clang/test/OpenMP/teams_generic_loop_codegen-1.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 // Test host codegen.
   6 // 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
   7 // 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
   8 // 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
   9 // 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
  10 // 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
  11 // 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
  12
  13 // 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}}"
  14 // 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
  15 // 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}}"
  16 // 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}}"
  17 // 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
  18 // 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}}"
  19 #ifdef CK1
  20
  21 int a[100];
  22
  23 int teams_argument_global(int n){
  24   int te = n / 128;
  25   int th = 128;
  26   // discard n_addr
  27
  28   #pragma omp target
  29   #pragma omp teams loop num_teams(te), thread_limit(th)
  30   for(int i = 0; i < n; i++) {
  31     a[i] = 0;
  32   }
  33
  34   #pragma omp target
  35   {{{
  36   #pragma omp teams loop
  37   for(int i = 0; i < n; i++) {
  38     a[i] = 0;
  39   }
  40   }}}
  41
  42   // outlined target regions
  43
  44
  45
  46
  47   return a[0];
  48 }
  49
  50 #endif // CK1
  51
  52 // Test host codegen.
  53 // 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
  54 // 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
  55 // 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
  56 // 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
  57 // 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
  58 // 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
  59
  60 // 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}}"
  61 // 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
  62 // 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}}"
  63 // 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}}"
  64 // 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
  65 // 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}}"
  66 #ifdef CK2
  67
  68 int teams_local_arg(void) {
  69   int n = 100;
  70   int a[n];
  71
  72   #pragma omp target
  73   #pragma omp teams loop
  74   for(int i = 0; i < n; i++) {
  75     a[i] = 0;
  76   }
  77
  78   // outlined target region
  79
  80
  81   return a[0];
  82 }
  83 #endif // CK2
  84
  85 // Test host codegen.
  86 // RUN: %clang_cc1 -DCK3 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17
  87 // RUN: %clang_cc1 -DCK3 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  88 // RUN: %clang_cc1 -DCK3 -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=CHECK17
  89 // RUN: %clang_cc1 -DCK3 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19
  90 // RUN: %clang_cc1 -DCK3 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  91 // RUN: %clang_cc1 -DCK3 -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=CHECK19
  92
  93 // RUN: %clang_cc1 -DCK3 -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}}"
  94 // RUN: %clang_cc1 -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  95 // RUN: %clang_cc1 -DCK3 -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}}"
  96 // RUN: %clang_cc1 -DCK3 -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}}"
  97 // RUN: %clang_cc1 -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  98 // RUN: %clang_cc1 -DCK3 -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}}"
  99 #ifdef CK3
 100
 101
 102 template <typename T, int X, long long Y>
 103 struct SS{
 104   T a[X];
 105   float b;
 106   int foo(void) {
 107
 108     #pragma omp target
 109     #pragma omp teams loop
 110     for(int i = 0; i < X; i++) {
 111       a[i] = (T)0;
 112     }
 113
 114       // outlined target region
 115
 116
 117     return a[0];
 118   }
 119 };
 120
 121 int teams_template_struct(void) {
 122   SS<int, 123, 456> V;
 123   return V.foo();
 124
 125 }
 126 #endif // CK3
 127
 128 // Test host codegen.
 129 // RUN: %clang_cc1 -DCK4 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK25
 130 // RUN: %clang_cc1 -DCK4 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
 131 // RUN: %clang_cc1 -DCK4 -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=CHECK25
 132 // RUN: %clang_cc1 -DCK4 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK27
 133 // RUN: %clang_cc1 -DCK4 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
 134 // RUN: %clang_cc1 -DCK4 -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=CHECK27
 135
 136 // RUN: %clang_cc1 -DCK4 -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}}"
 137 // RUN: %clang_cc1 -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
 138 // RUN: %clang_cc1 -DCK4 -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}}"
 139 // RUN: %clang_cc1 -DCK4 -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}}"
 140 // RUN: %clang_cc1 -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
 141 // RUN: %clang_cc1 -DCK4 -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}}"
 142
 143 #ifdef CK4
 144
 145 template <typename T, int n>
 146 int tmain(T argc) {
 147   T a[n];
 148   int te = n/128;
 149   int th = 128;
 150 #pragma omp target
 151 #pragma omp teams loop num_teams(te) thread_limit(th)
 152   for(int i = 0; i < n; i++) {
 153     a[i] = (T)0;
 154   }
 155   return 0;
 156 }
 157
 158 int main (int argc, char **argv) {
 159   int n = 100;
 160   int a[n];
 161 #pragma omp target
 162 #pragma omp teams loop
 163   for(int i = 0; i < n; i++) {
 164     a[i] = 0;
 165   }
 166   return tmain<int, 10>(argc);
 167 }
 168
 169
 170
 171
 172
 173
 174
 175 #endif // CK4
 176 #endif
 177 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
 178 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 179 // CHECK1-NEXT:  entry:
 180 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 181 // CHECK1-NEXT:    [[TE:%.*]] = alloca i32, align 4
 182 // CHECK1-NEXT:    [[TH:%.*]] = alloca i32, align 4
 183 // CHECK1-NEXT:    [[TE_CASTED:%.*]] = alloca i64, align 8
 184 // CHECK1-NEXT:    [[TH_CASTED:%.*]] = alloca i64, align 8
 185 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 186 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 8
 187 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 8
 188 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 8
 189 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 190 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 191 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 192 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 193 // CHECK1-NEXT:    [[N_CASTED4:%.*]] = alloca i64, align 8
 194 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x ptr], align 8
 195 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x ptr], align 8
 196 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x ptr], align 8
 197 // CHECK1-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
 198 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
 199 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
 200 // CHECK1-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 201 // CHECK1-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 202 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
 203 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
 204 // CHECK1-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
 205 // CHECK1-NEXT:    store i32 128, ptr [[TH]], align 4
 206 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
 207 // CHECK1-NEXT:    store i32 [[TMP1]], ptr [[TE_CASTED]], align 4
 208 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[TE_CASTED]], align 8
 209 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH]], align 4
 210 // CHECK1-NEXT:    store i32 [[TMP3]], ptr [[TH_CASTED]], align 4
 211 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TH_CASTED]], align 8
 212 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
 213 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
 214 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[N_CASTED]], align 8
 215 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 216 // CHECK1-NEXT:    store i64 [[TMP2]], ptr [[TMP7]], align 8
 217 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 218 // CHECK1-NEXT:    store i64 [[TMP2]], ptr [[TMP8]], align 8
 219 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 220 // CHECK1-NEXT:    store ptr null, ptr [[TMP9]], align 8
 221 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 222 // CHECK1-NEXT:    store i64 [[TMP4]], ptr [[TMP10]], align 8
 223 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 224 // CHECK1-NEXT:    store i64 [[TMP4]], ptr [[TMP11]], align 8
 225 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
 226 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 227 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 228 // CHECK1-NEXT:    store i64 [[TMP6]], ptr [[TMP13]], align 8
 229 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 230 // CHECK1-NEXT:    store i64 [[TMP6]], ptr [[TMP14]], align 8
 231 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
 232 // CHECK1-NEXT:    store ptr null, ptr [[TMP15]], align 8
 233 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 234 // CHECK1-NEXT:    store ptr @a, ptr [[TMP16]], align 8
 235 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 236 // CHECK1-NEXT:    store ptr @a, ptr [[TMP17]], align 8
 237 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
 238 // CHECK1-NEXT:    store ptr null, ptr [[TMP18]], align 8
 239 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 240 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 241 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TE]], align 4
 242 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TH]], align 4
 243 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[N_ADDR]], align 4
 244 // CHECK1-NEXT:    store i32 [[TMP23]], ptr [[DOTCAPTURE_EXPR_]], align 4
 245 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 246 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
 247 // CHECK1-NEXT:    [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
 248 // CHECK1-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
 249 // CHECK1-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 250 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 251 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP25]], 1
 252 // CHECK1-NEXT:    [[TMP26:%.*]] = zext i32 [[ADD]] to i64
 253 // CHECK1-NEXT:    [[TMP27:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP21]], 0
 254 // CHECK1-NEXT:    [[TMP28:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP22]], 0
 255 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 256 // CHECK1-NEXT:    store i32 2, ptr [[TMP29]], align 4
 257 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 258 // CHECK1-NEXT:    store i32 4, ptr [[TMP30]], align 4
 259 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 260 // CHECK1-NEXT:    store ptr [[TMP19]], ptr [[TMP31]], align 8
 261 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 262 // CHECK1-NEXT:    store ptr [[TMP20]], ptr [[TMP32]], align 8
 263 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 264 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP33]], align 8
 265 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 266 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP34]], align 8
 267 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 268 // CHECK1-NEXT:    store ptr null, ptr [[TMP35]], align 8
 269 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 270 // CHECK1-NEXT:    store ptr null, ptr [[TMP36]], align 8
 271 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 272 // CHECK1-NEXT:    store i64 [[TMP26]], ptr [[TMP37]], align 8
 273 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 274 // CHECK1-NEXT:    store i64 0, ptr [[TMP38]], align 8
 275 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 276 // CHECK1-NEXT:    store [3 x i32] [[TMP27]], ptr [[TMP39]], align 4
 277 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 278 // CHECK1-NEXT:    store [3 x i32] [[TMP28]], ptr [[TMP40]], align 4
 279 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 280 // CHECK1-NEXT:    store i32 0, ptr [[TMP41]], align 4
 281 // CHECK1-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, ptr [[KERNEL_ARGS]])
 282 // CHECK1-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
 283 // CHECK1-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 284 // CHECK1:       omp_offload.failed:
 285 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28(i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], ptr @a) #[[ATTR2:[0-9]+]]
 286 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 287 // CHECK1:       omp_offload.cont:
 288 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, ptr [[N_ADDR]], align 4
 289 // CHECK1-NEXT:    store i32 [[TMP44]], ptr [[N_CASTED4]], align 4
 290 // CHECK1-NEXT:    [[TMP45:%.*]] = load i64, ptr [[N_CASTED4]], align 8
 291 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
 292 // CHECK1-NEXT:    store i64 [[TMP45]], ptr [[TMP46]], align 8
 293 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
 294 // CHECK1-NEXT:    store i64 [[TMP45]], ptr [[TMP47]], align 8
 295 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
 296 // CHECK1-NEXT:    store ptr null, ptr [[TMP48]], align 8
 297 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
 298 // CHECK1-NEXT:    store ptr @a, ptr [[TMP49]], align 8
 299 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
 300 // CHECK1-NEXT:    store ptr @a, ptr [[TMP50]], align 8
 301 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
 302 // CHECK1-NEXT:    store ptr null, ptr [[TMP51]], align 8
 303 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
 304 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
 305 // CHECK1-NEXT:    [[TMP54:%.*]] = load i32, ptr [[N_ADDR]], align 4
 306 // CHECK1-NEXT:    store i32 [[TMP54]], ptr [[DOTCAPTURE_EXPR_9]], align 4
 307 // CHECK1-NEXT:    [[TMP55:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_9]], align 4
 308 // CHECK1-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP55]], 0
 309 // CHECK1-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
 310 // CHECK1-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
 311 // CHECK1-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_10]], align 4
 312 // CHECK1-NEXT:    [[TMP56:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_10]], align 4
 313 // CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP56]], 1
 314 // CHECK1-NEXT:    [[TMP57:%.*]] = zext i32 [[ADD14]] to i64
 315 // CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
 316 // CHECK1-NEXT:    store i32 2, ptr [[TMP58]], align 4
 317 // CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
 318 // CHECK1-NEXT:    store i32 2, ptr [[TMP59]], align 4
 319 // CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
 320 // CHECK1-NEXT:    store ptr [[TMP52]], ptr [[TMP60]], align 8
 321 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
 322 // CHECK1-NEXT:    store ptr [[TMP53]], ptr [[TMP61]], align 8
 323 // CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
 324 // CHECK1-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP62]], align 8
 325 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
 326 // CHECK1-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP63]], align 8
 327 // CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
 328 // CHECK1-NEXT:    store ptr null, ptr [[TMP64]], align 8
 329 // CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
 330 // CHECK1-NEXT:    store ptr null, ptr [[TMP65]], align 8
 331 // CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
 332 // CHECK1-NEXT:    store i64 [[TMP57]], ptr [[TMP66]], align 8
 333 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
 334 // CHECK1-NEXT:    store i64 0, ptr [[TMP67]], align 8
 335 // CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
 336 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP68]], align 4
 337 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
 338 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
 339 // CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 340 // CHECK1-NEXT:    store i32 0, ptr [[TMP70]], align 4
 341 // CHECK1-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.region_id, ptr [[KERNEL_ARGS15]])
 342 // CHECK1-NEXT:    [[TMP72:%.*]] = icmp ne i32 [[TMP71]], 0
 343 // CHECK1-NEXT:    br i1 [[TMP72]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 344 // CHECK1:       omp_offload.failed16:
 345 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34(i64 [[TMP45]], ptr @a) #[[ATTR2]]
 346 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
 347 // CHECK1:       omp_offload.cont17:
 348 // CHECK1-NEXT:    [[TMP73:%.*]] = load i32, ptr @a, align 4
 349 // CHECK1-NEXT:    ret i32 [[TMP73]]
 350 //
 351 //
 352 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28
 353 // CHECK1-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
 354 // CHECK1-NEXT:  entry:
 355 // CHECK1-NEXT:    [[TE_ADDR:%.*]] = alloca i64, align 8
 356 // CHECK1-NEXT:    [[TH_ADDR:%.*]] = alloca i64, align 8
 357 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 358 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 359 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
 360 // CHECK1-NEXT:    store i64 [[TE]], ptr [[TE_ADDR]], align 8
 361 // CHECK1-NEXT:    store i64 [[TH]], ptr [[TH_ADDR]], align 8
 362 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 363 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 364 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 365 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
 366 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
 367 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
 368 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined, ptr [[N_ADDR]], ptr [[TMP1]])
 369 // CHECK1-NEXT:    ret void
 370 //
 371 //
 372 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined
 373 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
 374 // CHECK1-NEXT:  entry:
 375 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 376 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 377 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
 378 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 379 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 380 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 381 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 382 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 383 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 384 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 385 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 386 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 387 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 388 // CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
 389 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 390 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 391 // CHECK1-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
 392 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 393 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
 394 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 395 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
 396 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
 397 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 398 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
 399 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 400 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 401 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 402 // CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
 403 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 404 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
 405 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 406 // CHECK1:       omp.precond.then:
 407 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 408 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 409 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_COMB_UB]], align 4
 410 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 411 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 412 // CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 413 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
 414 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 415 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 416 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 417 // CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
 418 // CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 419 // CHECK1:       cond.true:
 420 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 421 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 422 // CHECK1:       cond.false:
 423 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 424 // CHECK1-NEXT:    br label [[COND_END]]
 425 // CHECK1:       cond.end:
 426 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
 427 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 428 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 429 // CHECK1-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
 430 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 431 // CHECK1:       omp.inner.for.cond:
 432 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 433 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 434 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
 435 // CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 436 // CHECK1:       omp.inner.for.body:
 437 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 438 // CHECK1-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP15]] to i64
 439 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 440 // CHECK1-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
 441 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined, i64 [[TMP16]], i64 [[TMP18]], ptr [[TMP0]], ptr [[TMP1]])
 442 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 443 // CHECK1:       omp.inner.for.inc:
 444 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 445 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 446 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
 447 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 448 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 449 // CHECK1:       omp.inner.for.end:
 450 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 451 // CHECK1:       omp.loop.exit:
 452 // CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 453 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 454 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP22]])
 455 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 456 // CHECK1:       omp.precond.end:
 457 // CHECK1-NEXT:    ret void
 458 //
 459 //
 460 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined
 461 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
 462 // CHECK1-NEXT:  entry:
 463 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 464 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 465 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 466 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 467 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
 468 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 469 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 470 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 471 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 472 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 473 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 474 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 475 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 476 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 477 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 478 // CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
 479 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 480 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 481 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 482 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 483 // CHECK1-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
 484 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 485 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
 486 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 487 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
 488 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
 489 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 490 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
 491 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 492 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 493 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 494 // CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
 495 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 496 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
 497 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 498 // CHECK1:       omp.precond.then:
 499 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 500 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 501 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
 502 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 503 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP6]] to i32
 504 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 505 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 506 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 507 // CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
 508 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 509 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 510 // CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 511 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
 512 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 513 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 514 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 515 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
 516 // CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 517 // CHECK1:       cond.true:
 518 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 519 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 520 // CHECK1:       cond.false:
 521 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 522 // CHECK1-NEXT:    br label [[COND_END]]
 523 // CHECK1:       cond.end:
 524 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
 525 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 526 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 527 // CHECK1-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
 528 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 529 // CHECK1:       omp.inner.for.cond:
 530 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 531 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 532 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
 533 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 534 // CHECK1:       omp.inner.for.body:
 535 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 536 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
 537 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 538 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
 539 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I4]], align 4
 540 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
 541 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
 542 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 543 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 544 // CHECK1:       omp.body.continue:
 545 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 546 // CHECK1:       omp.inner.for.inc:
 547 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 548 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
 549 // CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
 550 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 551 // CHECK1:       omp.inner.for.end:
 552 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 553 // CHECK1:       omp.loop.exit:
 554 // CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 555 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 556 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 557 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 558 // CHECK1:       omp.precond.end:
 559 // CHECK1-NEXT:    ret void
 560 //
 561 //
 562 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34
 563 // CHECK1-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR3:[0-9]+]] {
 564 // CHECK1-NEXT:  entry:
 565 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 566 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 567 // CHECK1-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 568 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 569 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 570 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined, ptr [[N_ADDR]], ptr [[TMP0]])
 571 // CHECK1-NEXT:    ret void
 572 //
 573 //
 574 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined
 575 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
 576 // CHECK1-NEXT:  entry:
 577 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 578 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 579 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
 580 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 581 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 582 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 583 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 584 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 585 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 586 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 587 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 588 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 589 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 590 // CHECK1-NEXT:    [[I3:%.*]] = alloca i32, align 4
 591 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 592 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 593 // CHECK1-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
 594 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 595 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
 596 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 597 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
 598 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
 599 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 600 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
 601 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 602 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 603 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 604 // CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
 605 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 606 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
 607 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 608 // CHECK1:       omp.precond.then:
 609 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 610 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 611 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_COMB_UB]], align 4
 612 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 613 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 614 // CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 615 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
 616 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 617 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 618 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 619 // CHECK1-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
 620 // CHECK1-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 621 // CHECK1:       cond.true:
 622 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 623 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 624 // CHECK1:       cond.false:
 625 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 626 // CHECK1-NEXT:    br label [[COND_END]]
 627 // CHECK1:       cond.end:
 628 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
 629 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 630 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 631 // CHECK1-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
 632 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 633 // CHECK1:       omp.inner.for.cond:
 634 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 635 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 636 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
 637 // CHECK1-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 638 // CHECK1:       omp.inner.for.body:
 639 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 640 // CHECK1-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP15]] to i64
 641 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 642 // CHECK1-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
 643 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined, i64 [[TMP16]], i64 [[TMP18]], ptr [[TMP0]], ptr [[TMP1]])
 644 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 645 // CHECK1:       omp.inner.for.inc:
 646 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 647 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 648 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
 649 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 650 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 651 // CHECK1:       omp.inner.for.end:
 652 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 653 // CHECK1:       omp.loop.exit:
 654 // CHECK1-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 655 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
 656 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
 657 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 658 // CHECK1:       omp.precond.end:
 659 // CHECK1-NEXT:    ret void
 660 //
 661 //
 662 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined
 663 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
 664 // CHECK1-NEXT:  entry:
 665 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 666 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 667 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 668 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 669 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
 670 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 671 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 672 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 673 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 674 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 675 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 676 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 677 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 678 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 679 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 680 // CHECK1-NEXT:    [[I4:%.*]] = alloca i32, align 4
 681 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 682 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 683 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 684 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 685 // CHECK1-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
 686 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 687 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
 688 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 689 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
 690 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
 691 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 692 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
 693 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 694 // CHECK1-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 695 // CHECK1-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 696 // CHECK1-NEXT:    store i32 0, ptr [[I]], align 4
 697 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 698 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
 699 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
 700 // CHECK1:       omp.precond.then:
 701 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 702 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 703 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
 704 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 705 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP6]] to i32
 706 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 707 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
 708 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 709 // CHECK1-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
 710 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 711 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 712 // CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 713 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
 714 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 715 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 716 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 717 // CHECK1-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
 718 // CHECK1-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 719 // CHECK1:       cond.true:
 720 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 721 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 722 // CHECK1:       cond.false:
 723 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 724 // CHECK1-NEXT:    br label [[COND_END]]
 725 // CHECK1:       cond.end:
 726 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
 727 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 728 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 729 // CHECK1-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
 730 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 731 // CHECK1:       omp.inner.for.cond:
 732 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 733 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 734 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
 735 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 736 // CHECK1:       omp.inner.for.body:
 737 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 738 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
 739 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 740 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
 741 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I4]], align 4
 742 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
 743 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i64 0, i64 [[IDXPROM]]
 744 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
 745 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 746 // CHECK1:       omp.body.continue:
 747 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 748 // CHECK1:       omp.inner.for.inc:
 749 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 750 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
 751 // CHECK1-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
 752 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 753 // CHECK1:       omp.inner.for.end:
 754 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 755 // CHECK1:       omp.loop.exit:
 756 // CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 757 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 758 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 759 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
 760 // CHECK1:       omp.precond.end:
 761 // CHECK1-NEXT:    ret void
 762 //
 763 //
 764 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
 765 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
 766 // CHECK1-NEXT:  entry:
 767 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
 768 // CHECK1-NEXT:    ret void
 769 //
 770 //
 771 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
 772 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 773 // CHECK3-NEXT:  entry:
 774 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 775 // CHECK3-NEXT:    [[TE:%.*]] = alloca i32, align 4
 776 // CHECK3-NEXT:    [[TH:%.*]] = alloca i32, align 4
 777 // CHECK3-NEXT:    [[TE_CASTED:%.*]] = alloca i32, align 4
 778 // CHECK3-NEXT:    [[TH_CASTED:%.*]] = alloca i32, align 4
 779 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 780 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 4
 781 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 4
 782 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 4
 783 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 784 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 785 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 786 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 787 // CHECK3-NEXT:    [[N_CASTED4:%.*]] = alloca i32, align 4
 788 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x ptr], align 4
 789 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x ptr], align 4
 790 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x ptr], align 4
 791 // CHECK3-NEXT:    [[_TMP8:%.*]] = alloca i32, align 4
 792 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
 793 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
 794 // CHECK3-NEXT:    [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
 795 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 796 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N_ADDR]], align 4
 797 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
 798 // CHECK3-NEXT:    store i32 [[DIV]], ptr [[TE]], align 4
 799 // CHECK3-NEXT:    store i32 128, ptr [[TH]], align 4
 800 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE]], align 4
 801 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[TE_CASTED]], align 4
 802 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_CASTED]], align 4
 803 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH]], align 4
 804 // CHECK3-NEXT:    store i32 [[TMP3]], ptr [[TH_CASTED]], align 4
 805 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TH_CASTED]], align 4
 806 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_ADDR]], align 4
 807 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[N_CASTED]], align 4
 808 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N_CASTED]], align 4
 809 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 810 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[TMP7]], align 4
 811 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 812 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[TMP8]], align 4
 813 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 814 // CHECK3-NEXT:    store ptr null, ptr [[TMP9]], align 4
 815 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 816 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[TMP10]], align 4
 817 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 818 // CHECK3-NEXT:    store i32 [[TMP4]], ptr [[TMP11]], align 4
 819 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
 820 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 821 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 822 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[TMP13]], align 4
 823 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 824 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[TMP14]], align 4
 825 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
 826 // CHECK3-NEXT:    store ptr null, ptr [[TMP15]], align 4
 827 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 828 // CHECK3-NEXT:    store ptr @a, ptr [[TMP16]], align 4
 829 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 830 // CHECK3-NEXT:    store ptr @a, ptr [[TMP17]], align 4
 831 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
 832 // CHECK3-NEXT:    store ptr null, ptr [[TMP18]], align 4
 833 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 834 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 835 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TE]], align 4
 836 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TH]], align 4
 837 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[N_ADDR]], align 4
 838 // CHECK3-NEXT:    store i32 [[TMP23]], ptr [[DOTCAPTURE_EXPR_]], align 4
 839 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 840 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP24]], 0
 841 // CHECK3-NEXT:    [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
 842 // CHECK3-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
 843 // CHECK3-NEXT:    store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 844 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
 845 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP25]], 1
 846 // CHECK3-NEXT:    [[TMP26:%.*]] = zext i32 [[ADD]] to i64
 847 // CHECK3-NEXT:    [[TMP27:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP21]], 0
 848 // CHECK3-NEXT:    [[TMP28:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP22]], 0
 849 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 850 // CHECK3-NEXT:    store i32 2, ptr [[TMP29]], align 4
 851 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 852 // CHECK3-NEXT:    store i32 4, ptr [[TMP30]], align 4
 853 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 854 // CHECK3-NEXT:    store ptr [[TMP19]], ptr [[TMP31]], align 4
 855 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 856 // CHECK3-NEXT:    store ptr [[TMP20]], ptr [[TMP32]], align 4
 857 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 858 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP33]], align 4
 859 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 860 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP34]], align 4
 861 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 862 // CHECK3-NEXT:    store ptr null, ptr [[TMP35]], align 4
 863 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 864 // CHECK3-NEXT:    store ptr null, ptr [[TMP36]], align 4
 865 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 866 // CHECK3-NEXT:    store i64 [[TMP26]], ptr [[TMP37]], align 8
 867 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 868 // CHECK3-NEXT:    store i64 0, ptr [[TMP38]], align 8
 869 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 870 // CHECK3-NEXT:    store [3 x i32] [[TMP27]], ptr [[TMP39]], align 4
 871 // CHECK3-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 872 // CHECK3-NEXT:    store [3 x i32] [[TMP28]], ptr [[TMP40]], align 4
 873 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 874 // CHECK3-NEXT:    store i32 0, ptr [[TMP41]], align 4
 875 // CHECK3-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, ptr [[KERNEL_ARGS]])
 876 // CHECK3-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
 877 // CHECK3-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 878 // CHECK3:       omp_offload.failed:
 879 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28(i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], ptr @a) #[[ATTR2:[0-9]+]]
 880 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 881 // CHECK3:       omp_offload.cont:
 882 // CHECK3-NEXT:    [[TMP44:%.*]] = load i32, ptr [[N_ADDR]], align 4
 883 // CHECK3-NEXT:    store i32 [[TMP44]], ptr [[N_CASTED4]], align 4
 884 // CHECK3-NEXT:    [[TMP45:%.*]] = load i32, ptr [[N_CASTED4]], align 4
 885 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
 886 // CHECK3-NEXT:    store i32 [[TMP45]], ptr [[TMP46]], align 4
 887 // CHECK3-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
 888 // CHECK3-NEXT:    store i32 [[TMP45]], ptr [[TMP47]], align 4
 889 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
 890 // CHECK3-NEXT:    store ptr null, ptr [[TMP48]], align 4
 891 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
 892 // CHECK3-NEXT:    store ptr @a, ptr [[TMP49]], align 4
 893 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
 894 // CHECK3-NEXT:    store ptr @a, ptr [[TMP50]], align 4
 895 // CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
 896 // CHECK3-NEXT:    store ptr null, ptr [[TMP51]], align 4
 897 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
 898 // CHECK3-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
 899 // CHECK3-NEXT:    [[TMP54:%.*]] = load i32, ptr [[N_ADDR]], align 4
 900 // CHECK3-NEXT:    store i32 [[TMP54]], ptr [[DOTCAPTURE_EXPR_9]], align 4
 901 // CHECK3-NEXT:    [[TMP55:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_9]], align 4
 902 // CHECK3-NEXT:    [[SUB11:%.*]] = sub nsw i32 [[TMP55]], 0
 903 // CHECK3-NEXT:    [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
 904 // CHECK3-NEXT:    [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
 905 // CHECK3-NEXT:    store i32 [[SUB13]], ptr [[DOTCAPTURE_EXPR_10]], align 4
 906 // CHECK3-NEXT:    [[TMP56:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_10]], align 4
 907 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP56]], 1
 908 // CHECK3-NEXT:    [[TMP57:%.*]] = zext i32 [[ADD14]] to i64
 909 // CHECK3-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 0
 910 // CHECK3-NEXT:    store i32 2, ptr [[TMP58]], align 4
 911 // CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 1
 912 // CHECK3-NEXT:    store i32 2, ptr [[TMP59]], align 4
 913 // CHECK3-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 2
 914 // CHECK3-NEXT:    store ptr [[TMP52]], ptr [[TMP60]], align 4
 915 // CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 3
 916 // CHECK3-NEXT:    store ptr [[TMP53]], ptr [[TMP61]], align 4
 917 // CHECK3-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 4
 918 // CHECK3-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP62]], align 4
 919 // CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 5
 920 // CHECK3-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP63]], align 4
 921 // CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 6
 922 // CHECK3-NEXT:    store ptr null, ptr [[TMP64]], align 4
 923 // CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 7
 924 // CHECK3-NEXT:    store ptr null, ptr [[TMP65]], align 4
 925 // CHECK3-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 8
 926 // CHECK3-NEXT:    store i64 [[TMP57]], ptr [[TMP66]], align 8
 927 // CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 9
 928 // CHECK3-NEXT:    store i64 0, ptr [[TMP67]], align 8
 929 // CHECK3-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 10
 930 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP68]], align 4
 931 // CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 11
 932 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP69]], align 4
 933 // CHECK3-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS15]], i32 0, i32 12
 934 // CHECK3-NEXT:    store i32 0, ptr [[TMP70]], align 4
 935 // CHECK3-NEXT:    [[TMP71:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.region_id, ptr [[KERNEL_ARGS15]])
 936 // CHECK3-NEXT:    [[TMP72:%.*]] = icmp ne i32 [[TMP71]], 0
 937 // CHECK3-NEXT:    br i1 [[TMP72]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
 938 // CHECK3:       omp_offload.failed16:
 939 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34(i32 [[TMP45]], ptr @a) #[[ATTR2]]
 940 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
 941 // CHECK3:       omp_offload.cont17:
 942 // CHECK3-NEXT:    [[TMP73:%.*]] = load i32, ptr @a, align 4
 943 // CHECK3-NEXT:    ret i32 [[TMP73]]
 944 //
 945 //
 946 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28
 947 // CHECK3-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
 948 // CHECK3-NEXT:  entry:
 949 // CHECK3-NEXT:    [[TE_ADDR:%.*]] = alloca i32, align 4
 950 // CHECK3-NEXT:    [[TH_ADDR:%.*]] = alloca i32, align 4
 951 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
 952 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 953 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
 954 // CHECK3-NEXT:    store i32 [[TE]], ptr [[TE_ADDR]], align 4
 955 // CHECK3-NEXT:    store i32 [[TH]], ptr [[TH_ADDR]], align 4
 956 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
 957 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 958 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 959 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
 960 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
 961 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
 962 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined, ptr [[N_ADDR]], ptr [[TMP1]])
 963 // CHECK3-NEXT:    ret void
 964 //
 965 //
 966 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined
 967 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
 968 // CHECK3-NEXT:  entry:
 969 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 970 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 971 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
 972 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
 973 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 974 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 975 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 976 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
 977 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 978 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 979 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 980 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 981 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 982 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
 983 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 984 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 985 // CHECK3-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
 986 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
 987 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
 988 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
 989 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
 990 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
 991 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 992 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
 993 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 994 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
 995 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
 996 // CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
 997 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 998 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
 999 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1000 // CHECK3:       omp.precond.then:
1001 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1002 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1003 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_COMB_UB]], align 4
1004 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1005 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1006 // CHECK3-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1007 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
1008 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1009 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1010 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1011 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1012 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1013 // CHECK3:       cond.true:
1014 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1015 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1016 // CHECK3:       cond.false:
1017 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1018 // CHECK3-NEXT:    br label [[COND_END]]
1019 // CHECK3:       cond.end:
1020 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1021 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1022 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1023 // CHECK3-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
1024 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1025 // CHECK3:       omp.inner.for.cond:
1026 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1027 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1028 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1029 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1030 // CHECK3:       omp.inner.for.body:
1031 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1032 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1033 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined, i32 [[TMP15]], i32 [[TMP16]], ptr [[TMP0]], ptr [[TMP1]])
1034 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1035 // CHECK3:       omp.inner.for.inc:
1036 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1037 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1038 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
1039 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1040 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1041 // CHECK3:       omp.inner.for.end:
1042 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1043 // CHECK3:       omp.loop.exit:
1044 // CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1045 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
1046 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP20]])
1047 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1048 // CHECK3:       omp.precond.end:
1049 // CHECK3-NEXT:    ret void
1050 //
1051 //
1052 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.omp_outlined.omp_outlined
1053 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1054 // CHECK3-NEXT:  entry:
1055 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1056 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1057 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1058 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1059 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1060 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1061 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1062 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1063 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1064 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1065 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1066 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1067 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1068 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1069 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1070 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
1071 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1072 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1073 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1074 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1075 // CHECK3-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1076 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1077 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1078 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1079 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
1080 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
1081 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1082 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1083 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1084 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1085 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1086 // CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
1087 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1088 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1089 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1090 // CHECK3:       omp.precond.then:
1091 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1092 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1093 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
1094 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1095 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1096 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_LB]], align 4
1097 // CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
1098 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1099 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1100 // CHECK3-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1101 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
1102 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1103 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1104 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1105 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
1106 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1107 // CHECK3:       cond.true:
1108 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1109 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1110 // CHECK3:       cond.false:
1111 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1112 // CHECK3-NEXT:    br label [[COND_END]]
1113 // CHECK3:       cond.end:
1114 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
1115 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1116 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1117 // CHECK3-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
1118 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1119 // CHECK3:       omp.inner.for.cond:
1120 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1121 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1122 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
1123 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1124 // CHECK3:       omp.inner.for.body:
1125 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1126 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
1127 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1128 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
1129 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I3]], align 4
1130 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i32 0, i32 [[TMP18]]
1131 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1132 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1133 // CHECK3:       omp.body.continue:
1134 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1135 // CHECK3:       omp.inner.for.inc:
1136 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1137 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP19]], 1
1138 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
1139 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1140 // CHECK3:       omp.inner.for.end:
1141 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1142 // CHECK3:       omp.loop.exit:
1143 // CHECK3-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1144 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
1145 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
1146 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1147 // CHECK3:       omp.precond.end:
1148 // CHECK3-NEXT:    ret void
1149 //
1150 //
1151 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34
1152 // CHECK3-SAME: (i32 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR3:[0-9]+]] {
1153 // CHECK3-NEXT:  entry:
1154 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1155 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1156 // CHECK3-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
1157 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1158 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1159 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined, ptr [[N_ADDR]], ptr [[TMP0]])
1160 // CHECK3-NEXT:    ret void
1161 //
1162 //
1163 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined
1164 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1165 // CHECK3-NEXT:  entry:
1166 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1167 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1168 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1169 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1170 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1171 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1172 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1173 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1174 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1175 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1176 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1177 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1178 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1179 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
1180 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1181 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1182 // CHECK3-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1183 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1184 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1185 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1186 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
1187 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
1188 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1189 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1190 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1191 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1192 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1193 // CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
1194 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1195 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1196 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1197 // CHECK3:       omp.precond.then:
1198 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1199 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1200 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_COMB_UB]], align 4
1201 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1202 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1203 // CHECK3-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1204 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
1205 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP7]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1206 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1207 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1208 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1209 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1210 // CHECK3:       cond.true:
1211 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1212 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1213 // CHECK3:       cond.false:
1214 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1215 // CHECK3-NEXT:    br label [[COND_END]]
1216 // CHECK3:       cond.end:
1217 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1218 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1219 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1220 // CHECK3-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
1221 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1222 // CHECK3:       omp.inner.for.cond:
1223 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1224 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1225 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1226 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1227 // CHECK3:       omp.inner.for.body:
1228 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1229 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1230 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined, i32 [[TMP15]], i32 [[TMP16]], ptr [[TMP0]], ptr [[TMP1]])
1231 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1232 // CHECK3:       omp.inner.for.inc:
1233 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1234 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1235 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
1236 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1237 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1238 // CHECK3:       omp.inner.for.end:
1239 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1240 // CHECK3:       omp.loop.exit:
1241 // CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1242 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
1243 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP20]])
1244 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1245 // CHECK3:       omp.precond.end:
1246 // CHECK3-NEXT:    ret void
1247 //
1248 //
1249 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l34.omp_outlined.omp_outlined
1250 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1251 // CHECK3-NEXT:  entry:
1252 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1253 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1254 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1255 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1256 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1257 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1258 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1259 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1260 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1261 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1262 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1263 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1264 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1265 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1266 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1267 // CHECK3-NEXT:    [[I3:%.*]] = alloca i32, align 4
1268 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1269 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1270 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1271 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1272 // CHECK3-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1273 // CHECK3-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1274 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1275 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1276 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP0]], align 4
1277 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTCAPTURE_EXPR_]], align 4
1278 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1279 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1280 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1281 // CHECK3-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1282 // CHECK3-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1283 // CHECK3-NEXT:    store i32 0, ptr [[I]], align 4
1284 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1285 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1286 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1287 // CHECK3:       omp.precond.then:
1288 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1289 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1290 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_UB]], align 4
1291 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1292 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1293 // CHECK3-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_LB]], align 4
1294 // CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_UB]], align 4
1295 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1296 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1297 // CHECK3-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1298 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
1299 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP9]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1300 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1301 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1302 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
1303 // CHECK3-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1304 // CHECK3:       cond.true:
1305 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1306 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1307 // CHECK3:       cond.false:
1308 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1309 // CHECK3-NEXT:    br label [[COND_END]]
1310 // CHECK3:       cond.end:
1311 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
1312 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1313 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1314 // CHECK3-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
1315 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1316 // CHECK3:       omp.inner.for.cond:
1317 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1318 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1319 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
1320 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1321 // CHECK3:       omp.inner.for.body:
1322 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1323 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
1324 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1325 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
1326 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I3]], align 4
1327 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], ptr [[TMP1]], i32 0, i32 [[TMP18]]
1328 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1329 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1330 // CHECK3:       omp.body.continue:
1331 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1332 // CHECK3:       omp.inner.for.inc:
1333 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1334 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP19]], 1
1335 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
1336 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
1337 // CHECK3:       omp.inner.for.end:
1338 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1339 // CHECK3:       omp.loop.exit:
1340 // CHECK3-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1341 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
1342 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
1343 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
1344 // CHECK3:       omp.precond.end:
1345 // CHECK3-NEXT:    ret void
1346 //
1347 //
1348 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1349 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
1350 // CHECK3-NEXT:  entry:
1351 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
1352 // CHECK3-NEXT:    ret void
1353 //
1354 //
1355 // CHECK9-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1356 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
1357 // CHECK9-NEXT:  entry:
1358 // CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
1359 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
1360 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1361 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
1362 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
1363 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
1364 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
1365 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1366 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1367 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1368 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1369 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1370 // CHECK9-NEXT:    store i32 100, ptr [[N]], align 4
1371 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
1372 // CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1373 // CHECK9-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
1374 // CHECK9-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
1375 // CHECK9-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1376 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
1377 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
1378 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
1379 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
1380 // CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1381 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 24, i1 false)
1382 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1383 // CHECK9-NEXT:    store i64 [[TMP4]], ptr [[TMP6]], align 8
1384 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1385 // CHECK9-NEXT:    store i64 [[TMP4]], ptr [[TMP7]], align 8
1386 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1387 // CHECK9-NEXT:    store ptr null, ptr [[TMP8]], align 8
1388 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1389 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP9]], align 8
1390 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1391 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP10]], align 8
1392 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1393 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
1394 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1395 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 8
1396 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1397 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 8
1398 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1399 // CHECK9-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 8
1400 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1401 // CHECK9-NEXT:    store ptr null, ptr [[TMP15]], align 8
1402 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1403 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1404 // CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1405 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
1406 // CHECK9-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
1407 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1408 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
1409 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1410 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1411 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1412 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1413 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
1414 // CHECK9-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
1415 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1416 // CHECK9-NEXT:    store i32 2, ptr [[TMP23]], align 4
1417 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1418 // CHECK9-NEXT:    store i32 3, ptr [[TMP24]], align 4
1419 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1420 // CHECK9-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 8
1421 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1422 // CHECK9-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 8
1423 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1424 // CHECK9-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 8
1425 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1426 // CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 8
1427 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1428 // CHECK9-NEXT:    store ptr null, ptr [[TMP29]], align 8
1429 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1430 // CHECK9-NEXT:    store ptr null, ptr [[TMP30]], align 8
1431 // CHECK9-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1432 // CHECK9-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
1433 // CHECK9-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1434 // CHECK9-NEXT:    store i64 0, ptr [[TMP32]], align 8
1435 // CHECK9-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1436 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
1437 // CHECK9-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1438 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
1439 // CHECK9-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1440 // CHECK9-NEXT:    store i32 0, ptr [[TMP35]], align 4
1441 // CHECK9-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.region_id, ptr [[KERNEL_ARGS]])
1442 // CHECK9-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
1443 // CHECK9-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1444 // CHECK9:       omp_offload.failed:
1445 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72(i64 [[TMP4]], i64 [[TMP1]], ptr [[VLA]]) #[[ATTR4:[0-9]+]]
1446 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1447 // CHECK9:       omp_offload.cont:
1448 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[VLA]], i64 0
1449 // CHECK9-NEXT:    [[TMP38:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
1450 // CHECK9-NEXT:    [[TMP39:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
1451 // CHECK9-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP39]])
1452 // CHECK9-NEXT:    ret i32 [[TMP38]]
1453 //
1454 //
1455 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72
1456 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1457 // CHECK9-NEXT:  entry:
1458 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1459 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1460 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1461 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
1462 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1463 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1464 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1465 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1466 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
1467 // CHECK9-NEXT:    ret void
1468 //
1469 //
1470 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined
1471 // 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]+]] {
1472 // CHECK9-NEXT:  entry:
1473 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1474 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1475 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
1476 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1477 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1478 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1479 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1480 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1481 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1482 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1483 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1484 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1485 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1486 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1487 // CHECK9-NEXT:    [[I3:%.*]] = alloca i32, align 4
1488 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1489 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1490 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
1491 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1492 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1493 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
1494 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1495 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1496 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
1497 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1498 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1499 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1500 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1501 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1502 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1503 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1504 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1505 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1506 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1507 // CHECK9:       omp.precond.then:
1508 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1509 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1510 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB]], align 4
1511 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1512 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1513 // CHECK9-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1514 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1515 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1516 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1517 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1518 // CHECK9-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1519 // CHECK9-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1520 // CHECK9:       cond.true:
1521 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1522 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1523 // CHECK9:       cond.false:
1524 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1525 // CHECK9-NEXT:    br label [[COND_END]]
1526 // CHECK9:       cond.end:
1527 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1528 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1529 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1530 // CHECK9-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
1531 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1532 // CHECK9:       omp.inner.for.cond:
1533 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1534 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1535 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1536 // CHECK9-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1537 // CHECK9:       omp.inner.for.body:
1538 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1539 // CHECK9-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
1540 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1541 // CHECK9-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
1542 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined, i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
1543 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1544 // CHECK9:       omp.inner.for.inc:
1545 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1546 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1547 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
1548 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1549 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1550 // CHECK9:       omp.inner.for.end:
1551 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1552 // CHECK9:       omp.loop.exit:
1553 // CHECK9-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1554 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
1555 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP23]])
1556 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1557 // CHECK9:       omp.precond.end:
1558 // CHECK9-NEXT:    ret void
1559 //
1560 //
1561 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined
1562 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3]] {
1563 // CHECK9-NEXT:  entry:
1564 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1565 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1566 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1567 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1568 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
1569 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1570 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1571 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1572 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1573 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1574 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1575 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1576 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1577 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1578 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1579 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1580 // CHECK9-NEXT:    [[I4:%.*]] = alloca i32, align 4
1581 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1582 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1583 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1584 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1585 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
1586 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
1587 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1588 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
1589 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
1590 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1591 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
1592 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1593 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1594 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1595 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1596 // CHECK9-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1597 // CHECK9-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1598 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
1599 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1600 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1601 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1602 // CHECK9:       omp.precond.then:
1603 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1604 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1605 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
1606 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1607 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
1608 // CHECK9-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1609 // CHECK9-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
1610 // CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1611 // CHECK9-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
1612 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1613 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1614 // CHECK9-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1615 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
1616 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1617 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1618 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1619 // CHECK9-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
1620 // CHECK9-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1621 // CHECK9:       cond.true:
1622 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1623 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1624 // CHECK9:       cond.false:
1625 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1626 // CHECK9-NEXT:    br label [[COND_END]]
1627 // CHECK9:       cond.end:
1628 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
1629 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1630 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1631 // CHECK9-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
1632 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1633 // CHECK9:       omp.inner.for.cond:
1634 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1635 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1636 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
1637 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1638 // CHECK9:       omp.inner.for.body:
1639 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1640 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
1641 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1642 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
1643 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4
1644 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
1645 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[IDXPROM]]
1646 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1647 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1648 // CHECK9:       omp.body.continue:
1649 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1650 // CHECK9:       omp.inner.for.inc:
1651 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1652 // CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
1653 // CHECK9-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
1654 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1655 // CHECK9:       omp.inner.for.end:
1656 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1657 // CHECK9:       omp.loop.exit:
1658 // CHECK9-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1659 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
1660 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
1661 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1662 // CHECK9:       omp.precond.end:
1663 // CHECK9-NEXT:    ret void
1664 //
1665 //
1666 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1667 // CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
1668 // CHECK9-NEXT:  entry:
1669 // CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
1670 // CHECK9-NEXT:    ret void
1671 //
1672 //
1673 // CHECK11-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1674 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
1675 // CHECK11-NEXT:  entry:
1676 // CHECK11-NEXT:    [[N:%.*]] = alloca i32, align 4
1677 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
1678 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1679 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
1680 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
1681 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
1682 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
1683 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
1684 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1685 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1686 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1687 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1688 // CHECK11-NEXT:    store i32 100, ptr [[N]], align 4
1689 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
1690 // CHECK11-NEXT:    [[TMP1:%.*]] = call ptr @llvm.stacksave.p0()
1691 // CHECK11-NEXT:    store ptr [[TMP1]], ptr [[SAVED_STACK]], align 4
1692 // CHECK11-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
1693 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[__VLA_EXPR0]], align 4
1694 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N]], align 4
1695 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
1696 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
1697 // CHECK11-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
1698 // CHECK11-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
1699 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 24, i1 false)
1700 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1701 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP6]], align 4
1702 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1703 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[TMP7]], align 4
1704 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1705 // CHECK11-NEXT:    store ptr null, ptr [[TMP8]], align 4
1706 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1707 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP9]], align 4
1708 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1709 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP10]], align 4
1710 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1711 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
1712 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1713 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 4
1714 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1715 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 4
1716 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1717 // CHECK11-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 4
1718 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1719 // CHECK11-NEXT:    store ptr null, ptr [[TMP15]], align 4
1720 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1721 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1722 // CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1723 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
1724 // CHECK11-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
1725 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1726 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
1727 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1728 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1729 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1730 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1731 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
1732 // CHECK11-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
1733 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1734 // CHECK11-NEXT:    store i32 2, ptr [[TMP23]], align 4
1735 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1736 // CHECK11-NEXT:    store i32 3, ptr [[TMP24]], align 4
1737 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1738 // CHECK11-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 4
1739 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1740 // CHECK11-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 4
1741 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1742 // CHECK11-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 4
1743 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1744 // CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 4
1745 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1746 // CHECK11-NEXT:    store ptr null, ptr [[TMP29]], align 4
1747 // CHECK11-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1748 // CHECK11-NEXT:    store ptr null, ptr [[TMP30]], align 4
1749 // CHECK11-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1750 // CHECK11-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
1751 // CHECK11-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1752 // CHECK11-NEXT:    store i64 0, ptr [[TMP32]], align 8
1753 // CHECK11-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1754 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
1755 // CHECK11-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1756 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
1757 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1758 // CHECK11-NEXT:    store i32 0, ptr [[TMP35]], align 4
1759 // CHECK11-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.region_id, ptr [[KERNEL_ARGS]])
1760 // CHECK11-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
1761 // CHECK11-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1762 // CHECK11:       omp_offload.failed:
1763 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72(i32 [[TMP3]], i32 [[TMP0]], ptr [[VLA]]) #[[ATTR4:[0-9]+]]
1764 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1765 // CHECK11:       omp_offload.cont:
1766 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[VLA]], i32 0
1767 // CHECK11-NEXT:    [[TMP38:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
1768 // CHECK11-NEXT:    [[TMP39:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
1769 // CHECK11-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP39]])
1770 // CHECK11-NEXT:    ret i32 [[TMP38]]
1771 //
1772 //
1773 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72
1774 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1775 // CHECK11-NEXT:  entry:
1776 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1777 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1778 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1779 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
1780 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1781 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1782 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1783 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1784 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
1785 // CHECK11-NEXT:    ret void
1786 //
1787 //
1788 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined
1789 // 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]+]] {
1790 // CHECK11-NEXT:  entry:
1791 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1792 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1793 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1794 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1795 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1796 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1797 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1798 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1799 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1800 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1801 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1802 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1803 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1804 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1805 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
1806 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1807 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1808 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1809 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1810 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1811 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1812 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1813 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1814 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
1815 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1816 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1817 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1818 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1819 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1820 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1821 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
1822 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1823 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1824 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1825 // CHECK11:       omp.precond.then:
1826 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1827 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1828 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB]], align 4
1829 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1830 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1831 // CHECK11-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1832 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1833 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1834 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1835 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1836 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1837 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1838 // CHECK11:       cond.true:
1839 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1840 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1841 // CHECK11:       cond.false:
1842 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1843 // CHECK11-NEXT:    br label [[COND_END]]
1844 // CHECK11:       cond.end:
1845 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1846 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1847 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1848 // CHECK11-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
1849 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1850 // CHECK11:       omp.inner.for.cond:
1851 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1852 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1853 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1854 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1855 // CHECK11:       omp.inner.for.body:
1856 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1857 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1858 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined, i32 [[TMP16]], i32 [[TMP17]], ptr [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
1859 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1860 // CHECK11:       omp.inner.for.inc:
1861 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1862 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1863 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
1864 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1865 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1866 // CHECK11:       omp.inner.for.end:
1867 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1868 // CHECK11:       omp.loop.exit:
1869 // CHECK11-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1870 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
1871 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP21]])
1872 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
1873 // CHECK11:       omp.precond.end:
1874 // CHECK11-NEXT:    ret void
1875 //
1876 //
1877 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l72.omp_outlined.omp_outlined
1878 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3]] {
1879 // CHECK11-NEXT:  entry:
1880 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1881 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1882 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1883 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1884 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1885 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1886 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1887 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1888 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1889 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1890 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1891 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1892 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1893 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1894 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1895 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1896 // CHECK11-NEXT:    [[I3:%.*]] = alloca i32, align 4
1897 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1898 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1899 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1900 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1901 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1902 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1903 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1904 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1905 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1906 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1907 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
1908 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1909 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1910 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1911 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1912 // CHECK11-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1913 // CHECK11-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
1914 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
1915 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1916 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1917 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1918 // CHECK11:       omp.precond.then:
1919 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1920 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1921 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
1922 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1923 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1924 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_LB]], align 4
1925 // CHECK11-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_UB]], align 4
1926 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1927 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1928 // CHECK11-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1929 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
1930 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1931 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1932 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1933 // CHECK11-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
1934 // CHECK11-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1935 // CHECK11:       cond.true:
1936 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
1937 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1938 // CHECK11:       cond.false:
1939 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1940 // CHECK11-NEXT:    br label [[COND_END]]
1941 // CHECK11:       cond.end:
1942 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
1943 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1944 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1945 // CHECK11-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
1946 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1947 // CHECK11:       omp.inner.for.cond:
1948 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1949 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1950 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
1951 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1952 // CHECK11:       omp.inner.for.body:
1953 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1954 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
1955 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1956 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
1957 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4
1958 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP19]]
1959 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
1960 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1961 // CHECK11:       omp.body.continue:
1962 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1963 // CHECK11:       omp.inner.for.inc:
1964 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1965 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
1966 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
1967 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1968 // CHECK11:       omp.inner.for.end:
1969 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1970 // CHECK11:       omp.loop.exit:
1971 // CHECK11-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1972 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
1973 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
1974 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
1975 // CHECK11:       omp.precond.end:
1976 // CHECK11-NEXT:    ret void
1977 //
1978 //
1979 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1980 // CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
1981 // CHECK11-NEXT:  entry:
1982 // CHECK11-NEXT:    call void @__tgt_register_requires(i64 1)
1983 // CHECK11-NEXT:    ret void
1984 //
1985 //
1986 // CHECK17-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1987 // CHECK17-SAME: () #[[ATTR0:[0-9]+]] {
1988 // CHECK17-NEXT:  entry:
1989 // CHECK17-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1990 // CHECK17-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(496) [[V]])
1991 // CHECK17-NEXT:    ret i32 [[CALL]]
1992 //
1993 //
1994 // CHECK17-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1995 // CHECK17-SAME: (ptr noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat {
1996 // CHECK17-NEXT:  entry:
1997 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1998 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
1999 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
2000 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
2001 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2002 // CHECK17-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2003 // CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
2004 // CHECK17-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
2005 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
2006 // CHECK17-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2007 // CHECK17-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 8
2008 // CHECK17-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2009 // CHECK17-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
2010 // CHECK17-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2011 // CHECK17-NEXT:    store ptr null, ptr [[TMP2]], align 8
2012 // CHECK17-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2013 // CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2014 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2015 // CHECK17-NEXT:    store i32 2, ptr [[TMP5]], align 4
2016 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2017 // CHECK17-NEXT:    store i32 1, ptr [[TMP6]], align 4
2018 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2019 // CHECK17-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
2020 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2021 // CHECK17-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
2022 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2023 // CHECK17-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 8
2024 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2025 // CHECK17-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 8
2026 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2027 // CHECK17-NEXT:    store ptr null, ptr [[TMP11]], align 8
2028 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2029 // CHECK17-NEXT:    store ptr null, ptr [[TMP12]], align 8
2030 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2031 // CHECK17-NEXT:    store i64 123, ptr [[TMP13]], align 8
2032 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2033 // CHECK17-NEXT:    store i64 0, ptr [[TMP14]], align 8
2034 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2035 // CHECK17-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
2036 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2037 // CHECK17-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
2038 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2039 // CHECK17-NEXT:    store i32 0, ptr [[TMP17]], align 4
2040 // CHECK17-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.region_id, ptr [[KERNEL_ARGS]])
2041 // CHECK17-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2042 // CHECK17-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2043 // CHECK17:       omp_offload.failed:
2044 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108(ptr [[THIS1]]) #[[ATTR3:[0-9]+]]
2045 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2046 // CHECK17:       omp_offload.cont:
2047 // CHECK17-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
2048 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A2]], i64 0, i64 0
2049 // CHECK17-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
2050 // CHECK17-NEXT:    ret i32 [[TMP20]]
2051 //
2052 //
2053 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108
2054 // CHECK17-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2055 // CHECK17-NEXT:  entry:
2056 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
2057 // CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
2058 // CHECK17-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
2059 // CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined, ptr [[TMP0]])
2060 // CHECK17-NEXT:    ret void
2061 //
2062 //
2063 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined
2064 // CHECK17-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
2065 // CHECK17-NEXT:  entry:
2066 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2067 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2068 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
2069 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2070 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2071 // CHECK17-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2072 // CHECK17-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2073 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2074 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2075 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
2076 // CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2077 // CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2078 // CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
2079 // CHECK17-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
2080 // CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2081 // CHECK17-NEXT:    store i32 122, ptr [[DOTOMP_COMB_UB]], align 4
2082 // CHECK17-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2083 // CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2084 // CHECK17-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2085 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2086 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2087 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2088 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2089 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2090 // CHECK17:       cond.true:
2091 // CHECK17-NEXT:    br label [[COND_END:%.*]]
2092 // CHECK17:       cond.false:
2093 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2094 // CHECK17-NEXT:    br label [[COND_END]]
2095 // CHECK17:       cond.end:
2096 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2097 // CHECK17-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2098 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2099 // CHECK17-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2100 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2101 // CHECK17:       omp.inner.for.cond:
2102 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2103 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2104 // CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2105 // CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2106 // CHECK17:       omp.inner.for.body:
2107 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2108 // CHECK17-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2109 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2110 // CHECK17-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2111 // CHECK17-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
2112 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2113 // CHECK17:       omp.inner.for.inc:
2114 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2115 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2116 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2117 // CHECK17-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2118 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
2119 // CHECK17:       omp.inner.for.end:
2120 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2121 // CHECK17:       omp.loop.exit:
2122 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]])
2123 // CHECK17-NEXT:    ret void
2124 //
2125 //
2126 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined
2127 // CHECK17-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2]] {
2128 // CHECK17-NEXT:  entry:
2129 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2130 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2131 // CHECK17-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2132 // CHECK17-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2133 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
2134 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2135 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2136 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2137 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2138 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2139 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2140 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
2141 // CHECK17-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2142 // CHECK17-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2143 // CHECK17-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2144 // CHECK17-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2145 // CHECK17-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
2146 // CHECK17-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
2147 // CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2148 // CHECK17-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
2149 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2150 // CHECK17-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
2151 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2152 // CHECK17-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
2153 // CHECK17-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
2154 // CHECK17-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
2155 // CHECK17-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2156 // CHECK17-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2157 // CHECK17-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2158 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
2159 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2160 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2161 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
2162 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2163 // CHECK17:       cond.true:
2164 // CHECK17-NEXT:    br label [[COND_END:%.*]]
2165 // CHECK17:       cond.false:
2166 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2167 // CHECK17-NEXT:    br label [[COND_END]]
2168 // CHECK17:       cond.end:
2169 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2170 // CHECK17-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2171 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2172 // CHECK17-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
2173 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2174 // CHECK17:       omp.inner.for.cond:
2175 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2176 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2177 // CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2178 // CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2179 // CHECK17:       omp.inner.for.body:
2180 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2181 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2182 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2183 // CHECK17-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2184 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
2185 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
2186 // CHECK17-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
2187 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i64 0, i64 [[IDXPROM]]
2188 // CHECK17-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2189 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2190 // CHECK17:       omp.body.continue:
2191 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2192 // CHECK17:       omp.inner.for.inc:
2193 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2194 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2195 // CHECK17-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
2196 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
2197 // CHECK17:       omp.inner.for.end:
2198 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2199 // CHECK17:       omp.loop.exit:
2200 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
2201 // CHECK17-NEXT:    ret void
2202 //
2203 //
2204 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2205 // CHECK17-SAME: () #[[ATTR4:[0-9]+]] {
2206 // CHECK17-NEXT:  entry:
2207 // CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
2208 // CHECK17-NEXT:    ret void
2209 //
2210 //
2211 // CHECK19-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2212 // CHECK19-SAME: () #[[ATTR0:[0-9]+]] {
2213 // CHECK19-NEXT:  entry:
2214 // CHECK19-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2215 // CHECK19-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(496) [[V]])
2216 // CHECK19-NEXT:    ret i32 [[CALL]]
2217 //
2218 //
2219 // CHECK19-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2220 // CHECK19-SAME: (ptr noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2221 // CHECK19-NEXT:  entry:
2222 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2223 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
2224 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
2225 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
2226 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2227 // CHECK19-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2228 // CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2229 // CHECK19-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2230 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
2231 // CHECK19-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2232 // CHECK19-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 4
2233 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2234 // CHECK19-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
2235 // CHECK19-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2236 // CHECK19-NEXT:    store ptr null, ptr [[TMP2]], align 4
2237 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2238 // CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2239 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2240 // CHECK19-NEXT:    store i32 2, ptr [[TMP5]], align 4
2241 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2242 // CHECK19-NEXT:    store i32 1, ptr [[TMP6]], align 4
2243 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2244 // CHECK19-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
2245 // CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2246 // CHECK19-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
2247 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2248 // CHECK19-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 4
2249 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2250 // CHECK19-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 4
2251 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2252 // CHECK19-NEXT:    store ptr null, ptr [[TMP11]], align 4
2253 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2254 // CHECK19-NEXT:    store ptr null, ptr [[TMP12]], align 4
2255 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2256 // CHECK19-NEXT:    store i64 123, ptr [[TMP13]], align 8
2257 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2258 // CHECK19-NEXT:    store i64 0, ptr [[TMP14]], align 8
2259 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2260 // CHECK19-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
2261 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2262 // CHECK19-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
2263 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2264 // CHECK19-NEXT:    store i32 0, ptr [[TMP17]], align 4
2265 // CHECK19-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.region_id, ptr [[KERNEL_ARGS]])
2266 // CHECK19-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2267 // CHECK19-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2268 // CHECK19:       omp_offload.failed:
2269 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108(ptr [[THIS1]]) #[[ATTR3:[0-9]+]]
2270 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2271 // CHECK19:       omp_offload.cont:
2272 // CHECK19-NEXT:    [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
2273 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A2]], i32 0, i32 0
2274 // CHECK19-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
2275 // CHECK19-NEXT:    ret i32 [[TMP20]]
2276 //
2277 //
2278 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108
2279 // CHECK19-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2280 // CHECK19-NEXT:  entry:
2281 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2282 // CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2283 // CHECK19-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2284 // CHECK19-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined, ptr [[TMP0]])
2285 // CHECK19-NEXT:    ret void
2286 //
2287 //
2288 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined
2289 // CHECK19-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
2290 // CHECK19-NEXT:  entry:
2291 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2292 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2293 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2294 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2295 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2296 // CHECK19-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2297 // CHECK19-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2298 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2299 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2300 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
2301 // CHECK19-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2302 // CHECK19-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2303 // CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2304 // CHECK19-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2305 // CHECK19-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2306 // CHECK19-NEXT:    store i32 122, ptr [[DOTOMP_COMB_UB]], align 4
2307 // CHECK19-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2308 // CHECK19-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2309 // CHECK19-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2310 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2311 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2312 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2313 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2314 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2315 // CHECK19:       cond.true:
2316 // CHECK19-NEXT:    br label [[COND_END:%.*]]
2317 // CHECK19:       cond.false:
2318 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2319 // CHECK19-NEXT:    br label [[COND_END]]
2320 // CHECK19:       cond.end:
2321 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2322 // CHECK19-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2323 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2324 // CHECK19-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2325 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2326 // CHECK19:       omp.inner.for.cond:
2327 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2328 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2329 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2330 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2331 // CHECK19:       omp.inner.for.body:
2332 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2333 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2334 // CHECK19-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
2335 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2336 // CHECK19:       omp.inner.for.inc:
2337 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2338 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2339 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2340 // CHECK19-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2341 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
2342 // CHECK19:       omp.inner.for.end:
2343 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2344 // CHECK19:       omp.loop.exit:
2345 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP2]])
2346 // CHECK19-NEXT:    ret void
2347 //
2348 //
2349 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l108.omp_outlined.omp_outlined
2350 // CHECK19-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR2]] {
2351 // CHECK19-NEXT:  entry:
2352 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2353 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2354 // CHECK19-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2355 // CHECK19-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2356 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2357 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2358 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2359 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2360 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2361 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2362 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2363 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
2364 // CHECK19-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2365 // CHECK19-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2366 // CHECK19-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
2367 // CHECK19-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
2368 // CHECK19-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2369 // CHECK19-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2370 // CHECK19-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2371 // CHECK19-NEXT:    store i32 122, ptr [[DOTOMP_UB]], align 4
2372 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
2373 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
2374 // CHECK19-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
2375 // CHECK19-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
2376 // CHECK19-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2377 // CHECK19-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2378 // CHECK19-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2379 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
2380 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2381 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2382 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
2383 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2384 // CHECK19:       cond.true:
2385 // CHECK19-NEXT:    br label [[COND_END:%.*]]
2386 // CHECK19:       cond.false:
2387 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2388 // CHECK19-NEXT:    br label [[COND_END]]
2389 // CHECK19:       cond.end:
2390 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2391 // CHECK19-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2392 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2393 // CHECK19-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
2394 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2395 // CHECK19:       omp.inner.for.cond:
2396 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2397 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2398 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2399 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2400 // CHECK19:       omp.inner.for.body:
2401 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2402 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2403 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2404 // CHECK19-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2405 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
2406 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
2407 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], ptr [[A]], i32 0, i32 [[TMP11]]
2408 // CHECK19-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2409 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2410 // CHECK19:       omp.body.continue:
2411 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2412 // CHECK19:       omp.inner.for.inc:
2413 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2414 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
2415 // CHECK19-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
2416 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
2417 // CHECK19:       omp.inner.for.end:
2418 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2419 // CHECK19:       omp.loop.exit:
2420 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
2421 // CHECK19-NEXT:    ret void
2422 //
2423 //
2424 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2425 // CHECK19-SAME: () #[[ATTR4:[0-9]+]] {
2426 // CHECK19-NEXT:  entry:
2427 // CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
2428 // CHECK19-NEXT:    ret void
2429 //
2430 //
2431 // CHECK25-LABEL: define {{[^@]+}}@main
2432 // CHECK25-SAME: (i32 noundef signext [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
2433 // CHECK25-NEXT:  entry:
2434 // CHECK25-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
2435 // CHECK25-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
2436 // CHECK25-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
2437 // CHECK25-NEXT:    [[N:%.*]] = alloca i32, align 4
2438 // CHECK25-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
2439 // CHECK25-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
2440 // CHECK25-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
2441 // CHECK25-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
2442 // CHECK25-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
2443 // CHECK25-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
2444 // CHECK25-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
2445 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2446 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2447 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2448 // CHECK25-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2449 // CHECK25-NEXT:    store i32 0, ptr [[RETVAL]], align 4
2450 // CHECK25-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
2451 // CHECK25-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
2452 // CHECK25-NEXT:    store i32 100, ptr [[N]], align 4
2453 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
2454 // CHECK25-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
2455 // CHECK25-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
2456 // CHECK25-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 8
2457 // CHECK25-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
2458 // CHECK25-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
2459 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N]], align 4
2460 // CHECK25-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
2461 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
2462 // CHECK25-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
2463 // CHECK25-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 24, i1 false)
2464 // CHECK25-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2465 // CHECK25-NEXT:    store i64 [[TMP4]], ptr [[TMP6]], align 8
2466 // CHECK25-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2467 // CHECK25-NEXT:    store i64 [[TMP4]], ptr [[TMP7]], align 8
2468 // CHECK25-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2469 // CHECK25-NEXT:    store ptr null, ptr [[TMP8]], align 8
2470 // CHECK25-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2471 // CHECK25-NEXT:    store i64 [[TMP1]], ptr [[TMP9]], align 8
2472 // CHECK25-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2473 // CHECK25-NEXT:    store i64 [[TMP1]], ptr [[TMP10]], align 8
2474 // CHECK25-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2475 // CHECK25-NEXT:    store ptr null, ptr [[TMP11]], align 8
2476 // CHECK25-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2477 // CHECK25-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 8
2478 // CHECK25-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2479 // CHECK25-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 8
2480 // CHECK25-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2481 // CHECK25-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 8
2482 // CHECK25-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2483 // CHECK25-NEXT:    store ptr null, ptr [[TMP15]], align 8
2484 // CHECK25-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2485 // CHECK25-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2486 // CHECK25-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2487 // CHECK25-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
2488 // CHECK25-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
2489 // CHECK25-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2490 // CHECK25-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
2491 // CHECK25-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2492 // CHECK25-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2493 // CHECK25-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2494 // CHECK25-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2495 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
2496 // CHECK25-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
2497 // CHECK25-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2498 // CHECK25-NEXT:    store i32 2, ptr [[TMP23]], align 4
2499 // CHECK25-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2500 // CHECK25-NEXT:    store i32 3, ptr [[TMP24]], align 4
2501 // CHECK25-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2502 // CHECK25-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 8
2503 // CHECK25-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2504 // CHECK25-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 8
2505 // CHECK25-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2506 // CHECK25-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 8
2507 // CHECK25-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2508 // CHECK25-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 8
2509 // CHECK25-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2510 // CHECK25-NEXT:    store ptr null, ptr [[TMP29]], align 8
2511 // CHECK25-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2512 // CHECK25-NEXT:    store ptr null, ptr [[TMP30]], align 8
2513 // CHECK25-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2514 // CHECK25-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
2515 // CHECK25-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2516 // CHECK25-NEXT:    store i64 0, ptr [[TMP32]], align 8
2517 // CHECK25-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2518 // CHECK25-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
2519 // CHECK25-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2520 // CHECK25-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
2521 // CHECK25-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2522 // CHECK25-NEXT:    store i32 0, ptr [[TMP35]], align 4
2523 // CHECK25-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.region_id, ptr [[KERNEL_ARGS]])
2524 // CHECK25-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2525 // CHECK25-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2526 // CHECK25:       omp_offload.failed:
2527 // CHECK25-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161(i64 [[TMP4]], i64 [[TMP1]], ptr [[VLA]]) #[[ATTR4:[0-9]+]]
2528 // CHECK25-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2529 // CHECK25:       omp_offload.cont:
2530 // CHECK25-NEXT:    [[TMP38:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
2531 // CHECK25-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP38]])
2532 // CHECK25-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
2533 // CHECK25-NEXT:    [[TMP39:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
2534 // CHECK25-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP39]])
2535 // CHECK25-NEXT:    [[TMP40:%.*]] = load i32, ptr [[RETVAL]], align 4
2536 // CHECK25-NEXT:    ret i32 [[TMP40]]
2537 //
2538 //
2539 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161
2540 // CHECK25-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
2541 // CHECK25-NEXT:  entry:
2542 // CHECK25-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
2543 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2544 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
2545 // CHECK25-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
2546 // CHECK25-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
2547 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
2548 // CHECK25-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
2549 // CHECK25-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
2550 // CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined, ptr [[N_ADDR]], i64 [[TMP0]], ptr [[TMP1]])
2551 // CHECK25-NEXT:    ret void
2552 //
2553 //
2554 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined
2555 // CHECK25-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]+]] {
2556 // CHECK25-NEXT:  entry:
2557 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2558 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2559 // CHECK25-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
2560 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2561 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
2562 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2563 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2564 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2565 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2566 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
2567 // CHECK25-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2568 // CHECK25-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2569 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2570 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2571 // CHECK25-NEXT:    [[I3:%.*]] = alloca i32, align 4
2572 // CHECK25-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2573 // CHECK25-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2574 // CHECK25-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
2575 // CHECK25-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
2576 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
2577 // CHECK25-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
2578 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
2579 // CHECK25-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
2580 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
2581 // CHECK25-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
2582 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2583 // CHECK25-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2584 // CHECK25-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2585 // CHECK25-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2586 // CHECK25-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2587 // CHECK25-NEXT:    store i32 0, ptr [[I]], align 4
2588 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2589 // CHECK25-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2590 // CHECK25-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2591 // CHECK25:       omp.precond.then:
2592 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2593 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2594 // CHECK25-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB]], align 4
2595 // CHECK25-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2596 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2597 // CHECK25-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2598 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2599 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2600 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2601 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2602 // CHECK25-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2603 // CHECK25-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2604 // CHECK25:       cond.true:
2605 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2606 // CHECK25-NEXT:    br label [[COND_END:%.*]]
2607 // CHECK25:       cond.false:
2608 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2609 // CHECK25-NEXT:    br label [[COND_END]]
2610 // CHECK25:       cond.end:
2611 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2612 // CHECK25-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2613 // CHECK25-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2614 // CHECK25-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
2615 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2616 // CHECK25:       omp.inner.for.cond:
2617 // CHECK25-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2618 // CHECK25-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2619 // CHECK25-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2620 // CHECK25-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2621 // CHECK25:       omp.inner.for.body:
2622 // CHECK25-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2623 // CHECK25-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
2624 // CHECK25-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2625 // CHECK25-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
2626 // CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined, i64 [[TMP17]], i64 [[TMP19]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
2627 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2628 // CHECK25:       omp.inner.for.inc:
2629 // CHECK25-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2630 // CHECK25-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2631 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
2632 // CHECK25-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2633 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
2634 // CHECK25:       omp.inner.for.end:
2635 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2636 // CHECK25:       omp.loop.exit:
2637 // CHECK25-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2638 // CHECK25-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
2639 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP23]])
2640 // CHECK25-NEXT:    br label [[OMP_PRECOND_END]]
2641 // CHECK25:       omp.precond.end:
2642 // CHECK25-NEXT:    ret void
2643 //
2644 //
2645 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined
2646 // CHECK25-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3]] {
2647 // CHECK25-NEXT:  entry:
2648 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2649 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2650 // CHECK25-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2651 // CHECK25-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2652 // CHECK25-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
2653 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2654 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
2655 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2656 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2657 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2658 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2659 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
2660 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2661 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2662 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2663 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2664 // CHECK25-NEXT:    [[I4:%.*]] = alloca i32, align 4
2665 // CHECK25-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2666 // CHECK25-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2667 // CHECK25-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2668 // CHECK25-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2669 // CHECK25-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
2670 // CHECK25-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
2671 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
2672 // CHECK25-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
2673 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
2674 // CHECK25-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
2675 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
2676 // CHECK25-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
2677 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2678 // CHECK25-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2679 // CHECK25-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2680 // CHECK25-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2681 // CHECK25-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
2682 // CHECK25-NEXT:    store i32 0, ptr [[I]], align 4
2683 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
2684 // CHECK25-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2685 // CHECK25-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2686 // CHECK25:       omp.precond.then:
2687 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2688 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2689 // CHECK25-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
2690 // CHECK25-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2691 // CHECK25-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
2692 // CHECK25-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2693 // CHECK25-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
2694 // CHECK25-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
2695 // CHECK25-NEXT:    store i32 [[CONV3]], ptr [[DOTOMP_UB]], align 4
2696 // CHECK25-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2697 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2698 // CHECK25-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2699 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
2700 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2701 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2702 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2703 // CHECK25-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
2704 // CHECK25-NEXT:    br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2705 // CHECK25:       cond.true:
2706 // CHECK25-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
2707 // CHECK25-NEXT:    br label [[COND_END:%.*]]
2708 // CHECK25:       cond.false:
2709 // CHECK25-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2710 // CHECK25-NEXT:    br label [[COND_END]]
2711 // CHECK25:       cond.end:
2712 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
2713 // CHECK25-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2714 // CHECK25-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2715 // CHECK25-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
2716 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2717 // CHECK25:       omp.inner.for.cond:
2718 // CHECK25-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2719 // CHECK25-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2720 // CHECK25-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2721 // CHECK25-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2722 // CHECK25:       omp.inner.for.body:
2723 // CHECK25-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2724 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2725 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2726 // CHECK25-NEXT:    store i32 [[ADD]], ptr [[I4]], align 4
2727 // CHECK25-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I4]], align 4
2728 // CHECK25-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
2729 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[IDXPROM]]
2730 // CHECK25-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2731 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2732 // CHECK25:       omp.body.continue:
2733 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2734 // CHECK25:       omp.inner.for.inc:
2735 // CHECK25-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2736 // CHECK25-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
2737 // CHECK25-NEXT:    store i32 [[ADD7]], ptr [[DOTOMP_IV]], align 4
2738 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
2739 // CHECK25:       omp.inner.for.end:
2740 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2741 // CHECK25:       omp.loop.exit:
2742 // CHECK25-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2743 // CHECK25-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
2744 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
2745 // CHECK25-NEXT:    br label [[OMP_PRECOND_END]]
2746 // CHECK25:       omp.precond.end:
2747 // CHECK25-NEXT:    ret void
2748 //
2749 //
2750 // CHECK25-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
2751 // CHECK25-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR6:[0-9]+]] comdat {
2752 // CHECK25-NEXT:  entry:
2753 // CHECK25-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
2754 // CHECK25-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
2755 // CHECK25-NEXT:    [[TE:%.*]] = alloca i32, align 4
2756 // CHECK25-NEXT:    [[TH:%.*]] = alloca i32, align 4
2757 // CHECK25-NEXT:    [[TE_CASTED:%.*]] = alloca i64, align 8
2758 // CHECK25-NEXT:    [[TH_CASTED:%.*]] = alloca i64, align 8
2759 // CHECK25-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 8
2760 // CHECK25-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 8
2761 // CHECK25-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 8
2762 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2763 // CHECK25-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2764 // CHECK25-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
2765 // CHECK25-NEXT:    store i32 0, ptr [[TE]], align 4
2766 // CHECK25-NEXT:    store i32 128, ptr [[TH]], align 4
2767 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, ptr [[TE]], align 4
2768 // CHECK25-NEXT:    store i32 [[TMP0]], ptr [[TE_CASTED]], align 4
2769 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, ptr [[TE_CASTED]], align 8
2770 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
2771 // CHECK25-NEXT:    store i32 [[TMP2]], ptr [[TH_CASTED]], align 4
2772 // CHECK25-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TH_CASTED]], align 8
2773 // CHECK25-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2774 // CHECK25-NEXT:    store i64 [[TMP1]], ptr [[TMP4]], align 8
2775 // CHECK25-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2776 // CHECK25-NEXT:    store i64 [[TMP1]], ptr [[TMP5]], align 8
2777 // CHECK25-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2778 // CHECK25-NEXT:    store ptr null, ptr [[TMP6]], align 8
2779 // CHECK25-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2780 // CHECK25-NEXT:    store i64 [[TMP3]], ptr [[TMP7]], align 8
2781 // CHECK25-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2782 // CHECK25-NEXT:    store i64 [[TMP3]], ptr [[TMP8]], align 8
2783 // CHECK25-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2784 // CHECK25-NEXT:    store ptr null, ptr [[TMP9]], align 8
2785 // CHECK25-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2786 // CHECK25-NEXT:    store ptr [[A]], ptr [[TMP10]], align 8
2787 // CHECK25-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2788 // CHECK25-NEXT:    store ptr [[A]], ptr [[TMP11]], align 8
2789 // CHECK25-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2790 // CHECK25-NEXT:    store ptr null, ptr [[TMP12]], align 8
2791 // CHECK25-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2792 // CHECK25-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2793 // CHECK25-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TE]], align 4
2794 // CHECK25-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TH]], align 4
2795 // CHECK25-NEXT:    [[TMP17:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP15]], 0
2796 // CHECK25-NEXT:    [[TMP18:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP16]], 0
2797 // CHECK25-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2798 // CHECK25-NEXT:    store i32 2, ptr [[TMP19]], align 4
2799 // CHECK25-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2800 // CHECK25-NEXT:    store i32 3, ptr [[TMP20]], align 4
2801 // CHECK25-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2802 // CHECK25-NEXT:    store ptr [[TMP13]], ptr [[TMP21]], align 8
2803 // CHECK25-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2804 // CHECK25-NEXT:    store ptr [[TMP14]], ptr [[TMP22]], align 8
2805 // CHECK25-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2806 // CHECK25-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP23]], align 8
2807 // CHECK25-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2808 // CHECK25-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP24]], align 8
2809 // CHECK25-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2810 // CHECK25-NEXT:    store ptr null, ptr [[TMP25]], align 8
2811 // CHECK25-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2812 // CHECK25-NEXT:    store ptr null, ptr [[TMP26]], align 8
2813 // CHECK25-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2814 // CHECK25-NEXT:    store i64 10, ptr [[TMP27]], align 8
2815 // CHECK25-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2816 // CHECK25-NEXT:    store i64 0, ptr [[TMP28]], align 8
2817 // CHECK25-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2818 // CHECK25-NEXT:    store [3 x i32] [[TMP17]], ptr [[TMP29]], align 4
2819 // CHECK25-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2820 // CHECK25-NEXT:    store [3 x i32] [[TMP18]], ptr [[TMP30]], align 4
2821 // CHECK25-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2822 // CHECK25-NEXT:    store i32 0, ptr [[TMP31]], align 4
2823 // CHECK25-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 [[TMP15]], i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.region_id, ptr [[KERNEL_ARGS]])
2824 // CHECK25-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
2825 // CHECK25-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2826 // CHECK25:       omp_offload.failed:
2827 // CHECK25-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150(i64 [[TMP1]], i64 [[TMP3]], ptr [[A]]) #[[ATTR4]]
2828 // CHECK25-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2829 // CHECK25:       omp_offload.cont:
2830 // CHECK25-NEXT:    ret i32 0
2831 //
2832 //
2833 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150
2834 // CHECK25-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
2835 // CHECK25-NEXT:  entry:
2836 // CHECK25-NEXT:    [[TE_ADDR:%.*]] = alloca i64, align 8
2837 // CHECK25-NEXT:    [[TH_ADDR:%.*]] = alloca i64, align 8
2838 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
2839 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
2840 // CHECK25-NEXT:    store i64 [[TE]], ptr [[TE_ADDR]], align 8
2841 // CHECK25-NEXT:    store i64 [[TH]], ptr [[TH_ADDR]], align 8
2842 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
2843 // CHECK25-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 8
2844 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
2845 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
2846 // CHECK25-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
2847 // CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined, ptr [[TMP1]])
2848 // CHECK25-NEXT:    ret void
2849 //
2850 //
2851 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined
2852 // CHECK25-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
2853 // CHECK25-NEXT:  entry:
2854 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2855 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2856 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
2857 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2858 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2859 // CHECK25-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2860 // CHECK25-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2861 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2862 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2863 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
2864 // CHECK25-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2865 // CHECK25-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2866 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
2867 // CHECK25-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
2868 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2869 // CHECK25-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
2870 // CHECK25-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2871 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2872 // CHECK25-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2873 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
2874 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2875 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2876 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2877 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2878 // CHECK25:       cond.true:
2879 // CHECK25-NEXT:    br label [[COND_END:%.*]]
2880 // CHECK25:       cond.false:
2881 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2882 // CHECK25-NEXT:    br label [[COND_END]]
2883 // CHECK25:       cond.end:
2884 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2885 // CHECK25-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2886 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2887 // CHECK25-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
2888 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2889 // CHECK25:       omp.inner.for.cond:
2890 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2891 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2892 // CHECK25-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2893 // CHECK25-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2894 // CHECK25:       omp.inner.for.body:
2895 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2896 // CHECK25-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2897 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2898 // CHECK25-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2899 // CHECK25-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
2900 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2901 // CHECK25:       omp.inner.for.inc:
2902 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2903 // CHECK25-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2904 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2905 // CHECK25-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2906 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
2907 // CHECK25:       omp.inner.for.end:
2908 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2909 // CHECK25:       omp.loop.exit:
2910 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP2]])
2911 // CHECK25-NEXT:    ret void
2912 //
2913 //
2914 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined
2915 // CHECK25-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
2916 // CHECK25-NEXT:  entry:
2917 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
2918 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
2919 // CHECK25-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2920 // CHECK25-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2921 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
2922 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2923 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2924 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2925 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2926 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2927 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2928 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
2929 // CHECK25-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
2930 // CHECK25-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
2931 // CHECK25-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2932 // CHECK25-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2933 // CHECK25-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
2934 // CHECK25-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
2935 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2936 // CHECK25-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
2937 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
2938 // CHECK25-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
2939 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
2940 // CHECK25-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
2941 // CHECK25-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
2942 // CHECK25-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
2943 // CHECK25-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2944 // CHECK25-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2945 // CHECK25-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
2946 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
2947 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2948 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2949 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
2950 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2951 // CHECK25:       cond.true:
2952 // CHECK25-NEXT:    br label [[COND_END:%.*]]
2953 // CHECK25:       cond.false:
2954 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2955 // CHECK25-NEXT:    br label [[COND_END]]
2956 // CHECK25:       cond.end:
2957 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2958 // CHECK25-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2959 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2960 // CHECK25-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
2961 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2962 // CHECK25:       omp.inner.for.cond:
2963 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2964 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2965 // CHECK25-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2966 // CHECK25-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2967 // CHECK25:       omp.inner.for.body:
2968 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2969 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2970 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2971 // CHECK25-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2972 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
2973 // CHECK25-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
2974 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
2975 // CHECK25-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
2976 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2977 // CHECK25:       omp.body.continue:
2978 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2979 // CHECK25:       omp.inner.for.inc:
2980 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2981 // CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2982 // CHECK25-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
2983 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
2984 // CHECK25:       omp.inner.for.end:
2985 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2986 // CHECK25:       omp.loop.exit:
2987 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
2988 // CHECK25-NEXT:    ret void
2989 //
2990 //
2991 // CHECK25-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2992 // CHECK25-SAME: () #[[ATTR7:[0-9]+]] {
2993 // CHECK25-NEXT:  entry:
2994 // CHECK25-NEXT:    call void @__tgt_register_requires(i64 1)
2995 // CHECK25-NEXT:    ret void
2996 //
2997 //
2998 // CHECK27-LABEL: define {{[^@]+}}@main
2999 // CHECK27-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3000 // CHECK27-NEXT:  entry:
3001 // CHECK27-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3002 // CHECK27-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
3003 // CHECK27-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 4
3004 // CHECK27-NEXT:    [[N:%.*]] = alloca i32, align 4
3005 // CHECK27-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
3006 // CHECK27-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3007 // CHECK27-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
3008 // CHECK27-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
3009 // CHECK27-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
3010 // CHECK27-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
3011 // CHECK27-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
3012 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3013 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3014 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3015 // CHECK27-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3016 // CHECK27-NEXT:    store i32 0, ptr [[RETVAL]], align 4
3017 // CHECK27-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
3018 // CHECK27-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 4
3019 // CHECK27-NEXT:    store i32 100, ptr [[N]], align 4
3020 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
3021 // CHECK27-NEXT:    [[TMP1:%.*]] = call ptr @llvm.stacksave.p0()
3022 // CHECK27-NEXT:    store ptr [[TMP1]], ptr [[SAVED_STACK]], align 4
3023 // CHECK27-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
3024 // CHECK27-NEXT:    store i32 [[TMP0]], ptr [[__VLA_EXPR0]], align 4
3025 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, ptr [[N]], align 4
3026 // CHECK27-NEXT:    store i32 [[TMP2]], ptr [[N_CASTED]], align 4
3027 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_CASTED]], align 4
3028 // CHECK27-NEXT:    [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
3029 // CHECK27-NEXT:    [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
3030 // CHECK27-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 24, i1 false)
3031 // CHECK27-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3032 // CHECK27-NEXT:    store i32 [[TMP3]], ptr [[TMP6]], align 4
3033 // CHECK27-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3034 // CHECK27-NEXT:    store i32 [[TMP3]], ptr [[TMP7]], align 4
3035 // CHECK27-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3036 // CHECK27-NEXT:    store ptr null, ptr [[TMP8]], align 4
3037 // CHECK27-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3038 // CHECK27-NEXT:    store i32 [[TMP0]], ptr [[TMP9]], align 4
3039 // CHECK27-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3040 // CHECK27-NEXT:    store i32 [[TMP0]], ptr [[TMP10]], align 4
3041 // CHECK27-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3042 // CHECK27-NEXT:    store ptr null, ptr [[TMP11]], align 4
3043 // CHECK27-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3044 // CHECK27-NEXT:    store ptr [[VLA]], ptr [[TMP12]], align 4
3045 // CHECK27-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3046 // CHECK27-NEXT:    store ptr [[VLA]], ptr [[TMP13]], align 4
3047 // CHECK27-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3048 // CHECK27-NEXT:    store i64 [[TMP5]], ptr [[TMP14]], align 4
3049 // CHECK27-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3050 // CHECK27-NEXT:    store ptr null, ptr [[TMP15]], align 4
3051 // CHECK27-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3052 // CHECK27-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3053 // CHECK27-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3054 // CHECK27-NEXT:    [[TMP19:%.*]] = load i32, ptr [[N]], align 4
3055 // CHECK27-NEXT:    store i32 [[TMP19]], ptr [[DOTCAPTURE_EXPR_]], align 4
3056 // CHECK27-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
3057 // CHECK27-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP20]], 0
3058 // CHECK27-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3059 // CHECK27-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3060 // CHECK27-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
3061 // CHECK27-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
3062 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP21]], 1
3063 // CHECK27-NEXT:    [[TMP22:%.*]] = zext i32 [[ADD]] to i64
3064 // CHECK27-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
3065 // CHECK27-NEXT:    store i32 2, ptr [[TMP23]], align 4
3066 // CHECK27-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
3067 // CHECK27-NEXT:    store i32 3, ptr [[TMP24]], align 4
3068 // CHECK27-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
3069 // CHECK27-NEXT:    store ptr [[TMP16]], ptr [[TMP25]], align 4
3070 // CHECK27-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
3071 // CHECK27-NEXT:    store ptr [[TMP17]], ptr [[TMP26]], align 4
3072 // CHECK27-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
3073 // CHECK27-NEXT:    store ptr [[TMP18]], ptr [[TMP27]], align 4
3074 // CHECK27-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
3075 // CHECK27-NEXT:    store ptr @.offload_maptypes, ptr [[TMP28]], align 4
3076 // CHECK27-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
3077 // CHECK27-NEXT:    store ptr null, ptr [[TMP29]], align 4
3078 // CHECK27-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
3079 // CHECK27-NEXT:    store ptr null, ptr [[TMP30]], align 4
3080 // CHECK27-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
3081 // CHECK27-NEXT:    store i64 [[TMP22]], ptr [[TMP31]], align 8
3082 // CHECK27-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
3083 // CHECK27-NEXT:    store i64 0, ptr [[TMP32]], align 8
3084 // CHECK27-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
3085 // CHECK27-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP33]], align 4
3086 // CHECK27-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
3087 // CHECK27-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP34]], align 4
3088 // CHECK27-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
3089 // CHECK27-NEXT:    store i32 0, ptr [[TMP35]], align 4
3090 // CHECK27-NEXT:    [[TMP36:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.region_id, ptr [[KERNEL_ARGS]])
3091 // CHECK27-NEXT:    [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
3092 // CHECK27-NEXT:    br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3093 // CHECK27:       omp_offload.failed:
3094 // CHECK27-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161(i32 [[TMP3]], i32 [[TMP0]], ptr [[VLA]]) #[[ATTR4:[0-9]+]]
3095 // CHECK27-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3096 // CHECK27:       omp_offload.cont:
3097 // CHECK27-NEXT:    [[TMP38:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
3098 // CHECK27-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP38]])
3099 // CHECK27-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
3100 // CHECK27-NEXT:    [[TMP39:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
3101 // CHECK27-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP39]])
3102 // CHECK27-NEXT:    [[TMP40:%.*]] = load i32, ptr [[RETVAL]], align 4
3103 // CHECK27-NEXT:    ret i32 [[TMP40]]
3104 //
3105 //
3106 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161
3107 // CHECK27-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3108 // CHECK27-NEXT:  entry:
3109 // CHECK27-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3110 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3111 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
3112 // CHECK27-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
3113 // CHECK27-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
3114 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
3115 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
3116 // CHECK27-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
3117 // CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined, ptr [[N_ADDR]], i32 [[TMP0]], ptr [[TMP1]])
3118 // CHECK27-NEXT:    ret void
3119 //
3120 //
3121 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined
3122 // CHECK27-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]+]] {
3123 // CHECK27-NEXT:  entry:
3124 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
3125 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
3126 // CHECK27-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
3127 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3128 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
3129 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3130 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3131 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3132 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3133 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
3134 // CHECK27-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3135 // CHECK27-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3136 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3137 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3138 // CHECK27-NEXT:    [[I3:%.*]] = alloca i32, align 4
3139 // CHECK27-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
3140 // CHECK27-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
3141 // CHECK27-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
3142 // CHECK27-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
3143 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
3144 // CHECK27-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
3145 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
3146 // CHECK27-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
3147 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
3148 // CHECK27-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
3149 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
3150 // CHECK27-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3151 // CHECK27-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3152 // CHECK27-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3153 // CHECK27-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
3154 // CHECK27-NEXT:    store i32 0, ptr [[I]], align 4
3155 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
3156 // CHECK27-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3157 // CHECK27-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3158 // CHECK27:       omp.precond.then:
3159 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
3160 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
3161 // CHECK27-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB]], align 4
3162 // CHECK27-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3163 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3164 // CHECK27-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
3165 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
3166 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3167 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3168 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
3169 // CHECK27-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3170 // CHECK27-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3171 // CHECK27:       cond.true:
3172 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
3173 // CHECK27-NEXT:    br label [[COND_END:%.*]]
3174 // CHECK27:       cond.false:
3175 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3176 // CHECK27-NEXT:    br label [[COND_END]]
3177 // CHECK27:       cond.end:
3178 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3179 // CHECK27-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
3180 // CHECK27-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
3181 // CHECK27-NEXT:    store i32 [[TMP13]], ptr [[DOTOMP_IV]], align 4
3182 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3183 // CHECK27:       omp.inner.for.cond:
3184 // CHECK27-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3185 // CHECK27-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3186 // CHECK27-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3187 // CHECK27-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3188 // CHECK27:       omp.inner.for.body:
3189 // CHECK27-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
3190 // CHECK27-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3191 // CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined, i32 [[TMP16]], i32 [[TMP17]], ptr [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
3192 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3193 // CHECK27:       omp.inner.for.inc:
3194 // CHECK27-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3195 // CHECK27-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
3196 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
3197 // CHECK27-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
3198 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
3199 // CHECK27:       omp.inner.for.end:
3200 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3201 // CHECK27:       omp.loop.exit:
3202 // CHECK27-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
3203 // CHECK27-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
3204 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP21]])
3205 // CHECK27-NEXT:    br label [[OMP_PRECOND_END]]
3206 // CHECK27:       omp.precond.end:
3207 // CHECK27-NEXT:    ret void
3208 //
3209 //
3210 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l161.omp_outlined.omp_outlined
3211 // CHECK27-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3]] {
3212 // CHECK27-NEXT:  entry:
3213 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
3214 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
3215 // CHECK27-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3216 // CHECK27-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3217 // CHECK27-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
3218 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3219 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
3220 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3221 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3222 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3223 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3224 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
3225 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3226 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3227 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3228 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3229 // CHECK27-NEXT:    [[I3:%.*]] = alloca i32, align 4
3230 // CHECK27-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
3231 // CHECK27-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
3232 // CHECK27-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
3233 // CHECK27-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
3234 // CHECK27-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
3235 // CHECK27-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
3236 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
3237 // CHECK27-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
3238 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
3239 // CHECK27-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
3240 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP0]], align 4
3241 // CHECK27-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
3242 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
3243 // CHECK27-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3244 // CHECK27-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3245 // CHECK27-NEXT:    [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3246 // CHECK27-NEXT:    store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1]], align 4
3247 // CHECK27-NEXT:    store i32 0, ptr [[I]], align 4
3248 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
3249 // CHECK27-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3250 // CHECK27-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3251 // CHECK27:       omp.precond.then:
3252 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3253 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
3254 // CHECK27-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_UB]], align 4
3255 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
3256 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
3257 // CHECK27-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_LB]], align 4
3258 // CHECK27-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_UB]], align 4
3259 // CHECK27-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3260 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3261 // CHECK27-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
3262 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
3263 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP10]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3264 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3265 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
3266 // CHECK27-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3267 // CHECK27-NEXT:    br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3268 // CHECK27:       cond.true:
3269 // CHECK27-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1]], align 4
3270 // CHECK27-NEXT:    br label [[COND_END:%.*]]
3271 // CHECK27:       cond.false:
3272 // CHECK27-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3273 // CHECK27-NEXT:    br label [[COND_END]]
3274 // CHECK27:       cond.end:
3275 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3276 // CHECK27-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3277 // CHECK27-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3278 // CHECK27-NEXT:    store i32 [[TMP15]], ptr [[DOTOMP_IV]], align 4
3279 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3280 // CHECK27:       omp.inner.for.cond:
3281 // CHECK27-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3282 // CHECK27-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3283 // CHECK27-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
3284 // CHECK27-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3285 // CHECK27:       omp.inner.for.body:
3286 // CHECK27-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3287 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
3288 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3289 // CHECK27-NEXT:    store i32 [[ADD]], ptr [[I3]], align 4
3290 // CHECK27-NEXT:    [[TMP19:%.*]] = load i32, ptr [[I3]], align 4
3291 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP19]]
3292 // CHECK27-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
3293 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3294 // CHECK27:       omp.body.continue:
3295 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3296 // CHECK27:       omp.inner.for.inc:
3297 // CHECK27-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3298 // CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
3299 // CHECK27-NEXT:    store i32 [[ADD6]], ptr [[DOTOMP_IV]], align 4
3300 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
3301 // CHECK27:       omp.inner.for.end:
3302 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3303 // CHECK27:       omp.loop.exit:
3304 // CHECK27-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
3305 // CHECK27-NEXT:    [[TMP22:%.*]] = load i32, ptr [[TMP21]], align 4
3306 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP22]])
3307 // CHECK27-NEXT:    br label [[OMP_PRECOND_END]]
3308 // CHECK27:       omp.precond.end:
3309 // CHECK27-NEXT:    ret void
3310 //
3311 //
3312 // CHECK27-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
3313 // CHECK27-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR6:[0-9]+]] comdat {
3314 // CHECK27-NEXT:  entry:
3315 // CHECK27-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
3316 // CHECK27-NEXT:    [[A:%.*]] = alloca [10 x i32], align 4
3317 // CHECK27-NEXT:    [[TE:%.*]] = alloca i32, align 4
3318 // CHECK27-NEXT:    [[TH:%.*]] = alloca i32, align 4
3319 // CHECK27-NEXT:    [[TE_CASTED:%.*]] = alloca i32, align 4
3320 // CHECK27-NEXT:    [[TH_CASTED:%.*]] = alloca i32, align 4
3321 // CHECK27-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x ptr], align 4
3322 // CHECK27-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x ptr], align 4
3323 // CHECK27-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x ptr], align 4
3324 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3325 // CHECK27-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3326 // CHECK27-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
3327 // CHECK27-NEXT:    store i32 0, ptr [[TE]], align 4
3328 // CHECK27-NEXT:    store i32 128, ptr [[TH]], align 4
3329 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, ptr [[TE]], align 4
3330 // CHECK27-NEXT:    store i32 [[TMP0]], ptr [[TE_CASTED]], align 4
3331 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TE_CASTED]], align 4
3332 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TH]], align 4
3333 // CHECK27-NEXT:    store i32 [[TMP2]], ptr [[TH_CASTED]], align 4
3334 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_CASTED]], align 4
3335 // CHECK27-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3336 // CHECK27-NEXT:    store i32 [[TMP1]], ptr [[TMP4]], align 4
3337 // CHECK27-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3338 // CHECK27-NEXT:    store i32 [[TMP1]], ptr [[TMP5]], align 4
3339 // CHECK27-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3340 // CHECK27-NEXT:    store ptr null, ptr [[TMP6]], align 4
3341 // CHECK27-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3342 // CHECK27-NEXT:    store i32 [[TMP3]], ptr [[TMP7]], align 4
3343 // CHECK27-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3344 // CHECK27-NEXT:    store i32 [[TMP3]], ptr [[TMP8]], align 4
3345 // CHECK27-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3346 // CHECK27-NEXT:    store ptr null, ptr [[TMP9]], align 4
3347 // CHECK27-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3348 // CHECK27-NEXT:    store ptr [[A]], ptr [[TMP10]], align 4
3349 // CHECK27-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3350 // CHECK27-NEXT:    store ptr [[A]], ptr [[TMP11]], align 4
3351 // CHECK27-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3352 // CHECK27-NEXT:    store ptr null, ptr [[TMP12]], align 4
3353 // CHECK27-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3354 // CHECK27-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3355 // CHECK27-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TE]], align 4
3356 // CHECK27-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TH]], align 4
3357 // CHECK27-NEXT:    [[TMP17:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP15]], 0
3358 // CHECK27-NEXT:    [[TMP18:%.*]] = insertvalue [3 x i32] zeroinitializer, i32 [[TMP16]], 0
3359 // CHECK27-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
3360 // CHECK27-NEXT:    store i32 2, ptr [[TMP19]], align 4
3361 // CHECK27-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
3362 // CHECK27-NEXT:    store i32 3, ptr [[TMP20]], align 4
3363 // CHECK27-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
3364 // CHECK27-NEXT:    store ptr [[TMP13]], ptr [[TMP21]], align 4
3365 // CHECK27-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
3366 // CHECK27-NEXT:    store ptr [[TMP14]], ptr [[TMP22]], align 4
3367 // CHECK27-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
3368 // CHECK27-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP23]], align 4
3369 // CHECK27-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
3370 // CHECK27-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP24]], align 4
3371 // CHECK27-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
3372 // CHECK27-NEXT:    store ptr null, ptr [[TMP25]], align 4
3373 // CHECK27-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
3374 // CHECK27-NEXT:    store ptr null, ptr [[TMP26]], align 4
3375 // CHECK27-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
3376 // CHECK27-NEXT:    store i64 10, ptr [[TMP27]], align 8
3377 // CHECK27-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
3378 // CHECK27-NEXT:    store i64 0, ptr [[TMP28]], align 8
3379 // CHECK27-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
3380 // CHECK27-NEXT:    store [3 x i32] [[TMP17]], ptr [[TMP29]], align 4
3381 // CHECK27-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
3382 // CHECK27-NEXT:    store [3 x i32] [[TMP18]], ptr [[TMP30]], align 4
3383 // CHECK27-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
3384 // CHECK27-NEXT:    store i32 0, ptr [[TMP31]], align 4
3385 // CHECK27-NEXT:    [[TMP32:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 [[TMP15]], i32 [[TMP16]], ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.region_id, ptr [[KERNEL_ARGS]])
3386 // CHECK27-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
3387 // CHECK27-NEXT:    br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3388 // CHECK27:       omp_offload.failed:
3389 // CHECK27-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150(i32 [[TMP1]], i32 [[TMP3]], ptr [[A]]) #[[ATTR4]]
3390 // CHECK27-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3391 // CHECK27:       omp_offload.cont:
3392 // CHECK27-NEXT:    ret i32 0
3393 //
3394 //
3395 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150
3396 // CHECK27-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
3397 // CHECK27-NEXT:  entry:
3398 // CHECK27-NEXT:    [[TE_ADDR:%.*]] = alloca i32, align 4
3399 // CHECK27-NEXT:    [[TH_ADDR:%.*]] = alloca i32, align 4
3400 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
3401 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB3]])
3402 // CHECK27-NEXT:    store i32 [[TE]], ptr [[TE_ADDR]], align 4
3403 // CHECK27-NEXT:    store i32 [[TH]], ptr [[TH_ADDR]], align 4
3404 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
3405 // CHECK27-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[A_ADDR]], align 4
3406 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TE_ADDR]], align 4
3407 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TH_ADDR]], align 4
3408 // CHECK27-NEXT:    call void @__kmpc_push_num_teams(ptr @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
3409 // CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined, ptr [[TMP1]])
3410 // CHECK27-NEXT:    ret void
3411 //
3412 //
3413 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined
3414 // CHECK27-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
3415 // CHECK27-NEXT:  entry:
3416 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
3417 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
3418 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
3419 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3420 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3421 // CHECK27-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3422 // CHECK27-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3423 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3424 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3425 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
3426 // CHECK27-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
3427 // CHECK27-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
3428 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
3429 // CHECK27-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
3430 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
3431 // CHECK27-NEXT:    store i32 9, ptr [[DOTOMP_COMB_UB]], align 4
3432 // CHECK27-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3433 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3434 // CHECK27-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
3435 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
3436 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3437 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3438 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3439 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3440 // CHECK27:       cond.true:
3441 // CHECK27-NEXT:    br label [[COND_END:%.*]]
3442 // CHECK27:       cond.false:
3443 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3444 // CHECK27-NEXT:    br label [[COND_END]]
3445 // CHECK27:       cond.end:
3446 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3447 // CHECK27-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
3448 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
3449 // CHECK27-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
3450 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3451 // CHECK27:       omp.inner.for.cond:
3452 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3453 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3454 // CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3455 // CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3456 // CHECK27:       omp.inner.for.body:
3457 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
3458 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
3459 // CHECK27-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
3460 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3461 // CHECK27:       omp.inner.for.inc:
3462 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3463 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
3464 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
3465 // CHECK27-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
3466 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
3467 // CHECK27:       omp.inner.for.end:
3468 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3469 // CHECK27:       omp.loop.exit:
3470 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP2]])
3471 // CHECK27-NEXT:    ret void
3472 //
3473 //
3474 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l150.omp_outlined.omp_outlined
3475 // CHECK27-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR3]] {
3476 // CHECK27-NEXT:  entry:
3477 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
3478 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
3479 // CHECK27-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3480 // CHECK27-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3481 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
3482 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3483 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3484 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3485 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3486 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3487 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3488 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
3489 // CHECK27-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
3490 // CHECK27-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
3491 // CHECK27-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
3492 // CHECK27-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
3493 // CHECK27-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
3494 // CHECK27-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
3495 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
3496 // CHECK27-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
3497 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
3498 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
3499 // CHECK27-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
3500 // CHECK27-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
3501 // CHECK27-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
3502 // CHECK27-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
3503 // CHECK27-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
3504 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
3505 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
3506 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3507 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
3508 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3509 // CHECK27:       cond.true:
3510 // CHECK27-NEXT:    br label [[COND_END:%.*]]
3511 // CHECK27:       cond.false:
3512 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3513 // CHECK27-NEXT:    br label [[COND_END]]
3514 // CHECK27:       cond.end:
3515 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3516 // CHECK27-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
3517 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
3518 // CHECK27-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
3519 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3520 // CHECK27:       omp.inner.for.cond:
3521 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3522 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
3523 // CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3524 // CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3525 // CHECK27:       omp.inner.for.body:
3526 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3527 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3528 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3529 // CHECK27-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
3530 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
3531 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], ptr [[TMP0]], i32 0, i32 [[TMP11]]
3532 // CHECK27-NEXT:    store i32 0, ptr [[ARRAYIDX]], align 4
3533 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3534 // CHECK27:       omp.body.continue:
3535 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3536 // CHECK27:       omp.inner.for.inc:
3537 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
3538 // CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
3539 // CHECK27-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
3540 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
3541 // CHECK27:       omp.inner.for.end:
3542 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3543 // CHECK27:       omp.loop.exit:
3544 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
3545 // CHECK27-NEXT:    ret void
3546 //
3547 //
3548 // CHECK27-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3549 // CHECK27-SAME: () #[[ATTR7:[0-9]+]] {
3550 // CHECK27-NEXT:  entry:
3551 // CHECK27-NEXT:    call void @__tgt_register_requires(i64 1)
3552 // CHECK27-NEXT:    ret void
3553 //