entry:
 298 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 299 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 300 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 301 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 302 // CHECK1-NEXT:    ret void
 303 //
 304 //
 305 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 306 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
 307 // CHECK1-NEXT:  entry:
 308 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 309 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 310 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 311 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 312 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 313 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 314 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 315 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 316 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 317 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 318 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 319 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 320 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 321 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 322 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 323 // CHECK1-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 324 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 325 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 326 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 327 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 328 // 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)
 329 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 330 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 331 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 332 // CHECK1:       cond.true:
 333 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 334 // CHECK1:       cond.false:
 335 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 336 // CHECK1-NEXT:    br label [[COND_END]]
 337 // CHECK1:       cond.end:
 338 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 339 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 340 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 341 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 342 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 343 // CHECK1:       omp.inner.for.cond:
 344 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 345 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 346 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 347 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 348 // CHECK1:       omp.inner.for.body:
 349 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 350 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 351 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 352 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 353 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 354 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 355 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 356 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 357 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 358 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 359 // CHECK1:       omp.body.continue:
 360 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 361 // CHECK1:       omp.inner.for.inc:
 362 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 363 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 364 // CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 365 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 366 // CHECK1:       omp.inner.for.end:
 367 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 368 // CHECK1:       omp.loop.exit:
 369 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 370 // CHECK1-NEXT:    ret void
 371 //
 372 //
 373 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
 374 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 375 // CHECK1-NEXT:  entry:
 376 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 377 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 378 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 379 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.omp_outlined, ptr [[TMP0]])
 380 // CHECK1-NEXT:    ret void
 381 //
 382 //
 383 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.omp_outlined
 384 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2]] {
 385 // CHECK1-NEXT:  entry:
 386 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 387 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 388 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 389 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 390 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 391 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 392 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 393 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 394 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 395 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 396 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 397 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 398 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 399 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 400 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 401 // CHECK1-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 402 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 403 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 404 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 405 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 406 // 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)
 407 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 408 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 409 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 410 // CHECK1:       cond.true:
 411 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 412 // CHECK1:       cond.false:
 413 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 414 // CHECK1-NEXT:    br label [[COND_END]]
 415 // CHECK1:       cond.end:
 416 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 417 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 418 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 419 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 420 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 421 // CHECK1:       omp.inner.for.cond:
 422 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 423 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 424 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 425 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 426 // CHECK1:       omp.inner.for.body:
 427 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 428 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 429 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 430 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 431 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 432 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 433 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
 434 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 435 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 436 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 437 // CHECK1:       omp.body.continue:
 438 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 439 // CHECK1:       omp.inner.for.inc:
 440 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 441 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 442 // CHECK1-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 443 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 444 // CHECK1:       omp.inner.for.end:
 445 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 446 // CHECK1:       omp.loop.exit:
 447 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 448 // CHECK1-NEXT:    ret void
 449 //
 450 //
 451 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
 452 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 453 // CHECK1-NEXT:  entry:
 454 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 455 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 456 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 457 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.omp_outlined, ptr [[TMP0]])
 458 // CHECK1-NEXT:    ret void
 459 //
 460 //
 461 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.omp_outlined
 462 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2]] {
 463 // CHECK1-NEXT:  entry:
 464 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 465 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 466 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 467 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 468 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 469 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 470 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 471 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 472 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 473 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 474 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 475 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 476 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 477 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 478 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 479 // CHECK1-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 480 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 481 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 482 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 483 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 484 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 61)
 485 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 486 // CHECK1:       omp.dispatch.cond:
 487 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 488 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 489 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 490 // CHECK1:       cond.true:
 491 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 492 // CHECK1:       cond.false:
 493 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 494 // CHECK1-NEXT:    br label [[COND_END]]
 495 // CHECK1:       cond.end:
 496 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 497 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 498 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 499 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 500 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 501 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 502 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 503 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 504 // CHECK1:       omp.dispatch.body:
 505 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 506 // CHECK1:       omp.inner.for.cond:
 507 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8:![0-9]+]]
 508 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP8]]
 509 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 510 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 511 // CHECK1:       omp.inner.for.body:
 512 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 513 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 514 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 515 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP8]]
 516 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 517 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP8]]
 518 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
 519 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
 520 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP8]]
 521 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 522 // CHECK1:       omp.body.continue:
 523 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 524 // CHECK1:       omp.inner.for.inc:
 525 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 526 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
 527 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP8]]
 528 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
 529 // CHECK1:       omp.inner.for.end:
 530 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 531 // CHECK1:       omp.dispatch.inc:
 532 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 533 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 534 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
 535 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
 536 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 537 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 538 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
 539 // CHECK1-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
 540 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
 541 // CHECK1:       omp.dispatch.end:
 542 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 543 // CHECK1-NEXT:    ret void
 544 //
 545 //
 546 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
 547 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
 548 // CHECK1-NEXT:  entry:
 549 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
 550 // CHECK1-NEXT:    ret void
 551 //
 552 //
 553 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 554 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 555 // CHECK3-NEXT:  entry:
 556 // CHECK3-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 557 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(496) [[V]])
 558 // CHECK3-NEXT:    ret i32 [[CALL]]
 559 //
 560 //
 561 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 562 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
 563 // CHECK3-NEXT:  entry:
 564 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 565 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
 566 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
 567 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
 568 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 569 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 570 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x ptr], align 4
 571 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x ptr], align 4
 572 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x ptr], align 4
 573 // CHECK3-NEXT:    [[_TMP6:%.*]] = alloca i32, align 4
 574 // CHECK3-NEXT:    [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 575 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x ptr], align 4
 576 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x ptr], align 4
 577 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x ptr], align 4
 578 // CHECK3-NEXT:    [[_TMP14:%.*]] = alloca i32, align 4
 579 // CHECK3-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 580 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 581 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 582 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 583 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 584 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 4
 585 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 586 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
 587 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 588 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
 589 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 590 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 591 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 592 // CHECK3-NEXT:    store i32 2, ptr [[TMP5]], align 4
 593 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 594 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
 595 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 596 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
 597 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 598 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
 599 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 600 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 4
 601 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 602 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 4
 603 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 604 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
 605 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 606 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 607 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 608 // CHECK3-NEXT:    store i64 123, ptr [[TMP13]], align 8
 609 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 610 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
 611 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 612 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 613 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 614 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 615 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 616 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
 617 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 618 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 619 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 620 // CHECK3:       omp_offload.failed:
 621 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR3:[0-9]+]]
 622 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 623 // CHECK3:       omp_offload.cont:
 624 // CHECK3-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 625 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 626 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP20]], align 4
 627 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 628 // CHECK3-NEXT:    store ptr [[A2]], ptr [[TMP21]], align 4
 629 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
 630 // CHECK3-NEXT:    store ptr null, ptr [[TMP22]], align 4
 631 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
 632 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
 633 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 0
 634 // CHECK3-NEXT:    store i32 2, ptr [[TMP25]], align 4
 635 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 1
 636 // CHECK3-NEXT:    store i32 1, ptr [[TMP26]], align 4
 637 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 2
 638 // CHECK3-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 4
 639 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 3
 640 // CHECK3-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 4
 641 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 4
 642 // CHECK3-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP29]], align 4
 643 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 5
 644 // CHECK3-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP30]], align 4
 645 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 6
 646 // CHECK3-NEXT:    store ptr null, ptr [[TMP31]], align 4
 647 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 7
 648 // CHECK3-NEXT:    store ptr null, ptr [[TMP32]], align 4
 649 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 8
 650 // CHECK3-NEXT:    store i64 123, ptr [[TMP33]], align 8
 651 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 9
 652 // CHECK3-NEXT:    store i64 0, ptr [[TMP34]], align 8
 653 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 10
 654 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
 655 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 11
 656 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
 657 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS7]], i32 0, i32 12
 658 // CHECK3-NEXT:    store i32 0, ptr [[TMP37]], align 4
 659 // CHECK3-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.region_id, ptr [[KERNEL_ARGS7]])
 660 // CHECK3-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
 661 // CHECK3-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
 662 // CHECK3:       omp_offload.failed8:
 663 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(ptr [[THIS1]]) #[[ATTR3]]
 664 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT9]]
 665 // CHECK3:       omp_offload.cont9:
 666 // CHECK3-NEXT:    [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 667 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 668 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP40]], align 4
 669 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 670 // CHECK3-NEXT:    store ptr [[A10]], ptr [[TMP41]], align 4
 671 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
 672 // CHECK3-NEXT:    store ptr null, ptr [[TMP42]], align 4
 673 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
 674 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
 675 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
 676 // CHECK3-NEXT:    store i32 2, ptr [[TMP45]], align 4
 677 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
 678 // CHECK3-NEXT:    store i32 1, ptr [[TMP46]], align 4
 679 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
 680 // CHECK3-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 4
 681 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
 682 // CHECK3-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 4
 683 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
 684 // CHECK3-NEXT:    store ptr @.offload_sizes.3, ptr [[TMP49]], align 4
 685 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
 686 // CHECK3-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP50]], align 4
 687 // CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
 688 // CHECK3-NEXT:    store ptr null, ptr [[TMP51]], align 4
 689 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
 690 // CHECK3-NEXT:    store ptr null, ptr [[TMP52]], align 4
 691 // CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
 692 // CHECK3-NEXT:    store i64 123, ptr [[TMP53]], align 8
 693 // CHECK3-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
 694 // CHECK3-NEXT:    store i64 0, ptr [[TMP54]], align 8
 695 // CHECK3-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
 696 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
 697 // CHECK3-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
 698 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
 699 // CHECK3-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 700 // CHECK3-NEXT:    store i32 0, ptr [[TMP57]], align 4
 701 // CHECK3-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.region_id, ptr [[KERNEL_ARGS15]])
 702 // CHECK3-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
 703 // CHECK3-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 704 // CHECK3:       omp_offload.failed16:
 705 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(ptr [[THIS1]]) #[[ATTR3]]
 706 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
 707 // CHECK3:       omp_offload.cont17:
 708 // CHECK3-NEXT:    [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 709 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A18]], i32 0, i32 0
 710 // CHECK3-NEXT:    [[TMP60:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 711 // CHECK3-NEXT:    ret i32 [[TMP60]]
 712 //
 713 //
 714 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 715 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 716 // CHECK3-NEXT:  entry:
 717 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 718 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 719 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 720 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 721 // CHECK3-NEXT:    ret void
 722 //
 723 //
 724 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 725 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
 726 // CHECK3-NEXT:  entry:
 727 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 728 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 729 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 730 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 731 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 732 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 733 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 734 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 735 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 736 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 737 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 738 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 739 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 740 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 741 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 742 // CHECK3-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 743 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 744 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 745 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 746 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 747 // 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)
 748 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 749 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 750 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 751 // CHECK3:       cond.true:
 752 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 753 // CHECK3:       cond.false:
 754 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 755 // CHECK3-NEXT:    br label [[COND_END]]
 756 // CHECK3:       cond.end:
 757 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 758 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 759 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 760 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 761 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 762 // CHECK3:       omp.inner.for.cond:
 763 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 764 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 765 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 766 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 767 // CHECK3:       omp.inner.for.body:
 768 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 769 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 770 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 771 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 772 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 773 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 774 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP9]]
 775 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 776 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 777 // CHECK3:       omp.body.continue:
 778 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 779 // CHECK3:       omp.inner.for.inc:
 780 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 781 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 782 // CHECK3-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 783 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 784 // CHECK3:       omp.inner.for.end:
 785 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 786 // CHECK3:       omp.loop.exit:
 787 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 788 // CHECK3-NEXT:    ret void
 789 //
 790 //
 791 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
 792 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 793 // CHECK3-NEXT:  entry:
 794 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 795 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 796 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 797 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.omp_outlined, ptr [[TMP0]])
 798 // CHECK3-NEXT:    ret void
 799 //
 800 //
 801 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.omp_outlined
 802 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2]] {
 803 // CHECK3-NEXT:  entry:
 804 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 805 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 806 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 807 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 808 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 809 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 810 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 811 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 812 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 813 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 814 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 815 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 816 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 817 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 818 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 819 // CHECK3-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 820 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 821 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 822 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 823 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 824 // 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)
 825 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 826 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 827 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 828 // CHECK3:       cond.true:
 829 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 830 // CHECK3:       cond.false:
 831 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 832 // CHECK3-NEXT:    br label [[COND_END]]
 833 // CHECK3:       cond.end:
 834 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 835 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 836 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 837 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 838 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 839 // CHECK3:       omp.inner.for.cond:
 840 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 841 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 842 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 843 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 844 // CHECK3:       omp.inner.for.body:
 845 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 846 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
 847 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 848 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 849 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 850 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
 851 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP9]]
 852 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 853 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 854 // CHECK3:       omp.body.continue:
 855 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 856 // CHECK3:       omp.inner.for.inc:
 857 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 858 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
 859 // CHECK3-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 860 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 861 // CHECK3:       omp.inner.for.end:
 862 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 863 // CHECK3:       omp.loop.exit:
 864 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 865 // CHECK3-NEXT:    ret void
 866 //
 867 //
 868 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
 869 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 870 // CHECK3-NEXT:  entry:
 871 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 872 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 873 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 874 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.omp_outlined, ptr [[TMP0]])
 875 // CHECK3-NEXT:    ret void
 876 //
 877 //
 878 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.omp_outlined
 879 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2]] {
 880 // CHECK3-NEXT:  entry:
 881 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 882 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 883 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 884 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 885 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 886 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 887 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 888 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 889 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 890 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 891 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 892 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 893 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 894 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 895 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 896 // CHECK3-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
 897 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 898 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 899 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 900 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 901 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 61)
 902 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
 903 // CHECK3:       omp.dispatch.cond:
 904 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 905 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
 906 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 907 // CHECK3:       cond.true:
 908 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 909 // CHECK3:       cond.false:
 910 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 911 // CHECK3-NEXT:    br label [[COND_END]]
 912 // CHECK3:       cond.end:
 913 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 914 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 915 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 916 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 917 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 918 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 919 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 920 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
 921 // CHECK3:       omp.dispatch.body:
 922 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 923 // CHECK3:       omp.inner.for.cond:
 924 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9:![0-9]+]]
 925 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP9]]
 926 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 927 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 928 // CHECK3:       omp.inner.for.body:
 929 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 930 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
 931 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 932 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 933 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 934 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP9]]
 935 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP11]]
 936 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP9]]
 937 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 938 // CHECK3:       omp.body.continue:
 939 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 940 // CHECK3:       omp.inner.for.inc:
 941 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 942 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
 943 // CHECK3-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP9]]
 944 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
 945 // CHECK3:       omp.inner.for.end:
 946 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
 947 // CHECK3:       omp.dispatch.inc:
 948 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 949 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 950 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
 951 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
 952 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 953 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 954 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
 955 // CHECK3-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
 956 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
 957 // CHECK3:       omp.dispatch.end:
 958 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 959 // CHECK3-NEXT:    ret void
 960 //
 961 //
 962 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
 963 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
 964 // CHECK3-NEXT:  entry:
 965 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
 966 // CHECK3-NEXT:    ret void
 967 //
 968 //
 969 // CHECK9-LABEL: define {{[^@]+}}@main
 970 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 971 // CHECK9-NEXT:  entry:
 972 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 973 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 974 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
 975 // CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
 976 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
 977 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
 978 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 979 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
 980 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
 981 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
 982 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
 983 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 984 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 985 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 986 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 987 // CHECK9-NEXT:    [[N_CASTED3:%.*]] = alloca i64, align 8
 988 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x ptr], align 8
 989 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x ptr], align 8
 990 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x ptr], align 8
 991 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 8
 992 // CHECK9-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
 993 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
 994 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
 995 // CHECK9-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 996 // CHECK9-NEXT:    [[N_CASTED18:%.*]] = alloca i64, align 8
 997 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [3 x ptr], align 8
 998 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [3 x ptr], align 8
 999 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [3 x ptr], align 8
1000 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES22:%.*]] = alloca [3 x i64], align 8
1001 // CHECK9-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
1002 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
1003 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
1004 // CHECK9-NEXT:    [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1005 // CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
1006 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1007 // CHECK9-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
1008 // CHECK9-NEXT:    store i32 100, ptr [[N]], align 4
1009 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
1010 // CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1011 // CHECK9-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
1012 // CHECK9-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
1013 // CHECK9-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1014 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
1015 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
1016 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
1017 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
1018 // CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1019 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 24, i1 false)
1020 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1021 // CHECK9-NEXT:    store i64 [[TMP4]], ptr [[TMP6]], align 8
1022 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1023 // CHECK9-NEXT:    store i64 [[TMP4]], ptr [[TMP7]], align 8
1024 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1025 // CHECK9-NEXT:    store ptr null, ptr [[TMP8]], align 8
1026 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1027 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP9]], align 8
1028 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1029 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP10]], align 8
1030 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1031 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
1032 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1033 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 8
1034 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1035 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 8
1036 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1037 // CHECK9-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 8
1038 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1039 // CHECK9-NEXT:    store ptr null, ptr [[TMP15]], align 8
1040 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1041 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1042 // CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1043 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
1044 // CHECK9-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
1045 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1046 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
1047 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1048 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1049 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1050 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1051 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
1052 // CHECK9-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
1053 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1054 // CHECK9-NEXT:    store i32 2, ptr [[TMP23]], align 4
1055 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1056 // CHECK9-NEXT:    store i32 3, ptr [[TMP24]], align 4
1057 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1058 // CHECK9-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 8
1059 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1060 // CHECK9-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 8
1061 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1062 // CHECK9-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 8
1063 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1064 // CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 8
1065 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1066 // CHECK9-NEXT:    store ptr null, ptr [[TMP29]], align 8
1067 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1068 // CHECK9-NEXT:    store ptr null, ptr [[TMP30]], align 8
1069 // CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1070 // CHECK9-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
1071 // CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1072 // CHECK9-NEXT:    store i64 0, ptr [[TMP32]], align 8
1073 // CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1074 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
1075 // CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1076 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
1077 // CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1078 // CHECK9-NEXT:    store i32 0, ptr [[TMP35]], align 4
1079 // CHECK9-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100.region_id, ptr [[KERNEL_ARGS]])
1080 // CHECK9-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
1081 // CHECK9-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1082 // CHECK9:       omp_offload.failed:
1083 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100(i64 [[TMP4]], i64 [[TMP1]], ptr [[VLA]]) #[[ATTR4:[0-9]+]]
1084 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1085 // CHECK9:       omp_offload.cont:
1086 // CHECK9-NEXT:    [[TMP38:%.*]] = load i32, ptr [[N]], align 4
1087 // CHECK9-NEXT:    store i32 [[TMP38]], ptr [[N_CASTED3]], align 4
1088 // CHECK9-NEXT:    [[TMP39:%.*]] = load i64, ptr [[N_CASTED3]], align 8
1089 // CHECK9-NEXT:    [[TMP40:%.*]] = mul nuw i64 [[TMP1]], 4
1090 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES7]], ptr align 8 @.offload_sizes.1, i64 24, i1 false)
1091 // CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
1092 // CHECK9-NEXT:    store i64 [[TMP39]], ptr [[TMP41]], align 8
1093 // CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
1094 // CHECK9-NEXT:    store i64 [[TMP39]], ptr [[TMP42]], align 8
1095 // CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 0
1096 // CHECK9-NEXT:    store ptr null, ptr [[TMP43]], align 8
1097 // CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
1098 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP44]], align 8
1099 // CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
1100 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP45]], align 8
1101 // CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 1
1102 // CHECK9-NEXT:    store ptr null, ptr [[TMP46]], align 8
1103 // CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
1104 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP47]], align 8
1105 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
1106 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP48]], align 8
1107 // CHECK9-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
1108 // CHECK9-NEXT:    store i64 [[TMP40]], ptr [[TMP49]], align 8
1109 // CHECK9-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i64 0, i64 2
1110 // CHECK9-NEXT:    store ptr null, ptr [[TMP50]], align 8
1111 // CHECK9-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
1112 // CHECK9-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
1113 // CHECK9-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
1114 // CHECK9-NEXT:    [[TMP54:%.*]] = load i32, ptr [[N]], align 4
1115 // CHECK9-NEXT:    store i32 [[TMP54]], ptr [[DOTCAPTURE_EXPR_9]], align 4
1116 // CHECK9-NEXT:    [[TMP55:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_9]], align 4
1117 // CHECK9-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP55]], 0
1118 // CHECK9-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
1119 // CHECK9-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
1120 // CHECK9-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_10]], align 4
1121 // CHECK9-NEXT:    [[TMP56:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_10]], align 4
1122 // CHECK9-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP56]], 1
1123 // CHECK9-NEXT:    [[TMP57:%.*]] = zext i32 [[ADD14]] to i64
1124 // CHECK9-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
1125 // CHECK9-NEXT:    store i32 2, ptr [[TMP58]], align 4
1126 // CHECK9-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
1127 // CHECK9-NEXT:    store i32 3, ptr [[TMP59]], align 4
1128 // CHECK9-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
1129 // CHECK9-NEXT:    store ptr [[TMP51]], ptr [[TMP60]], align 8
1130 // CHECK9-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
1131 // CHECK9-NEXT:    store ptr [[TMP52]], ptr [[TMP61]], align 8
1132 // CHECK9-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
1133 // CHECK9-NEXT:    store ptr [[TMP53]], ptr [[TMP62]], align 8
1134 // CHECK9-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
1135 // CHECK9-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP63]], align 8
1136 // CHECK9-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
1137 // CHECK9-NEXT:    store ptr null, ptr [[TMP64]], align 8
1138 // CHECK9-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
1139 // CHECK9-NEXT:    store ptr null, ptr [[TMP65]], align 8
1140 // CHECK9-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
1141 // CHECK9-NEXT:    store i64 [[TMP57]], ptr [[TMP66]], align 8
1142 // CHECK9-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
1143 // CHECK9-NEXT:    store i64 0, ptr [[TMP67]], align 8
1144 // CHECK9-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
1145 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP68]], align 4
1146 // CHECK9-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
1147 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
1148 // CHECK9-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
1149 // CHECK9-NEXT:    store i32 0, ptr [[TMP70]], align 4
1150 // CHECK9-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.region_id, ptr [[KERNEL_ARGS15]])
1151 // CHECK9-NEXT:    [[TMP72:%.*]] = icmp ne i32 [[TMP71]], 0
1152 // CHECK9-NEXT:    br i1 [[TMP72]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
1153 // CHECK9:       omp_offload.failed16:
1154 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105(i64 [[TMP39]], i64 [[TMP1]], ptr [[VLA]]) #[[ATTR4]]
1155 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
1156 // CHECK9:       omp_offload.cont17:
1157 // CHECK9-NEXT:    [[TMP73:%.*]] = load i32, ptr [[N]], align 4
1158 // CHECK9-NEXT:    store i32 [[TMP73]], ptr [[N_CASTED18]], align 4
1159 // CHECK9-NEXT:    [[TMP74:%.*]] = load i64, ptr [[N_CASTED18]], align 8
1160 // CHECK9-NEXT:    [[TMP75:%.*]] = mul nuw i64 [[TMP1]], 4
1161 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES22]], ptr align 8 @.offload_sizes.3, i64 24, i1 false)
1162 // CHECK9-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
1163 // CHECK9-NEXT:    store i64 [[TMP74]], ptr [[TMP76]], align 8
1164 // CHECK9-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
1165 // CHECK9-NEXT:    store i64 [[TMP74]], ptr [[TMP77]], align 8
1166 // CHECK9-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS21]], i64 0, i64 0
1167 // CHECK9-NEXT:    store ptr null, ptr [[TMP78]], align 8
1168 // CHECK9-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
1169 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP79]], align 8
1170 // CHECK9-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
1171 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP80]], align 8
1172 // CHECK9-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS21]], i64 0, i64 1
1173 // CHECK9-NEXT:    store ptr null, ptr [[TMP81]], align 8
1174 // CHECK9-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
1175 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP82]], align 8
1176 // CHECK9-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
1177 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP83]], align 8
1178 // CHECK9-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES22]], i32 0, i32 2
1179 // CHECK9-NEXT:    store i64 [[TMP75]], ptr [[TMP84]], align 8
1180 // CHECK9-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS21]], i64 0, i64 2
1181 // CHECK9-NEXT:    store ptr null, ptr [[TMP85]], align 8
1182 // CHECK9-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
1183 // CHECK9-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
1184 // CHECK9-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
1185 // CHECK9-NEXT:    [[TMP89:%.*]] = load i32, ptr [[N]], align 4
1186 // CHECK9-NEXT:    store i32 [[TMP89]], ptr [[DOTCAPTURE_EXPR_24]], align 4
1187 // CHECK9-NEXT:    [[TMP90:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_24]], align 4
1188 // CHECK9-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[TMP90]], 0
1189 // CHECK9-NEXT:    [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
1190 // CHECK9-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
1191 // CHECK9-NEXT:    store i32 [[SUB28]], ptr [[DOTCAPTURE_EXPR_25]], align 4
1192 // CHECK9-NEXT:    [[TMP91:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_25]], align 4
1193 // CHECK9-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP91]], 1
1194 // CHECK9-NEXT:    [[TMP92:%.*]] = zext i32 [[ADD29]] to i64
1195 // CHECK9-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 0
1196 // CHECK9-NEXT:    store i32 2, ptr [[TMP93]], align 4
1197 // CHECK9-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 1
1198 // CHECK9-NEXT:    store i32 3, ptr [[TMP94]], align 4
1199 // CHECK9-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 2
1200 // CHECK9-NEXT:    store ptr [[TMP86]], ptr [[TMP95]], align 8
1201 // CHECK9-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 3
1202 // CHECK9-NEXT:    store ptr [[TMP87]], ptr [[TMP96]], align 8
1203 // CHECK9-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 4
1204 // CHECK9-NEXT:    store ptr [[TMP88]], ptr [[TMP97]], align 8
1205 // CHECK9-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 5
1206 // CHECK9-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP98]], align 8
1207 // CHECK9-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 6
1208 // CHECK9-NEXT:    store ptr null, ptr [[TMP99]], align 8
1209 // CHECK9-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 7
1210 // CHECK9-NEXT:    store ptr null, ptr [[TMP100]], align 8
1211 // CHECK9-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 8
1212 // CHECK9-NEXT:    store i64 [[TMP92]], ptr [[TMP101]], align 8
1213 // CHECK9-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 9
1214 // CHECK9-NEXT:    store i64 0, ptr [[TMP102]], align 8
1215 // CHECK9-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 10
1216 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP103]], align 4
1217 // CHECK9-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 11
1218 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP104]], align 4
1219 // CHECK9-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 12
1220 // CHECK9-NEXT:    store i32 0, ptr [[TMP105]], align 4
1221 // CHECK9-NEXT:    [[TMP106:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.region_id, ptr [[KERNEL_ARGS30]])
1222 // CHECK9-NEXT:    [[TMP107:%.*]] = icmp ne i32 [[TMP106]], 0
1223 // CHECK9-NEXT:    br i1 [[TMP107]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
1224 // CHECK9:       omp_offload.failed31:
1225 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110(i64 [[TMP74]], i64 [[TMP1]], ptr [[VLA]]) #[[ATTR4]]
1226 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT32]]
1227 // CHECK9:       omp_offload.cont32:
1228 // CHECK9-NEXT:    [[TMP108:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
1229 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP108]])
1230 // CHECK9-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
1231 // CHECK9-NEXT:    [[TMP109:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
1232 // CHECK9-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP109]])
1233 // CHECK9-NEXT:    [[TMP110:%.*]] = load i32, ptr [[RETVAL]], align 4
1234 // CHECK9-NEXT:    ret i32 [[TMP110]]
1235 //
1236 //
1237 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100
1238 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1239 // CHECK9-NEXT:  entry:
1240 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1241 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1242 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1243 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1244 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1245 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1246 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1247 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1248 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
1249 // CHECK9-NEXT:    ret void
1250 //
1251 //
1252 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100.omp_outlined
1253 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3:[0-9]+]] {
1254 // CHECK9-NEXT:  entry:
1255 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1256 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1257 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
1258 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1259 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1260 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1261 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1262 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1263 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1264 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1265 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1266 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1267 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1268 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1269 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
1270 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1271 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1272 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
1273 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1274 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1275 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
1276 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1277 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1278 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
1279 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1280 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1281 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1282 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1283 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1284 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1285 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1286 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1287 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1288 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1289 // CHECK9:       omp.precond.then:
1290 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1291 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1292 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
1293 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1294 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1295 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1296 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1297 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1298 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1299 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1300 // CHECK9-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1301 // CHECK9-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1302 // CHECK9:       cond.true:
1303 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1304 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1305 // CHECK9:       cond.false:
1306 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1307 // CHECK9-NEXT:    br label [[COND_END]]
1308 // CHECK9:       cond.end:
1309 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1310 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1311 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1312 // CHECK9-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
1313 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1314 // CHECK9:       omp.inner.for.cond:
1315 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1316 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1317 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1318 // CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1319 // CHECK9:       omp.inner.for.body:
1320 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1321 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1322 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1323 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
1324 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
1325 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1326 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[IDXPROM]]
1327 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1328 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1329 // CHECK9:       omp.body.continue:
1330 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1331 // CHECK9:       omp.inner.for.inc:
1332 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1333 // CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1334 // CHECK9-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
1335 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1336 // CHECK9:       omp.inner.for.end:
1337 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1338 // CHECK9:       omp.loop.exit:
1339 // CHECK9-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1340 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
1341 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
1342 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1343 // CHECK9:       omp.precond.end:
1344 // CHECK9-NEXT:    ret void
1345 //
1346 //
1347 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105
1348 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1349 // CHECK9-NEXT:  entry:
1350 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1351 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1352 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1353 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1354 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1355 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1356 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1357 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1358 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
1359 // CHECK9-NEXT:    ret void
1360 //
1361 //
1362 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.omp_outlined
1363 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3]] {
1364 // CHECK9-NEXT:  entry:
1365 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1366 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1367 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
1368 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1369 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1370 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1371 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1372 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1373 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1374 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1375 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1376 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1377 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1378 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1379 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
1380 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1381 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1382 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
1383 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1384 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1385 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
1386 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1387 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1388 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
1389 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1390 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1391 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1392 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1393 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1394 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1395 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1396 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1397 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1398 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1399 // CHECK9:       omp.precond.then:
1400 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1401 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1402 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
1403 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1404 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1405 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1406 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1407 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1408 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1409 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1410 // CHECK9-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1411 // CHECK9-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1412 // CHECK9:       cond.true:
1413 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1414 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1415 // CHECK9:       cond.false:
1416 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1417 // CHECK9-NEXT:    br label [[COND_END]]
1418 // CHECK9:       cond.end:
1419 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1420 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1421 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1422 // CHECK9-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
1423 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1424 // CHECK9:       omp.inner.for.cond:
1425 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1426 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1427 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1428 // CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1429 // CHECK9:       omp.inner.for.body:
1430 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1431 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1432 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1433 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
1434 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
1435 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1436 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[IDXPROM]]
1437 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1438 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1439 // CHECK9:       omp.body.continue:
1440 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1441 // CHECK9:       omp.inner.for.inc:
1442 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1443 // CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1444 // CHECK9-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
1445 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1446 // CHECK9:       omp.inner.for.end:
1447 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1448 // CHECK9:       omp.loop.exit:
1449 // CHECK9-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1450 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
1451 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
1452 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1453 // CHECK9:       omp.precond.end:
1454 // CHECK9-NEXT:    ret void
1455 //
1456 //
1457 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110
1458 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1459 // CHECK9-NEXT:  entry:
1460 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1461 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1462 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1463 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1464 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1465 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1466 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1467 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1468 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1469 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1470 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
1471 // CHECK9-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
1472 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1473 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
1474 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR__CASTED]], align 8
1475 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]], i64 [[TMP4]])
1476 // CHECK9-NEXT:    ret void
1477 //
1478 //
1479 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.omp_outlined
1480 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
1481 // CHECK9-NEXT:  entry:
1482 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1483 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1484 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
1485 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1486 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1487 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1488 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1489 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1490 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1491 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1492 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1493 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1494 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1495 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1496 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1497 // CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
1498 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1499 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1500 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
1501 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1502 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1503 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 8
1504 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
1505 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1506 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1507 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
1508 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1509 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1510 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1511 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1512 // CHECK9-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
1513 // CHECK9-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
1514 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1515 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1516 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1517 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1518 // CHECK9:       omp.precond.then:
1519 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1520 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1521 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
1522 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1523 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1524 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
1525 // CHECK9-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1526 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
1527 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP9]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
1528 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1529 // CHECK9:       omp.dispatch.cond:
1530 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1531 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1532 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
1533 // CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1534 // CHECK9:       cond.true:
1535 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1536 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1537 // CHECK9:       cond.false:
1538 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1539 // CHECK9-NEXT:    br label [[COND_END]]
1540 // CHECK9:       cond.end:
1541 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
1542 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1543 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1544 // CHECK9-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
1545 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1546 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1547 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
1548 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1549 // CHECK9:       omp.dispatch.body:
1550 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1551 // CHECK9:       omp.inner.for.cond:
1552 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11:![0-9]+]]
1553 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP11]]
1554 // CHECK9-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
1555 // CHECK9-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1556 // CHECK9:       omp.inner.for.body:
1557 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
1558 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
1559 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1560 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP11]]
1561 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP11]]
1562 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
1563 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[IDXPROM]]
1564 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP11]]
1565 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1566 // CHECK9:       omp.body.continue:
1567 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1568 // CHECK9:       omp.inner.for.inc:
1569 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
1570 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP21]], 1
1571 // CHECK9-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP11]]
1572 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1573 // CHECK9:       omp.inner.for.end:
1574 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1575 // CHECK9:       omp.dispatch.inc:
1576 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1577 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1578 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
1579 // CHECK9-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_LB]], align 4
1580 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1581 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1582 // CHECK9-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
1583 // CHECK9-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_UB]], align 4
1584 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
1585 // CHECK9:       omp.dispatch.end:
1586 // CHECK9-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1587 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4
1588 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP27]])
1589 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1590 // CHECK9:       omp.precond.end:
1591 // CHECK9-NEXT:    ret void
1592 //
1593 //
1594 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
1595 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR6:[0-9]+]] comdat {
1596 // CHECK9-NEXT:  entry:
1597 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1598 // CHECK9-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
1599 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
1600 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
1601 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
1602 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1603 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1604 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x ptr], align 8
1605 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x ptr], align 8
1606 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x ptr], align 8
1607 // CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
1608 // CHECK9-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1609 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x ptr], align 8
1610 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x ptr], align 8
1611 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x ptr], align 8
1612 // CHECK9-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
1613 // CHECK9-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1614 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1615 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1616 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP0]], align 8
1617 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1618 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
1619 // CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1620 // CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
1621 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1622 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1623 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1624 // CHECK9-NEXT:    store i32 2, ptr [[TMP5]], align 4
1625 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1626 // CHECK9-NEXT:    store i32 1, ptr [[TMP6]], align 4
1627 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1628 // CHECK9-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
1629 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1630 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
1631 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1632 // CHECK9-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP9]], align 8
1633 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1634 // CHECK9-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP10]], align 8
1635 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1636 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
1637 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1638 // CHECK9-NEXT:    store ptr null, ptr [[TMP12]], align 8
1639 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1640 // CHECK9-NEXT:    store i64 10, ptr [[TMP13]], align 8
1641 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1642 // CHECK9-NEXT:    store i64 0, ptr [[TMP14]], align 8
1643 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1644 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
1645 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1646 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
1647 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1648 // CHECK9-NEXT:    store i32 0, ptr [[TMP17]], align 4
1649 // CHECK9-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.region_id, ptr [[KERNEL_ARGS]])
1650 // CHECK9-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1651 // CHECK9-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1652 // CHECK9:       omp_offload.failed:
1653 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79(ptr [[A]]) #[[ATTR4]]
1654 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1655 // CHECK9:       omp_offload.cont:
1656 // CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1657 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP20]], align 8
1658 // CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1659 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP21]], align 8
1660 // CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
1661 // CHECK9-NEXT:    store ptr null, ptr [[TMP22]], align 8
1662 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1663 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1664 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
1665 // CHECK9-NEXT:    store i32 2, ptr [[TMP25]], align 4
1666 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
1667 // CHECK9-NEXT:    store i32 1, ptr [[TMP26]], align 4
1668 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
1669 // CHECK9-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
1670 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
1671 // CHECK9-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
1672 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
1673 // CHECK9-NEXT:    store ptr @.offload_sizes.7, ptr [[TMP29]], align 8
1674 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
1675 // CHECK9-NEXT:    store ptr @.offload_maptypes.8, ptr [[TMP30]], align 8
1676 // CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
1677 // CHECK9-NEXT:    store ptr null, ptr [[TMP31]], align 8
1678 // CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
1679 // CHECK9-NEXT:    store ptr null, ptr [[TMP32]], align 8
1680 // CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
1681 // CHECK9-NEXT:    store i64 10, ptr [[TMP33]], align 8
1682 // CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
1683 // CHECK9-NEXT:    store i64 0, ptr [[TMP34]], align 8
1684 // CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
1685 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
1686 // CHECK9-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
1687 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
1688 // CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
1689 // CHECK9-NEXT:    store i32 0, ptr [[TMP37]], align 4
1690 // CHECK9-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, ptr [[KERNEL_ARGS5]])
1691 // CHECK9-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
1692 // CHECK9-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
1693 // CHECK9:       omp_offload.failed6:
1694 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84(ptr [[A]]) #[[ATTR4]]
1695 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
1696 // CHECK9:       omp_offload.cont7:
1697 // CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
1698 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP40]], align 8
1699 // CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
1700 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP41]], align 8
1701 // CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
1702 // CHECK9-NEXT:    store ptr null, ptr [[TMP42]], align 8
1703 // CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
1704 // CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
1705 // CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
1706 // CHECK9-NEXT:    store i32 2, ptr [[TMP45]], align 4
1707 // CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
1708 // CHECK9-NEXT:    store i32 1, ptr [[TMP46]], align 4
1709 // CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
1710 // CHECK9-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 8
1711 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
1712 // CHECK9-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 8
1713 // CHECK9-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
1714 // CHECK9-NEXT:    store ptr @.offload_sizes.9, ptr [[TMP49]], align 8
1715 // CHECK9-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
1716 // CHECK9-NEXT:    store ptr @.offload_maptypes.10, ptr [[TMP50]], align 8
1717 // CHECK9-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
1718 // CHECK9-NEXT:    store ptr null, ptr [[TMP51]], align 8
1719 // CHECK9-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
1720 // CHECK9-NEXT:    store ptr null, ptr [[TMP52]], align 8
1721 // CHECK9-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
1722 // CHECK9-NEXT:    store i64 10, ptr [[TMP53]], align 8
1723 // CHECK9-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
1724 // CHECK9-NEXT:    store i64 0, ptr [[TMP54]], align 8
1725 // CHECK9-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
1726 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
1727 // CHECK9-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
1728 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
1729 // CHECK9-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
1730 // CHECK9-NEXT:    store i32 0, ptr [[TMP57]], align 4
1731 // CHECK9-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.region_id, ptr [[KERNEL_ARGS12]])
1732 // CHECK9-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
1733 // CHECK9-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
1734 // CHECK9:       omp_offload.failed13:
1735 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89(ptr [[A]]) #[[ATTR4]]
1736 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
1737 // CHECK9:       omp_offload.cont14:
1738 // CHECK9-NEXT:    ret i32 0
1739 //
1740 //
1741 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79
1742 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1743 // CHECK9-NEXT:  entry:
1744 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1745 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1746 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1747 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.omp_outlined, ptr [[TMP0]])
1748 // CHECK9-NEXT:    ret void
1749 //
1750 //
1751 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.omp_outlined
1752 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
1753 // CHECK9-NEXT:  entry:
1754 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1755 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1756 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1757 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1758 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1759 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1760 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1761 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1762 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1763 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1764 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1765 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1766 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1767 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1768 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1769 // CHECK9-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
1770 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1771 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1772 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1773 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1774 // 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)
1775 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1776 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1777 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1778 // CHECK9:       cond.true:
1779 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1780 // CHECK9:       cond.false:
1781 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1782 // CHECK9-NEXT:    br label [[COND_END]]
1783 // CHECK9:       cond.end:
1784 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1785 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1786 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1787 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1788 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1789 // CHECK9:       omp.inner.for.cond:
1790 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1791 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1792 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1793 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1794 // CHECK9:       omp.inner.for.body:
1795 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1796 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1797 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1798 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1799 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
1800 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
1801 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1802 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1803 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1804 // CHECK9:       omp.body.continue:
1805 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1806 // CHECK9:       omp.inner.for.inc:
1807 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1808 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1809 // CHECK9-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
1810 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1811 // CHECK9:       omp.inner.for.end:
1812 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1813 // CHECK9:       omp.loop.exit:
1814 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1815 // CHECK9-NEXT:    ret void
1816 //
1817 //
1818 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
1819 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1820 // CHECK9-NEXT:  entry:
1821 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1822 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1823 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1824 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined, ptr [[TMP0]])
1825 // CHECK9-NEXT:    ret void
1826 //
1827 //
1828 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined
1829 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
1830 // CHECK9-NEXT:  entry:
1831 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1832 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1833 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1834 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1835 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1836 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1837 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1838 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1839 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1840 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1841 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1842 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1843 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1844 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1845 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1846 // CHECK9-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
1847 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1848 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1849 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1850 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1851 // 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)
1852 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1853 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1854 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1855 // CHECK9:       cond.true:
1856 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1857 // CHECK9:       cond.false:
1858 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1859 // CHECK9-NEXT:    br label [[COND_END]]
1860 // CHECK9:       cond.end:
1861 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1862 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1863 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1864 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1865 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1866 // CHECK9:       omp.inner.for.cond:
1867 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1868 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1869 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1870 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1871 // CHECK9:       omp.inner.for.body:
1872 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1873 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1874 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1875 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1876 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
1877 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
1878 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1879 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1880 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1881 // CHECK9:       omp.body.continue:
1882 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1883 // CHECK9:       omp.inner.for.inc:
1884 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1885 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1886 // CHECK9-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
1887 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1888 // CHECK9:       omp.inner.for.end:
1889 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1890 // CHECK9:       omp.loop.exit:
1891 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1892 // CHECK9-NEXT:    ret void
1893 //
1894 //
1895 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89
1896 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1897 // CHECK9-NEXT:  entry:
1898 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1899 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1900 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1901 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.omp_outlined, ptr [[TMP0]])
1902 // CHECK9-NEXT:    ret void
1903 //
1904 //
1905 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.omp_outlined
1906 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
1907 // CHECK9-NEXT:  entry:
1908 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1909 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1910 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1911 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1912 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1913 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1914 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1915 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1916 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1917 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1918 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1919 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1920 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1921 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1922 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1923 // CHECK9-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
1924 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1925 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1926 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1927 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1928 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 10)
1929 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1930 // CHECK9:       omp.dispatch.cond:
1931 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1932 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1933 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1934 // CHECK9:       cond.true:
1935 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1936 // CHECK9:       cond.false:
1937 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1938 // CHECK9-NEXT:    br label [[COND_END]]
1939 // CHECK9:       cond.end:
1940 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1941 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1942 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1943 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1944 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1945 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1946 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1947 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1948 // CHECK9:       omp.dispatch.body:
1949 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1950 // CHECK9:       omp.inner.for.cond:
1951 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14:![0-9]+]]
1952 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP14]]
1953 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1954 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1955 // CHECK9:       omp.inner.for.body:
1956 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
1957 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1958 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1959 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP14]]
1960 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP14]]
1961 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1962 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1963 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP14]]
1964 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1965 // CHECK9:       omp.body.continue:
1966 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1967 // CHECK9:       omp.inner.for.inc:
1968 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
1969 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
1970 // CHECK9-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP14]]
1971 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
1972 // CHECK9:       omp.inner.for.end:
1973 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1974 // CHECK9:       omp.dispatch.inc:
1975 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1976 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1977 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1978 // CHECK9-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
1979 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1980 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1981 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
1982 // CHECK9-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
1983 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
1984 // CHECK9:       omp.dispatch.end:
1985 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1986 // CHECK9-NEXT:    ret void
1987 //
1988 //
1989 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1990 // CHECK9-SAME: () #[[ATTR7:[0-9]+]] {
1991 // CHECK9-NEXT:  entry:
1992 // CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
1993 // CHECK9-NEXT:    ret void
1994 //
1995 //
1996 // CHECK11-LABEL: define {{[^@]+}}@main
1997 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1998 // CHECK11-NEXT:  entry:
1999 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
2000 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
2001 // CHECK11-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 4
2002 // CHECK11-NEXT:    [[N:%.*]] = alloca i32, align 4
2003 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
2004 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
2005 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
2006 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
2007 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
2008 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
2009 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
2010 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2011 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2012 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2013 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2014 // CHECK11-NEXT:    [[N_CASTED3:%.*]] = alloca i32, align 4
2015 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x ptr], align 4
2016 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x ptr], align 4
2017 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x ptr], align 4
2018 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
2019 // CHECK11-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
2020 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
2021 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
2022 // CHECK11-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2023 // CHECK11-NEXT:    [[N_CASTED18:%.*]] = alloca i32, align 4
2024 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [3 x ptr], align 4
2025 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS20:%.*]] = alloca [3 x ptr], align 4
2026 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [3 x ptr], align 4
2027 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES22:%.*]] = alloca [3 x i64], align 4
2028 // CHECK11-NEXT:    [[_TMP23:%.*]] = alloca i32, align 4
2029 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
2030 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
2031 // CHECK11-NEXT:    [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2032 // CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
2033 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
2034 // CHECK11-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 4
2035 // CHECK11-NEXT:    store i32 100, ptr [[N]], align 4
2036 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
2037 // CHECK11-NEXT:    [[TMP1:%.*]] = call ptr @llvm.stacksave.p0()
2038 // CHECK11-NEXT:    store ptr [[TMP1]], ptr [[SAVED_STACK]], align 4
2039 // CHECK11-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
2040 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[__VLA_EXPR0]], align 4
2041 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N]], align 4
2042 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
2043 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
2044 // CHECK11-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
2045 // CHECK11-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
2046 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 24, i1 false)
2047 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2048 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP6]], align 4
2049 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2050 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP7]], align 4
2051 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2052 // CHECK11-NEXT:    store ptr null, ptr [[TMP8]], align 4
2053 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2054 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP9]], align 4
2055 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2056 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP10]], align 4
2057 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2058 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
2059 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2060 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 4
2061 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2062 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 4
2063 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2064 // CHECK11-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 4
2065 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2066 // CHECK11-NEXT:    store ptr null, ptr [[TMP15]], align 4
2067 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2068 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2069 // CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2070 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
2071 // CHECK11-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
2072 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2073 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
2074 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2075 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2076 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2077 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2078 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
2079 // CHECK11-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
2080 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2081 // CHECK11-NEXT:    store i32 2, ptr [[TMP23]], align 4
2082 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2083 // CHECK11-NEXT:    store i32 3, ptr [[TMP24]], align 4
2084 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2085 // CHECK11-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 4
2086 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2087 // CHECK11-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 4
2088 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2089 // CHECK11-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 4
2090 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2091 // CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 4
2092 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2093 // CHECK11-NEXT:    store ptr null, ptr [[TMP29]], align 4
2094 // CHECK11-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2095 // CHECK11-NEXT:    store ptr null, ptr [[TMP30]], align 4
2096 // CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2097 // CHECK11-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
2098 // CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2099 // CHECK11-NEXT:    store i64 0, ptr [[TMP32]], align 8
2100 // CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2101 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
2102 // CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2103 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
2104 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2105 // CHECK11-NEXT:    store i32 0, ptr [[TMP35]], align 4
2106 // CHECK11-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100.region_id, ptr [[KERNEL_ARGS]])
2107 // CHECK11-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2108 // CHECK11-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2109 // CHECK11:       omp_offload.failed:
2110 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100(i32 [[TMP3]], i32 [[TMP0]], ptr [[VLA]]) #[[ATTR4:[0-9]+]]
2111 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2112 // CHECK11:       omp_offload.cont:
2113 // CHECK11-NEXT:    [[TMP38:%.*]] = load i32, ptr [[N]], align 4
2114 // CHECK11-NEXT:    store i32 [[TMP38]], ptr [[N_CASTED3]], align 4
2115 // CHECK11-NEXT:    [[TMP39:%.*]] = load i32, ptr [[N_CASTED3]], align 4
2116 // CHECK11-NEXT:    [[TMP40:%.*]] = mul nuw i32 [[TMP0]], 4
2117 // CHECK11-NEXT:    [[TMP41:%.*]] = sext i32 [[TMP40]] to i64
2118 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES7]], ptr align 4 @.offload_sizes.1, i32 24, i1 false)
2119 // CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
2120 // CHECK11-NEXT:    store i32 [[TMP39]], ptr [[TMP42]], align 4
2121 // CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
2122 // CHECK11-NEXT:    store i32 [[TMP39]], ptr [[TMP43]], align 4
2123 // CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
2124 // CHECK11-NEXT:    store ptr null, ptr [[TMP44]], align 4
2125 // CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
2126 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP45]], align 4
2127 // CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
2128 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP46]], align 4
2129 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
2130 // CHECK11-NEXT:    store ptr null, ptr [[TMP47]], align 4
2131 // CHECK11-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
2132 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP48]], align 4
2133 // CHECK11-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
2134 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP49]], align 4
2135 // CHECK11-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
2136 // CHECK11-NEXT:    store i64 [[TMP41]], ptr [[TMP50]], align 4
2137 // CHECK11-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
2138 // CHECK11-NEXT:    store ptr null, ptr [[TMP51]], align 4
2139 // CHECK11-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
2140 // CHECK11-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
2141 // CHECK11-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
2142 // CHECK11-NEXT:    [[TMP55:%.*]] = load i32, ptr [[N]], align 4
2143 // CHECK11-NEXT:    store i32 [[TMP55]], ptr [[DOTCAPTURE_EXPR_9]], align 4
2144 // CHECK11-NEXT:    [[TMP56:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_9]], align 4
2145 // CHECK11-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP56]], 0
2146 // CHECK11-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
2147 // CHECK11-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
2148 // CHECK11-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_10]], align 4
2149 // CHECK11-NEXT:    [[TMP57:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_10]], align 4
2150 // CHECK11-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP57]], 1
2151 // CHECK11-NEXT:    [[TMP58:%.*]] = zext i32 [[ADD14]] to i64
2152 // CHECK11-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
2153 // CHECK11-NEXT:    store i32 2, ptr [[TMP59]], align 4
2154 // CHECK11-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
2155 // CHECK11-NEXT:    store i32 3, ptr [[TMP60]], align 4
2156 // CHECK11-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
2157 // CHECK11-NEXT:    store ptr [[TMP52]], ptr [[TMP61]], align 4
2158 // CHECK11-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
2159 // CHECK11-NEXT:    store ptr [[TMP53]], ptr [[TMP62]], align 4
2160 // CHECK11-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
2161 // CHECK11-NEXT:    store ptr [[TMP54]], ptr [[TMP63]], align 4
2162 // CHECK11-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
2163 // CHECK11-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP64]], align 4
2164 // CHECK11-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
2165 // CHECK11-NEXT:    store ptr null, ptr [[TMP65]], align 4
2166 // CHECK11-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
2167 // CHECK11-NEXT:    store ptr null, ptr [[TMP66]], align 4
2168 // CHECK11-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
2169 // CHECK11-NEXT:    store i64 [[TMP58]], ptr [[TMP67]], align 8
2170 // CHECK11-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
2171 // CHECK11-NEXT:    store i64 0, ptr [[TMP68]], align 8
2172 // CHECK11-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
2173 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
2174 // CHECK11-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
2175 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP70]], align 4
2176 // CHECK11-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
2177 // CHECK11-NEXT:    store i32 0, ptr [[TMP71]], align 4
2178 // CHECK11-NEXT:    [[TMP72:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.region_id, ptr [[KERNEL_ARGS15]])
2179 // CHECK11-NEXT:    [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0
2180 // CHECK11-NEXT:    br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
2181 // CHECK11:       omp_offload.failed16:
2182 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105(i32 [[TMP39]], i32 [[TMP0]], ptr [[VLA]]) #[[ATTR4]]
2183 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
2184 // CHECK11:       omp_offload.cont17:
2185 // CHECK11-NEXT:    [[TMP74:%.*]] = load i32, ptr [[N]], align 4
2186 // CHECK11-NEXT:    store i32 [[TMP74]], ptr [[N_CASTED18]], align 4
2187 // CHECK11-NEXT:    [[TMP75:%.*]] = load i32, ptr [[N_CASTED18]], align 4
2188 // CHECK11-NEXT:    [[TMP76:%.*]] = mul nuw i32 [[TMP0]], 4
2189 // CHECK11-NEXT:    [[TMP77:%.*]] = sext i32 [[TMP76]] to i64
2190 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES22]], ptr align 4 @.offload_sizes.3, i32 24, i1 false)
2191 // CHECK11-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
2192 // CHECK11-NEXT:    store i32 [[TMP75]], ptr [[TMP78]], align 4
2193 // CHECK11-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
2194 // CHECK11-NEXT:    store i32 [[TMP75]], ptr [[TMP79]], align 4
2195 // CHECK11-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
2196 // CHECK11-NEXT:    store ptr null, ptr [[TMP80]], align 4
2197 // CHECK11-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
2198 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP81]], align 4
2199 // CHECK11-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
2200 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP82]], align 4
2201 // CHECK11-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 1
2202 // CHECK11-NEXT:    store ptr null, ptr [[TMP83]], align 4
2203 // CHECK11-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
2204 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP84]], align 4
2205 // CHECK11-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
2206 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP85]], align 4
2207 // CHECK11-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES22]], i32 0, i32 2
2208 // CHECK11-NEXT:    store i64 [[TMP77]], ptr [[TMP86]], align 4
2209 // CHECK11-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 2
2210 // CHECK11-NEXT:    store ptr null, ptr [[TMP87]], align 4
2211 // CHECK11-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
2212 // CHECK11-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
2213 // CHECK11-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
2214 // CHECK11-NEXT:    [[TMP91:%.*]] = load i32, ptr [[N]], align 4
2215 // CHECK11-NEXT:    store i32 [[TMP91]], ptr [[DOTCAPTURE_EXPR_24]], align 4
2216 // CHECK11-NEXT:    [[TMP92:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_24]], align 4
2217 // CHECK11-NEXT:    [[SUB26:%.*]] = sub nsw i32 [[TMP92]], 0
2218 // CHECK11-NEXT:    [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
2219 // CHECK11-NEXT:    [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
2220 // CHECK11-NEXT:    store i32 [[SUB28]], ptr [[DOTCAPTURE_EXPR_25]], align 4
2221 // CHECK11-NEXT:    [[TMP93:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_25]], align 4
2222 // CHECK11-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP93]], 1
2223 // CHECK11-NEXT:    [[TMP94:%.*]] = zext i32 [[ADD29]] to i64
2224 // CHECK11-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 0
2225 // CHECK11-NEXT:    store i32 2, ptr [[TMP95]], align 4
2226 // CHECK11-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 1
2227 // CHECK11-NEXT:    store i32 3, ptr [[TMP96]], align 4
2228 // CHECK11-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 2
2229 // CHECK11-NEXT:    store ptr [[TMP88]], ptr [[TMP97]], align 4
2230 // CHECK11-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 3
2231 // CHECK11-NEXT:    store ptr [[TMP89]], ptr [[TMP98]], align 4
2232 // CHECK11-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 4
2233 // CHECK11-NEXT:    store ptr [[TMP90]], ptr [[TMP99]], align 4
2234 // CHECK11-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 5
2235 // CHECK11-NEXT:    store ptr @.offload_maptypes.4, ptr [[TMP100]], align 4
2236 // CHECK11-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 6
2237 // CHECK11-NEXT:    store ptr null, ptr [[TMP101]], align 4
2238 // CHECK11-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 7
2239 // CHECK11-NEXT:    store ptr null, ptr [[TMP102]], align 4
2240 // CHECK11-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 8
2241 // CHECK11-NEXT:    store i64 [[TMP94]], ptr [[TMP103]], align 8
2242 // CHECK11-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 9
2243 // CHECK11-NEXT:    store i64 0, ptr [[TMP104]], align 8
2244 // CHECK11-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 10
2245 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP105]], align 4
2246 // CHECK11-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 11
2247 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP106]], align 4
2248 // CHECK11-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS30]], i32 0, i32 12
2249 // CHECK11-NEXT:    store i32 0, ptr [[TMP107]], align 4
2250 // CHECK11-NEXT:    [[TMP108:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.region_id, ptr [[KERNEL_ARGS30]])
2251 // CHECK11-NEXT:    [[TMP109:%.*]] = icmp ne i32 [[TMP108]], 0
2252 // CHECK11-NEXT:    br i1 [[TMP109]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
2253 // CHECK11:       omp_offload.failed31:
2254 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110(i32 [[TMP75]], i32 [[TMP0]], ptr [[VLA]]) #[[ATTR4]]
2255 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT32]]
2256 // CHECK11:       omp_offload.cont32:
2257 // CHECK11-NEXT:    [[TMP110:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
2258 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP110]])
2259 // CHECK11-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
2260 // CHECK11-NEXT:    [[TMP111:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
2261 // CHECK11-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP111]])
2262 // CHECK11-NEXT:    [[TMP112:%.*]] = load i32, ptr [[RETVAL]], align 4
2263 // CHECK11-NEXT:    ret i32 [[TMP112]]
2264 //
2265 //
2266 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100
2267 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
2268 // CHECK11-NEXT:  entry:
2269 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2270 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2271 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2272 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2273 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2274 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2275 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2276 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2277 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
2278 // CHECK11-NEXT:    ret void
2279 //
2280 //
2281 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100.omp_outlined
2282 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3:[0-9]+]] {
2283 // CHECK11-NEXT:  entry:
2284 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2285 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2286 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
2287 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2288 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2289 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2290 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2291 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2292 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2293 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2294 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2295 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2296 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2297 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2298 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
2299 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2300 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2301 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
2302 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2303 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2304 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
2305 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2306 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2307 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
2308 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
2309 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2310 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2311 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2312 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2313 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2314 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
2315 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2316 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2317 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2318 // CHECK11:       omp.precond.then:
2319 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2320 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2321 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
2322 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2323 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2324 // CHECK11-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2325 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2326 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2327 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2328 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2329 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2330 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2331 // CHECK11:       cond.true:
2332 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2333 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2334 // CHECK11:       cond.false:
2335 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2336 // CHECK11-NEXT:    br label [[COND_END]]
2337 // CHECK11:       cond.end:
2338 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2339 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2340 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2341 // CHECK11-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
2342 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2343 // CHECK11:       omp.inner.for.cond:
2344 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2345 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2346 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2347 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2348 // CHECK11:       omp.inner.for.body:
2349 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2350 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
2351 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2352 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
2353 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
2354 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP17]]
2355 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2356 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2357 // CHECK11:       omp.body.continue:
2358 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2359 // CHECK11:       omp.inner.for.inc:
2360 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2361 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
2362 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
2363 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2364 // CHECK11:       omp.inner.for.end:
2365 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2366 // CHECK11:       omp.loop.exit:
2367 // CHECK11-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2368 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
2369 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
2370 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
2371 // CHECK11:       omp.precond.end:
2372 // CHECK11-NEXT:    ret void
2373 //
2374 //
2375 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105
2376 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2377 // CHECK11-NEXT:  entry:
2378 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2379 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2380 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2381 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2382 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2383 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2384 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2385 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2386 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
2387 // CHECK11-NEXT:    ret void
2388 //
2389 //
2390 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.omp_outlined
2391 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3]] {
2392 // CHECK11-NEXT:  entry:
2393 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2394 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2395 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
2396 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2397 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2398 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2399 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2400 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2401 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2402 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2403 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2404 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2405 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2406 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2407 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
2408 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2409 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2410 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
2411 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2412 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2413 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
2414 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2415 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2416 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
2417 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
2418 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2419 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2420 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2421 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2422 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2423 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
2424 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2425 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2426 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2427 // CHECK11:       omp.precond.then:
2428 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2429 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2430 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
2431 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2432 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2433 // CHECK11-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2434 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2435 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2436 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2437 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2438 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2439 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2440 // CHECK11:       cond.true:
2441 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2442 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2443 // CHECK11:       cond.false:
2444 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2445 // CHECK11-NEXT:    br label [[COND_END]]
2446 // CHECK11:       cond.end:
2447 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2448 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2449 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2450 // CHECK11-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
2451 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2452 // CHECK11:       omp.inner.for.cond:
2453 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2454 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2455 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2456 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2457 // CHECK11:       omp.inner.for.body:
2458 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2459 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
2460 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2461 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
2462 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[I3]], align 4
2463 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP17]]
2464 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2465 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2466 // CHECK11:       omp.body.continue:
2467 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2468 // CHECK11:       omp.inner.for.inc:
2469 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2470 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
2471 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
2472 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2473 // CHECK11:       omp.inner.for.end:
2474 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2475 // CHECK11:       omp.loop.exit:
2476 // CHECK11-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2477 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
2478 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP20]])
2479 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
2480 // CHECK11:       omp.precond.end:
2481 // CHECK11-NEXT:    ret void
2482 //
2483 //
2484 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110
2485 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2486 // CHECK11-NEXT:  entry:
2487 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2488 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2489 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2490 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2491 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
2492 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
2493 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2494 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2495 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2496 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2497 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N_ADDR]], align 4
2498 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
2499 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2500 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
2501 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__CASTED]], align 4
2502 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]], i32 [[TMP4]])
2503 // CHECK11-NEXT:    ret void
2504 //
2505 //
2506 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.omp_outlined
2507 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
2508 // CHECK11-NEXT:  entry:
2509 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2510 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2511 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
2512 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2513 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2514 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2515 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2516 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2517 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2518 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2519 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2520 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2521 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2522 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2523 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2524 // CHECK11-NEXT:    [[I4:%.*]] = alloca i32, align 4
2525 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2526 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2527 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
2528 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
2529 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2530 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
2531 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
2532 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
2533 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2534 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
2535 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2536 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2537 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2538 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2539 // CHECK11-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
2540 // CHECK11-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2]], align 4
2541 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
2542 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2543 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2544 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2545 // CHECK11:       omp.precond.then:
2546 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2547 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
2548 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
2549 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2550 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2551 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ADDR]], align 4
2552 // CHECK11-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2553 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
2554 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP9]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
2555 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2556 // CHECK11:       omp.dispatch.cond:
2557 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2558 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
2559 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
2560 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2561 // CHECK11:       cond.true:
2562 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
2563 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2564 // CHECK11:       cond.false:
2565 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2566 // CHECK11-NEXT:    br label [[COND_END]]
2567 // CHECK11:       cond.end:
2568 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
2569 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2570 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2571 // CHECK11-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
2572 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2573 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2574 // CHECK11-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
2575 // CHECK11-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2576 // CHECK11:       omp.dispatch.body:
2577 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2578 // CHECK11:       omp.inner.for.cond:
2579 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12:![0-9]+]]
2580 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP12]]
2581 // CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
2582 // CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2583 // CHECK11:       omp.inner.for.body:
2584 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
2585 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
2586 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2587 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP12]]
2588 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[I4]], align 4, !llvm.access.group [[ACC_GRP12]]
2589 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP20]]
2590 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP12]]
2591 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2592 // CHECK11:       omp.body.continue:
2593 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2594 // CHECK11:       omp.inner.for.inc:
2595 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
2596 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP21]], 1
2597 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP12]]
2598 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
2599 // CHECK11:       omp.inner.for.end:
2600 // CHECK11-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2601 // CHECK11:       omp.dispatch.inc:
2602 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2603 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2604 // CHECK11-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
2605 // CHECK11-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_LB]], align 4
2606 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2607 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2608 // CHECK11-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
2609 // CHECK11-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_UB]], align 4
2610 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND]]
2611 // CHECK11:       omp.dispatch.end:
2612 // CHECK11-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2613 // CHECK11-NEXT:    [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4
2614 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP27]])
2615 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
2616 // CHECK11:       omp.precond.end:
2617 // CHECK11-NEXT:    ret void
2618 //
2619 //
2620 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
2621 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR6:[0-9]+]] comdat {
2622 // CHECK11-NEXT:  entry:
2623 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
2624 // CHECK11-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
2625 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
2626 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
2627 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
2628 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2629 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2630 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x ptr], align 4
2631 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x ptr], align 4
2632 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x ptr], align 4
2633 // CHECK11-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
2634 // CHECK11-NEXT:    [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2635 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x ptr], align 4
2636 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x ptr], align 4
2637 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x ptr], align 4
2638 // CHECK11-NEXT:    [[_TMP11:%.*]] = alloca i32, align 4
2639 // CHECK11-NEXT:    [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2640 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
2641 // CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2642 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP0]], align 4
2643 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2644 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
2645 // CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2646 // CHECK11-NEXT:    store ptr null, ptr [[TMP2]], align 4
2647 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2648 // CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2649 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2650 // CHECK11-NEXT:    store i32 2, ptr [[TMP5]], align 4
2651 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2652 // CHECK11-NEXT:    store i32 1, ptr [[TMP6]], align 4
2653 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2654 // CHECK11-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
2655 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2656 // CHECK11-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
2657 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2658 // CHECK11-NEXT:    store ptr @.offload_sizes.5, ptr [[TMP9]], align 4
2659 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2660 // CHECK11-NEXT:    store ptr @.offload_maptypes.6, ptr [[TMP10]], align 4
2661 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2662 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
2663 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2664 // CHECK11-NEXT:    store ptr null, ptr [[TMP12]], align 4
2665 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2666 // CHECK11-NEXT:    store i64 10, ptr [[TMP13]], align 8
2667 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2668 // CHECK11-NEXT:    store i64 0, ptr [[TMP14]], align 8
2669 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2670 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
2671 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2672 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
2673 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2674 // CHECK11-NEXT:    store i32 0, ptr [[TMP17]], align 4
2675 // CHECK11-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.region_id, ptr [[KERNEL_ARGS]])
2676 // CHECK11-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2677 // CHECK11-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2678 // CHECK11:       omp_offload.failed:
2679 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79(ptr [[A]]) #[[ATTR4]]
2680 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2681 // CHECK11:       omp_offload.cont:
2682 // CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2683 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP20]], align 4
2684 // CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2685 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP21]], align 4
2686 // CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
2687 // CHECK11-NEXT:    store ptr null, ptr [[TMP22]], align 4
2688 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2689 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2690 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 0
2691 // CHECK11-NEXT:    store i32 2, ptr [[TMP25]], align 4
2692 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 1
2693 // CHECK11-NEXT:    store i32 1, ptr [[TMP26]], align 4
2694 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 2
2695 // CHECK11-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 4
2696 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 3
2697 // CHECK11-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 4
2698 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 4
2699 // CHECK11-NEXT:    store ptr @.offload_sizes.7, ptr [[TMP29]], align 4
2700 // CHECK11-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 5
2701 // CHECK11-NEXT:    store ptr @.offload_maptypes.8, ptr [[TMP30]], align 4
2702 // CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 6
2703 // CHECK11-NEXT:    store ptr null, ptr [[TMP31]], align 4
2704 // CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 7
2705 // CHECK11-NEXT:    store ptr null, ptr [[TMP32]], align 4
2706 // CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 8
2707 // CHECK11-NEXT:    store i64 10, ptr [[TMP33]], align 8
2708 // CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 9
2709 // CHECK11-NEXT:    store i64 0, ptr [[TMP34]], align 8
2710 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 10
2711 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
2712 // CHECK11-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 11
2713 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
2714 // CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS5]], i32 0, i32 12
2715 // CHECK11-NEXT:    store i32 0, ptr [[TMP37]], align 4
2716 // CHECK11-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, ptr [[KERNEL_ARGS5]])
2717 // CHECK11-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
2718 // CHECK11-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
2719 // CHECK11:       omp_offload.failed6:
2720 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84(ptr [[A]]) #[[ATTR4]]
2721 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT7]]
2722 // CHECK11:       omp_offload.cont7:
2723 // CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2724 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP40]], align 4
2725 // CHECK11-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2726 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP41]], align 4
2727 // CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
2728 // CHECK11-NEXT:    store ptr null, ptr [[TMP42]], align 4
2729 // CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2730 // CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2731 // CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 0
2732 // CHECK11-NEXT:    store i32 2, ptr [[TMP45]], align 4
2733 // CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 1
2734 // CHECK11-NEXT:    store i32 1, ptr [[TMP46]], align 4
2735 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 2
2736 // CHECK11-NEXT:    store ptr [[TMP43]], ptr [[TMP47]], align 4
2737 // CHECK11-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 3
2738 // CHECK11-NEXT:    store ptr [[TMP44]], ptr [[TMP48]], align 4
2739 // CHECK11-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 4
2740 // CHECK11-NEXT:    store ptr @.offload_sizes.9, ptr [[TMP49]], align 4
2741 // CHECK11-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 5
2742 // CHECK11-NEXT:    store ptr @.offload_maptypes.10, ptr [[TMP50]], align 4
2743 // CHECK11-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 6
2744 // CHECK11-NEXT:    store ptr null, ptr [[TMP51]], align 4
2745 // CHECK11-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 7
2746 // CHECK11-NEXT:    store ptr null, ptr [[TMP52]], align 4
2747 // CHECK11-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 8
2748 // CHECK11-NEXT:    store i64 10, ptr [[TMP53]], align 8
2749 // CHECK11-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 9
2750 // CHECK11-NEXT:    store i64 0, ptr [[TMP54]], align 8
2751 // CHECK11-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 10
2752 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP55]], align 4
2753 // CHECK11-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 11
2754 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP56]], align 4
2755 // CHECK11-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS12]], i32 0, i32 12
2756 // CHECK11-NEXT:    store i32 0, ptr [[TMP57]], align 4
2757 // CHECK11-NEXT:    [[TMP58:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.region_id, ptr [[KERNEL_ARGS12]])
2758 // CHECK11-NEXT:    [[TMP59:%.*]] = icmp ne i32 [[TMP58]], 0
2759 // CHECK11-NEXT:    br i1 [[TMP59]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
2760 // CHECK11:       omp_offload.failed13:
2761 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89(ptr [[A]]) #[[ATTR4]]
2762 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
2763 // CHECK11:       omp_offload.cont14:
2764 // CHECK11-NEXT:    ret i32 0
2765 //
2766 //
2767 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79
2768 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2769 // CHECK11-NEXT:  entry:
2770 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2771 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2772 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2773 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.omp_outlined, ptr [[TMP0]])
2774 // CHECK11-NEXT:    ret void
2775 //
2776 //
2777 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.omp_outlined
2778 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
2779 // CHECK11-NEXT:  entry:
2780 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2781 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2782 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2783 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2784 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2785 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2786 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2787 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2788 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2789 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2790 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2791 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2792 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2793 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2794 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2795 // CHECK11-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
2796 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2797 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2798 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2799 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2800 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2801 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2802 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2803 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2804 // CHECK11:       cond.true:
2805 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2806 // CHECK11:       cond.false:
2807 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2808 // CHECK11-NEXT:    br label [[COND_END]]
2809 // CHECK11:       cond.end:
2810 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2811 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2812 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2813 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2814 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2815 // CHECK11:       omp.inner.for.cond:
2816 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2817 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2818 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2819 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2820 // CHECK11:       omp.inner.for.body:
2821 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2822 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2823 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2824 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2825 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
2826 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP9]]
2827 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2828 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2829 // CHECK11:       omp.body.continue:
2830 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2831 // CHECK11:       omp.inner.for.inc:
2832 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2833 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2834 // CHECK11-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
2835 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2836 // CHECK11:       omp.inner.for.end:
2837 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2838 // CHECK11:       omp.loop.exit:
2839 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
2840 // CHECK11-NEXT:    ret void
2841 //
2842 //
2843 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
2844 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2845 // CHECK11-NEXT:  entry:
2846 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2847 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2848 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2849 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined, ptr [[TMP0]])
2850 // CHECK11-NEXT:    ret void
2851 //
2852 //
2853 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.omp_outlined
2854 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
2855 // CHECK11-NEXT:  entry:
2856 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2857 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2858 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2859 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2860 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2861 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2862 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2863 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2864 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2865 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2866 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2867 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2868 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2869 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2870 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2871 // CHECK11-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
2872 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2873 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2874 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2875 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2876 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2877 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2878 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2879 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2880 // CHECK11:       cond.true:
2881 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2882 // CHECK11:       cond.false:
2883 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2884 // CHECK11-NEXT:    br label [[COND_END]]
2885 // CHECK11:       cond.end:
2886 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2887 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2888 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2889 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2890 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2891 // CHECK11:       omp.inner.for.cond:
2892 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2893 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2894 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2895 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2896 // CHECK11:       omp.inner.for.body:
2897 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2898 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2899 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2900 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2901 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[I]], align 4
2902 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP9]]
2903 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2904 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2905 // CHECK11:       omp.body.continue:
2906 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2907 // CHECK11:       omp.inner.for.inc:
2908 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2909 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2910 // CHECK11-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
2911 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
2912 // CHECK11:       omp.inner.for.end:
2913 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2914 // CHECK11:       omp.loop.exit:
2915 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
2916 // CHECK11-NEXT:    ret void
2917 //
2918 //
2919 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89
2920 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2921 // CHECK11-NEXT:  entry:
2922 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2923 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2924 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2925 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.omp_outlined, ptr [[TMP0]])
2926 // CHECK11-NEXT:    ret void
2927 //
2928 //
2929 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.omp_outlined
2930 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
2931 // CHECK11-NEXT:  entry:
2932 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2933 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2934 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
2935 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2936 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2937 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2938 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2939 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2940 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2941 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
2942 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2943 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2944 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
2945 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
2946 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2947 // CHECK11-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
2948 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2949 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2950 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2951 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2952 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 91, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 10)
2953 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2954 // CHECK11:       omp.dispatch.cond:
2955 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2956 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2957 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2958 // CHECK11:       cond.true:
2959 // CHECK11-NEXT:    br label [[COND_END:%.*]]
2960 // CHECK11:       cond.false:
2961 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2962 // CHECK11-NEXT:    br label [[COND_END]]
2963 // CHECK11:       cond.end:
2964 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2965 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2966 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2967 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2968 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2969 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2970 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2971 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2972 // CHECK11:       omp.dispatch.body:
2973 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2974 // CHECK11:       omp.inner.for.cond:
2975 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15:![0-9]+]]
2976 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP15]]
2977 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2978 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2979 // CHECK11:       omp.inner.for.body:
2980 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
2981 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2982 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2983 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
2984 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP15]]
2985 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP11]]
2986 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4, !llvm.access.group [[ACC_GRP15]]
2987 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2988 // CHECK11:       omp.body.continue:
2989 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2990 // CHECK11:       omp.inner.for.inc:
2991 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
2992 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2993 // CHECK11-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP15]]
2994 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
2995 // CHECK11:       omp.inner.for.end:
2996 // CHECK11-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2997 // CHECK11:       omp.dispatch.inc:
2998 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2999 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
3000 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
3001 // CHECK11-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_LB]], align 4
3002 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3003 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
3004 // CHECK11-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
3005 // CHECK11-NEXT:    store i32 [[ADD5]], ptr [[DOTOMP_UB]], align 4
3006 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND]]
3007 // CHECK11:       omp.dispatch.end:
3008 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
3009 // CHECK11-NEXT:    ret void
3010 //
3011 //
3012 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3013 // CHECK11-SAME: () #[[ATTR7:[0-9]+]] {
3014 // CHECK11-NEXT:  entry:
3015 // CHECK11-NEXT:    call void @__tgt_register_requires(i64 1)
3016 // CHECK11-NEXT:    ret void
3017 //