clang/test/OpenMP/teams_generic_loop_collapse_codegen.cpp

   1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
   2 // expected-no-diagnostics
   3 #ifndef HEADER
   4 #define HEADER
   5
   6 // Test host codegen.
   7 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
   8 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
   9 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
  10 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
  11 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  12 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
  13
  14 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  15 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  16 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  17 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  18 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  19 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  20 #ifdef CK1
  21
  22 template <typename T, int X, long long Y>
  23 struct SS{
  24   T a[X][Y];
  25
  26   int foo(void) {
  27
  28     #pragma omp target
  29     #pragma omp teams loop collapse(2)
  30     for(int i = 0; i < X; i++) {
  31       for(int j = 0; j < Y; j++) {
  32         a[i][j] = (T)0;
  33       }
  34     }
  35
  36     // discard loop variables not needed here
  37
  38
  39     return a[0][0];
  40   }
  41 };
  42
  43 int teams_template_struct(void) {
  44   SS<int, 123, 456> V;
  45   return V.foo();
  46
  47 }
  48 #endif // CK1
  49
  50 // Test host codegen.
  51 // RUN: %clang_cc1 -DCK2 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
  52 // 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
  53 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
  54 // RUN: %clang_cc1 -DCK2 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
  55 // 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
  56 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
  57
  58 // RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  59 // 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
  60 // 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 %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  61 // RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  62 // 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
  63 // 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 %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  64 #ifdef CK2
  65
  66 template <typename T, int n, int m>
  67 int tmain(T argc) {
  68   T a[n][m];
  69   #pragma omp target
  70   #pragma omp teams loop collapse(2)
  71   for(int i = 0; i < n; i++) {
  72     for(int j = 0; j < m; j++) {
  73       a[i][j] = (T)0;
  74     }
  75   }
  76   return 0;
  77 }
  78
  79 int main (int argc, char **argv) {
  80   int n = 100;
  81   int m = 2;
  82   int a[n][m];
  83   #pragma omp target
  84   #pragma omp teams loop collapse(2)
  85   for(int i = 0; i < n; i++) {
  86     for(int j = 0; j < m; j++) {
  87       a[i][j] = 0;
  88     }
  89   }
  90   return tmain<int, 10, 2>(argc);
  91 }
  92
  93
  94
  95
  96
  97
  98
  99
 100 // discard loop variables not needed here
 101
 102
 103 #endif // CK2
 104 #endif // #ifndef HEADER
 105 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 106 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 107 // CHECK1-NEXT:  entry:
 108 // CHECK1-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 109 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(224352) [[V]])
 110 // CHECK1-NEXT:    ret i32 [[CALL]]
 111 //
 112 //
 113 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 114 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat {
 115 // CHECK1-NEXT:  entry:
 116 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 117 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 118 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 119 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 120 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 121 // CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 122 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 123 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 124 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 125 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 126 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 127 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 8
 128 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 129 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
 130 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 131 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 132 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 133 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 134 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 135 // CHECK1-NEXT:    store i32 3, ptr [[TMP5]], align 4
 136 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 137 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
 138 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 139 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 140 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 141 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 142 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 143 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 8
 144 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 145 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 8
 146 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 147 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 148 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 149 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 150 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 151 // CHECK1-NEXT:    store i64 56088, ptr [[TMP13]], align 8
 152 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 153 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
 154 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 155 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 156 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 157 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 158 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 159 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
 160 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 161 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 162 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 163 // CHECK1:       omp_offload.failed:
 164 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 165 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 166 // CHECK1:       omp_offload.cont:
 167 // CHECK1-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 168 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A3]], i64 0, i64 0
 169 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 0
 170 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
 171 // CHECK1-NEXT:    ret i32 [[TMP20]]
 172 //
 173 //
 174 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 175 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 176 // CHECK1-NEXT:  entry:
 177 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 178 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 179 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 180 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 181 // CHECK1-NEXT:    ret void
 182 //
 183 //
 184 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 185 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 186 // CHECK1-NEXT:  entry:
 187 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 188 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 189 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 190 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 191 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 192 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 193 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 194 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 195 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 196 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 197 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 198 // CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
 199 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 200 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 201 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 202 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 203 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 204 // CHECK1-NEXT:    store i32 56087, ptr [[DOTOMP_COMB_UB]], align 4
 205 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 206 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 207 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 208 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 209 // CHECK1-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)
 210 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 211 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
 212 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 213 // CHECK1:       cond.true:
 214 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 215 // CHECK1:       cond.false:
 216 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 217 // CHECK1-NEXT:    br label [[COND_END]]
 218 // CHECK1:       cond.end:
 219 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 220 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 221 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 222 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 223 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 224 // CHECK1:       omp.inner.for.cond:
 225 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 226 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 227 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 228 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 229 // CHECK1:       omp.inner.for.body:
 230 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 231 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 456
 232 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 233 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 234 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 235 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 236 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 237 // CHECK1-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 456
 238 // CHECK1-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
 239 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
 240 // CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 241 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
 242 // CHECK1-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
 243 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 244 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
 245 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
 246 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i64 0, i64 [[IDXPROM]]
 247 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
 248 // CHECK1-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP12]] to i64
 249 // CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM7]]
 250 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX8]], align 4
 251 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 252 // CHECK1:       omp.body.continue:
 253 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 254 // CHECK1:       omp.inner.for.inc:
 255 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 256 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP13]], 1
 257 // CHECK1-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
 258 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 259 // CHECK1:       omp.inner.for.end:
 260 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 261 // CHECK1:       omp.loop.exit:
 262 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 263 // CHECK1-NEXT:    ret void
 264 //
 265 //
 266 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 267 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 268 // CHECK3-NEXT:  entry:
 269 // CHECK3-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 270 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(224352) [[V]])
 271 // CHECK3-NEXT:    ret i32 [[CALL]]
 272 //
 273 //
 274 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 275 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
 276 // CHECK3-NEXT:  entry:
 277 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 278 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
 279 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
 280 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
 281 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 282 // CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 283 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 284 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 285 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 286 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 287 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 288 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 4
 289 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 290 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
 291 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 292 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
 293 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 294 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 295 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 296 // CHECK3-NEXT:    store i32 3, ptr [[TMP5]], align 4
 297 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 298 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
 299 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 300 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
 301 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 302 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
 303 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 304 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 4
 305 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 306 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 4
 307 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 308 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
 309 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 310 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 311 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 312 // CHECK3-NEXT:    store i64 56088, ptr [[TMP13]], align 8
 313 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 314 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
 315 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 316 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 317 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 318 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 319 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 320 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
 321 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 322 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 323 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 324 // CHECK3:       omp_offload.failed:
 325 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 326 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 327 // CHECK3:       omp_offload.cont:
 328 // CHECK3-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 329 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A3]], i32 0, i32 0
 330 // CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 0
 331 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
 332 // CHECK3-NEXT:    ret i32 [[TMP20]]
 333 //
 334 //
 335 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 336 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 337 // CHECK3-NEXT:  entry:
 338 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 339 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 340 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 341 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 342 // CHECK3-NEXT:    ret void
 343 //
 344 //
 345 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 346 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 347 // CHECK3-NEXT:  entry:
 348 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 349 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 350 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 351 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 352 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 353 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 354 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 355 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 356 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 357 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 358 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 359 // CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
 360 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 361 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 362 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 363 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 364 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 365 // CHECK3-NEXT:    store i32 56087, ptr [[DOTOMP_COMB_UB]], align 4
 366 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 367 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 368 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 369 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 370 // CHECK3-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)
 371 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 372 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
 373 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 374 // CHECK3:       cond.true:
 375 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 376 // CHECK3:       cond.false:
 377 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 378 // CHECK3-NEXT:    br label [[COND_END]]
 379 // CHECK3:       cond.end:
 380 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 381 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 382 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 383 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 384 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 385 // CHECK3:       omp.inner.for.cond:
 386 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 387 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 388 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 389 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 390 // CHECK3:       omp.inner.for.body:
 391 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 392 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 456
 393 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 394 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 395 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 396 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 397 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 398 // CHECK3-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 456
 399 // CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
 400 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
 401 // CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 402 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
 403 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
 404 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 405 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
 406 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i32 0, i32 [[TMP11]]
 407 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
 408 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP12]]
 409 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
 410 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 411 // CHECK3:       omp.body.continue:
 412 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 413 // CHECK3:       omp.inner.for.inc:
 414 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 415 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
 416 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
 417 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 418 // CHECK3:       omp.inner.for.end:
 419 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 420 // CHECK3:       omp.loop.exit:
 421 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 422 // CHECK3-NEXT:    ret void
 423 //
 424 //
 425 // CHECK9-LABEL: define {{[^@]+}}@main
 426 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 427 // CHECK9-NEXT:  entry:
 428 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 429 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 430 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
 431 // CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
 432 // CHECK9-NEXT:    [[M:%.*]] = alloca i32, align 4
 433 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
 434 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
 435 // CHECK9-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
 436 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 437 // CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
 438 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 8
 439 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 8
 440 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 8
 441 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
 442 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 443 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 444 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 445 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 446 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
 447 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 448 // CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 449 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 450 // CHECK9-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
 451 // CHECK9-NEXT:    store i32 100, ptr [[N]], align 4
 452 // CHECK9-NEXT:    store i32 2, ptr [[M]], align 4
 453 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 454 // CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
 455 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[M]], align 4
 456 // CHECK9-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
 457 // CHECK9-NEXT:    [[TMP4:%.*]] = call ptr @llvm.stacksave.p0()
 458 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[SAVED_STACK]], align 8
 459 // CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
 460 // CHECK9-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 4
 461 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
 462 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[__VLA_EXPR1]], align 8
 463 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N]], align 4
 464 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED]], align 4
 465 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED]], align 8
 466 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[M]], align 4
 467 // CHECK9-NEXT:    store i32 [[TMP8]], ptr [[M_CASTED]], align 4
 468 // CHECK9-NEXT:    [[TMP9:%.*]] = load i64, ptr [[M_CASTED]], align 8
 469 // CHECK9-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
 470 // CHECK9-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
 471 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 40, i1 false)
 472 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 473 // CHECK9-NEXT:    store i64 [[TMP7]], ptr [[TMP12]], align 8
 474 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 475 // CHECK9-NEXT:    store i64 [[TMP7]], ptr [[TMP13]], align 8
 476 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 477 // CHECK9-NEXT:    store ptr null, ptr [[TMP14]], align 8
 478 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 479 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[TMP15]], align 8
 480 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 481 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[TMP16]], align 8
 482 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
 483 // CHECK9-NEXT:    store ptr null, ptr [[TMP17]], align 8
 484 // CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 485 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP18]], align 8
 486 // CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 487 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP19]], align 8
 488 // CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
 489 // CHECK9-NEXT:    store ptr null, ptr [[TMP20]], align 8
 490 // CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 491 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP21]], align 8
 492 // CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 493 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP22]], align 8
 494 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
 495 // CHECK9-NEXT:    store ptr null, ptr [[TMP23]], align 8
 496 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 497 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP24]], align 8
 498 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 499 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP25]], align 8
 500 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 4
 501 // CHECK9-NEXT:    store i64 [[TMP11]], ptr [[TMP26]], align 8
 502 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
 503 // CHECK9-NEXT:    store ptr null, ptr [[TMP27]], align 8
 504 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 505 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 506 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
 507 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, ptr [[N]], align 4
 508 // CHECK9-NEXT:    store i32 [[TMP31]], ptr [[DOTCAPTURE_EXPR_]], align 4
 509 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, ptr [[M]], align 4
 510 // CHECK9-NEXT:    store i32 [[TMP32]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 511 // CHECK9-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 512 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP33]], 0
 513 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 514 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 515 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 516 // CHECK9-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[TMP34]], 0
 517 // CHECK9-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
 518 // CHECK9-NEXT:    [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
 519 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
 520 // CHECK9-NEXT:    [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
 521 // CHECK9-NEXT:    store i64 [[SUB7]], ptr [[DOTCAPTURE_EXPR_3]], align 8
 522 // CHECK9-NEXT:    [[TMP35:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8
 523 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP35]], 1
 524 // CHECK9-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 525 // CHECK9-NEXT:    store i32 3, ptr [[TMP36]], align 4
 526 // CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 527 // CHECK9-NEXT:    store i32 5, ptr [[TMP37]], align 4
 528 // CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 529 // CHECK9-NEXT:    store ptr [[TMP28]], ptr [[TMP38]], align 8
 530 // CHECK9-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 531 // CHECK9-NEXT:    store ptr [[TMP29]], ptr [[TMP39]], align 8
 532 // CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 533 // CHECK9-NEXT:    store ptr [[TMP30]], ptr [[TMP40]], align 8
 534 // CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 535 // CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP41]], align 8
 536 // CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 537 // CHECK9-NEXT:    store ptr null, ptr [[TMP42]], align 8
 538 // CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 539 // CHECK9-NEXT:    store ptr null, ptr [[TMP43]], align 8
 540 // CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 541 // CHECK9-NEXT:    store i64 [[ADD]], ptr [[TMP44]], align 8
 542 // CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 543 // CHECK9-NEXT:    store i64 0, ptr [[TMP45]], align 8
 544 // CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 545 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP46]], align 4
 546 // CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 547 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4
 548 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 549 // CHECK9-NEXT:    store i32 0, ptr [[TMP48]], align 4
 550 // CHECK9-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.region_id, ptr [[KERNEL_ARGS]])
 551 // CHECK9-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
 552 // CHECK9-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 553 // CHECK9:       omp_offload.failed:
 554 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83(i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP1]], i64 [[TMP3]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
 555 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 556 // CHECK9:       omp_offload.cont:
 557 // CHECK9-NEXT:    [[TMP51:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
 558 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef signext [[TMP51]])
 559 // CHECK9-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
 560 // CHECK9-NEXT:    [[TMP52:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
 561 // CHECK9-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP52]])
 562 // CHECK9-NEXT:    [[TMP53:%.*]] = load i32, ptr [[RETVAL]], align 4
 563 // CHECK9-NEXT:    ret i32 [[TMP53]]
 564 //
 565 //
 566 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83
 567 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
 568 // CHECK9-NEXT:  entry:
 569 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 570 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
 571 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 572 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 573 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 574 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 575 // CHECK9-NEXT:    store i64 [[M]], ptr [[M_ADDR]], align 8
 576 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 577 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 578 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 579 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 580 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 581 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 582 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i64 [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
 583 // CHECK9-NEXT:    ret void
 584 //
 585 //
 586 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined
 587 // CHECK9-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(4) [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
 588 // CHECK9-NEXT:  entry:
 589 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 590 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 591 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
 592 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 8
 593 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 594 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 595 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 596 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 597 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 598 // CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
 599 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 600 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
 601 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
 602 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 603 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
 604 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
 605 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
 606 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 607 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 608 // CHECK9-NEXT:    [[I11:%.*]] = alloca i32, align 4
 609 // CHECK9-NEXT:    [[J12:%.*]] = alloca i32, align 4
 610 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 611 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 612 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
 613 // CHECK9-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 8
 614 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 615 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 616 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 617 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
 618 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 8
 619 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 620 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 621 // CHECK9-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 622 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
 623 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
 624 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
 625 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
 626 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 627 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
 628 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 629 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 630 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 631 // CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
 632 // CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 633 // CHECK9-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
 634 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
 635 // CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
 636 // CHECK9-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
 637 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
 638 // CHECK9-NEXT:    store i32 0, ptr [[J]], align 4
 639 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 640 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
 641 // CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
 642 // CHECK9:       land.lhs.true:
 643 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 644 // CHECK9-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
 645 // CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
 646 // CHECK9:       omp.precond.then:
 647 // CHECK9-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
 648 // CHECK9-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 649 // CHECK9-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_COMB_UB]], align 8
 650 // CHECK9-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
 651 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 652 // CHECK9-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 653 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
 654 // CHECK9-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP13]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
 655 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 656 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 657 // CHECK9-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
 658 // CHECK9-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 659 // CHECK9:       cond.true:
 660 // CHECK9-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 661 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 662 // CHECK9:       cond.false:
 663 // CHECK9-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 664 // CHECK9-NEXT:    br label [[COND_END]]
 665 // CHECK9:       cond.end:
 666 // CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
 667 // CHECK9-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
 668 // CHECK9-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
 669 // CHECK9-NEXT:    store i64 [[TMP18]], ptr [[DOTOMP_IV]], align 8
 670 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 671 // CHECK9:       omp.inner.for.cond:
 672 // CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 673 // CHECK9-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 674 // CHECK9-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
 675 // CHECK9-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 676 // CHECK9:       omp.inner.for.body:
 677 // CHECK9-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 678 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 679 // CHECK9-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP22]], 0
 680 // CHECK9-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
 681 // CHECK9-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
 682 // CHECK9-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
 683 // CHECK9-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP21]], [[CONV18]]
 684 // CHECK9-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
 685 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
 686 // CHECK9-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
 687 // CHECK9-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
 688 // CHECK9-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 689 // CHECK9-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 690 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 691 // CHECK9-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP25]], 0
 692 // CHECK9-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
 693 // CHECK9-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
 694 // CHECK9-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
 695 // CHECK9-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP24]], [[CONV25]]
 696 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 697 // CHECK9-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP26]], 0
 698 // CHECK9-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
 699 // CHECK9-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
 700 // CHECK9-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
 701 // CHECK9-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
 702 // CHECK9-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP23]], [[MUL31]]
 703 // CHECK9-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
 704 // CHECK9-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
 705 // CHECK9-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
 706 // CHECK9-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
 707 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[I11]], align 4
 708 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP27]] to i64
 709 // CHECK9-NEXT:    [[TMP28:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP3]]
 710 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i64 [[TMP28]]
 711 // CHECK9-NEXT:    [[TMP29:%.*]] = load i32, ptr [[J12]], align 4
 712 // CHECK9-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP29]] to i64
 713 // CHECK9-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i64 [[IDXPROM36]]
 714 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX37]], align 4
 715 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 716 // CHECK9:       omp.body.continue:
 717 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 718 // CHECK9:       omp.inner.for.inc:
 719 // CHECK9-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 720 // CHECK9-NEXT:    [[ADD38:%.*]] = add nsw i64 [[TMP30]], 1
 721 // CHECK9-NEXT:    store i64 [[ADD38]], ptr [[DOTOMP_IV]], align 8
 722 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 723 // CHECK9:       omp.inner.for.end:
 724 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 725 // CHECK9:       omp.loop.exit:
 726 // CHECK9-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 727 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
 728 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP32]])
 729 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 730 // CHECK9:       omp.precond.end:
 731 // CHECK9-NEXT:    ret void
 732 //
 733 //
 734 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
 735 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
 736 // CHECK9-NEXT:  entry:
 737 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 738 // CHECK9-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
 739 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 740 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 741 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 742 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 743 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 744 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 745 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 746 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 747 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP0]], align 8
 748 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 749 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
 750 // CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 751 // CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
 752 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 753 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 754 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 755 // CHECK9-NEXT:    store i32 3, ptr [[TMP5]], align 4
 756 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 757 // CHECK9-NEXT:    store i32 1, ptr [[TMP6]], align 4
 758 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 759 // CHECK9-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 760 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 761 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 762 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 763 // CHECK9-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP9]], align 8
 764 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 765 // CHECK9-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP10]], align 8
 766 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 767 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
 768 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 769 // CHECK9-NEXT:    store ptr null, ptr [[TMP12]], align 8
 770 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 771 // CHECK9-NEXT:    store i64 20, ptr [[TMP13]], align 8
 772 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 773 // CHECK9-NEXT:    store i64 0, ptr [[TMP14]], align 8
 774 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 775 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 776 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 777 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 778 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 779 // CHECK9-NEXT:    store i32 0, ptr [[TMP17]], align 4
 780 // CHECK9-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.region_id, ptr [[KERNEL_ARGS]])
 781 // CHECK9-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 782 // CHECK9-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 783 // CHECK9:       omp_offload.failed:
 784 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69(ptr [[A]]) #[[ATTR3]]
 785 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 786 // CHECK9:       omp_offload.cont:
 787 // CHECK9-NEXT:    ret i32 0
 788 //
 789 //
 790 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69
 791 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
 792 // CHECK9-NEXT:  entry:
 793 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 794 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 795 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 796 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
 797 // CHECK9-NEXT:    ret void
 798 //
 799 //
 800 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined
 801 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
 802 // CHECK9-NEXT:  entry:
 803 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 804 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 805 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 806 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 807 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 808 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 809 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 810 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 811 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 812 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 813 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 814 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
 815 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 816 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 817 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 818 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 819 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 820 // CHECK9-NEXT:    store i32 19, ptr [[DOTOMP_COMB_UB]], align 4
 821 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 822 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 823 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 824 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 825 // CHECK9-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)
 826 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 827 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
 828 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 829 // CHECK9:       cond.true:
 830 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 831 // CHECK9:       cond.false:
 832 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 833 // CHECK9-NEXT:    br label [[COND_END]]
 834 // CHECK9:       cond.end:
 835 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 836 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 837 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 838 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 839 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 840 // CHECK9:       omp.inner.for.cond:
 841 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 842 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 843 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 844 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 845 // CHECK9:       omp.inner.for.body:
 846 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 847 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 2
 848 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 849 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 850 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 851 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 852 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 853 // CHECK9-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 2
 854 // CHECK9-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
 855 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
 856 // CHECK9-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 857 // CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
 858 // CHECK9-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
 859 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
 860 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
 861 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
 862 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
 863 // CHECK9-NEXT:    [[IDXPROM7:%.*]] = sext i32 [[TMP12]] to i64
 864 // CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM7]]
 865 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX8]], align 4
 866 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 867 // CHECK9:       omp.body.continue:
 868 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 869 // CHECK9:       omp.inner.for.inc:
 870 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 871 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP13]], 1
 872 // CHECK9-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
 873 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 874 // CHECK9:       omp.inner.for.end:
 875 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 876 // CHECK9:       omp.loop.exit:
 877 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 878 // CHECK9-NEXT:    ret void
 879 //
 880 //
 881 // CHECK11-LABEL: define {{[^@]+}}@main
 882 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 883 // CHECK11-NEXT:  entry:
 884 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 885 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 886 // CHECK11-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 4
 887 // CHECK11-NEXT:    [[N:%.*]] = alloca i32, align 4
 888 // CHECK11-NEXT:    [[M:%.*]] = alloca i32, align 4
 889 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
 890 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
 891 // CHECK11-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
 892 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
 893 // CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
 894 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 4
 895 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 4
 896 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 4
 897 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
 898 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 899 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 900 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 901 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 902 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
 903 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 904 // CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 905 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 906 // CHECK11-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 4
 907 // CHECK11-NEXT:    store i32 100, ptr [[N]], align 4
 908 // CHECK11-NEXT:    store i32 2, ptr [[M]], align 4
 909 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 910 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[M]], align 4
 911 // CHECK11-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
 912 // CHECK11-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 4
 913 // CHECK11-NEXT:    [[TMP3:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
 914 // CHECK11-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP3]], align 4
 915 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[__VLA_EXPR0]], align 4
 916 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[__VLA_EXPR1]], align 4
 917 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N]], align 4
 918 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
 919 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_CASTED]], align 4
 920 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[M]], align 4
 921 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[M_CASTED]], align 4
 922 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[M_CASTED]], align 4
 923 // CHECK11-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
 924 // CHECK11-NEXT:    [[TMP9:%.*]] = mul nuw i32 [[TMP8]], 4
 925 // CHECK11-NEXT:    [[TMP10:%.*]] = sext i32 [[TMP9]] to i64
 926 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 40, i1 false)
 927 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 928 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP11]], align 4
 929 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 930 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP12]], align 4
 931 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 932 // CHECK11-NEXT:    store ptr null, ptr [[TMP13]], align 4
 933 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 934 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[TMP14]], align 4
 935 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 936 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[TMP15]], align 4
 937 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
 938 // CHECK11-NEXT:    store ptr null, ptr [[TMP16]], align 4
 939 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 940 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP17]], align 4
 941 // CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 942 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP18]], align 4
 943 // CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
 944 // CHECK11-NEXT:    store ptr null, ptr [[TMP19]], align 4
 945 // CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 946 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP20]], align 4
 947 // CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 948 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP21]], align 4
 949 // CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
 950 // CHECK11-NEXT:    store ptr null, ptr [[TMP22]], align 4
 951 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 952 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP23]], align 4
 953 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 954 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP24]], align 4
 955 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 4
 956 // CHECK11-NEXT:    store i64 [[TMP10]], ptr [[TMP25]], align 4
 957 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
 958 // CHECK11-NEXT:    store ptr null, ptr [[TMP26]], align 4
 959 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 960 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 961 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
 962 // CHECK11-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N]], align 4
 963 // CHECK11-NEXT:    store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR_]], align 4
 964 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32, ptr [[M]], align 4
 965 // CHECK11-NEXT:    store i32 [[TMP31]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 966 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 967 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
 968 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 969 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 970 // CHECK11-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 971 // CHECK11-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[TMP33]], 0
 972 // CHECK11-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
 973 // CHECK11-NEXT:    [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
 974 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
 975 // CHECK11-NEXT:    [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
 976 // CHECK11-NEXT:    store i64 [[SUB7]], ptr [[DOTCAPTURE_EXPR_3]], align 8
 977 // CHECK11-NEXT:    [[TMP34:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8
 978 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP34]], 1
 979 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 980 // CHECK11-NEXT:    store i32 3, ptr [[TMP35]], align 4
 981 // CHECK11-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 982 // CHECK11-NEXT:    store i32 5, ptr [[TMP36]], align 4
 983 // CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 984 // CHECK11-NEXT:    store ptr [[TMP27]], ptr [[TMP37]], align 4
 985 // CHECK11-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 986 // CHECK11-NEXT:    store ptr [[TMP28]], ptr [[TMP38]], align 4
 987 // CHECK11-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 988 // CHECK11-NEXT:    store ptr [[TMP29]], ptr [[TMP39]], align 4
 989 // CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 990 // CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP40]], align 4
 991 // CHECK11-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 992 // CHECK11-NEXT:    store ptr null, ptr [[TMP41]], align 4
 993 // CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 994 // CHECK11-NEXT:    store ptr null, ptr [[TMP42]], align 4
 995 // CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 996 // CHECK11-NEXT:    store i64 [[ADD]], ptr [[TMP43]], align 8
 997 // CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 998 // CHECK11-NEXT:    store i64 0, ptr [[TMP44]], align 8
 999 // CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1000 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP45]], align 4
1001 // CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1002 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP46]], align 4
1003 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1004 // CHECK11-NEXT:    store i32 0, ptr [[TMP47]], align 4
1005 // CHECK11-NEXT:    [[TMP48:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.region_id, ptr [[KERNEL_ARGS]])
1006 // CHECK11-NEXT:    [[TMP49:%.*]] = icmp ne i32 [[TMP48]], 0
1007 // CHECK11-NEXT:    br i1 [[TMP49]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1008 // CHECK11:       omp_offload.failed:
1009 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83(i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP0]], i32 [[TMP1]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
1010 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1011 // CHECK11:       omp_offload.cont:
1012 // CHECK11-NEXT:    [[TMP50:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
1013 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef [[TMP50]])
1014 // CHECK11-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
1015 // CHECK11-NEXT:    [[TMP51:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
1016 // CHECK11-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP51]])
1017 // CHECK11-NEXT:    [[TMP52:%.*]] = load i32, ptr [[RETVAL]], align 4
1018 // CHECK11-NEXT:    ret i32 [[TMP52]]
1019 //
1020 //
1021 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83
1022 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1023 // CHECK11-NEXT:  entry:
1024 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1025 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
1026 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1027 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1028 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1029 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
1030 // CHECK11-NEXT:    store i32 [[M]], ptr [[M_ADDR]], align 4
1031 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1032 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1033 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1034 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1035 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1036 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1037 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i32 [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
1038 // CHECK11-NEXT:    ret void
1039 //
1040 //
1041 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined
1042 // CHECK11-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(4) [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1043 // CHECK11-NEXT:  entry:
1044 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1045 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1046 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1047 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 4
1048 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1049 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1050 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1051 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1052 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1053 // CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
1054 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1055 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1056 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
1057 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1058 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1059 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
1060 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
1061 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1062 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1063 // CHECK11-NEXT:    [[I11:%.*]] = alloca i32, align 4
1064 // CHECK11-NEXT:    [[J12:%.*]] = alloca i32, align 4
1065 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1066 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1067 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1068 // CHECK11-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 4
1069 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1070 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1071 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1072 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1073 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4
1074 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1075 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1076 // CHECK11-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1077 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
1078 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
1079 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
1080 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
1081 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1082 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
1083 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1084 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
1085 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1086 // CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
1087 // CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1088 // CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
1089 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
1090 // CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
1091 // CHECK11-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
1092 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
1093 // CHECK11-NEXT:    store i32 0, ptr [[J]], align 4
1094 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1095 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
1096 // CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
1097 // CHECK11:       land.lhs.true:
1098 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1099 // CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
1100 // CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
1101 // CHECK11:       omp.precond.then:
1102 // CHECK11-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
1103 // CHECK11-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1104 // CHECK11-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_COMB_UB]], align 8
1105 // CHECK11-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
1106 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1107 // CHECK11-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1108 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
1109 // CHECK11-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP13]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
1110 // CHECK11-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1111 // CHECK11-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1112 // CHECK11-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
1113 // CHECK11-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1114 // CHECK11:       cond.true:
1115 // CHECK11-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1116 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1117 // CHECK11:       cond.false:
1118 // CHECK11-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1119 // CHECK11-NEXT:    br label [[COND_END]]
1120 // CHECK11:       cond.end:
1121 // CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
1122 // CHECK11-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
1123 // CHECK11-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
1124 // CHECK11-NEXT:    store i64 [[TMP18]], ptr [[DOTOMP_IV]], align 8
1125 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1126 // CHECK11:       omp.inner.for.cond:
1127 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1128 // CHECK11-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1129 // CHECK11-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
1130 // CHECK11-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1131 // CHECK11:       omp.inner.for.body:
1132 // CHECK11-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1133 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1134 // CHECK11-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP22]], 0
1135 // CHECK11-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
1136 // CHECK11-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
1137 // CHECK11-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
1138 // CHECK11-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP21]], [[CONV18]]
1139 // CHECK11-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
1140 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
1141 // CHECK11-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
1142 // CHECK11-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
1143 // CHECK11-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1144 // CHECK11-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1145 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1146 // CHECK11-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP25]], 0
1147 // CHECK11-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
1148 // CHECK11-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
1149 // CHECK11-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
1150 // CHECK11-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP24]], [[CONV25]]
1151 // CHECK11-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1152 // CHECK11-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP26]], 0
1153 // CHECK11-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
1154 // CHECK11-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
1155 // CHECK11-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
1156 // CHECK11-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
1157 // CHECK11-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP23]], [[MUL31]]
1158 // CHECK11-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
1159 // CHECK11-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
1160 // CHECK11-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
1161 // CHECK11-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
1162 // CHECK11-NEXT:    [[TMP27:%.*]] = load i32, ptr [[I11]], align 4
1163 // CHECK11-NEXT:    [[TMP28:%.*]] = mul nsw i32 [[TMP27]], [[TMP3]]
1164 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 [[TMP28]]
1165 // CHECK11-NEXT:    [[TMP29:%.*]] = load i32, ptr [[J12]], align 4
1166 // CHECK11-NEXT:    [[ARRAYIDX36:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 [[TMP29]]
1167 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX36]], align 4
1168 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1169 // CHECK11:       omp.body.continue:
1170 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1171 // CHECK11:       omp.inner.for.inc:
1172 // CHECK11-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1173 // CHECK11-NEXT:    [[ADD37:%.*]] = add nsw i64 [[TMP30]], 1
1174 // CHECK11-NEXT:    store i64 [[ADD37]], ptr [[DOTOMP_IV]], align 8
1175 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1176 // CHECK11:       omp.inner.for.end:
1177 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1178 // CHECK11:       omp.loop.exit:
1179 // CHECK11-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1180 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
1181 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP32]])
1182 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
1183 // CHECK11:       omp.precond.end:
1184 // CHECK11-NEXT:    ret void
1185 //
1186 //
1187 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1188 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1189 // CHECK11-NEXT:  entry:
1190 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1191 // CHECK11-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1192 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
1193 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
1194 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
1195 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1196 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1197 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1198 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1199 // CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1200 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP0]], align 4
1201 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1202 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
1203 // CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1204 // CHECK11-NEXT:    store ptr null, ptr [[TMP2]], align 4
1205 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1206 // CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1207 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1208 // CHECK11-NEXT:    store i32 3, ptr [[TMP5]], align 4
1209 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1210 // CHECK11-NEXT:    store i32 1, ptr [[TMP6]], align 4
1211 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1212 // CHECK11-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
1213 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1214 // CHECK11-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
1215 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1216 // CHECK11-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP9]], align 4
1217 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1218 // CHECK11-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP10]], align 4
1219 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1220 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
1221 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1222 // CHECK11-NEXT:    store ptr null, ptr [[TMP12]], align 4
1223 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1224 // CHECK11-NEXT:    store i64 20, ptr [[TMP13]], align 8
1225 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1226 // CHECK11-NEXT:    store i64 0, ptr [[TMP14]], align 8
1227 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1228 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
1229 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1230 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
1231 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1232 // CHECK11-NEXT:    store i32 0, ptr [[TMP17]], align 4
1233 // CHECK11-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.region_id, ptr [[KERNEL_ARGS]])
1234 // CHECK11-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1235 // CHECK11-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1236 // CHECK11:       omp_offload.failed:
1237 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69(ptr [[A]]) #[[ATTR3]]
1238 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1239 // CHECK11:       omp_offload.cont:
1240 // CHECK11-NEXT:    ret i32 0
1241 //
1242 //
1243 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69
1244 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1245 // CHECK11-NEXT:  entry:
1246 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1247 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1248 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1249 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
1250 // CHECK11-NEXT:    ret void
1251 //
1252 //
1253 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined
1254 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1255 // CHECK11-NEXT:  entry:
1256 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1257 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1258 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1259 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1260 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1261 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1262 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1263 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1264 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1265 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1266 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1267 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1268 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1269 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1270 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1271 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1272 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1273 // CHECK11-NEXT:    store i32 19, ptr [[DOTOMP_COMB_UB]], align 4
1274 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1275 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1276 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1277 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1278 // CHECK11-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)
1279 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1280 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
1281 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1282 // CHECK11:       cond.true:
1283 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1284 // CHECK11:       cond.false:
1285 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1286 // CHECK11-NEXT:    br label [[COND_END]]
1287 // CHECK11:       cond.end:
1288 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1289 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1290 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1291 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1292 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1293 // CHECK11:       omp.inner.for.cond:
1294 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1295 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1296 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1297 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1298 // CHECK11:       omp.inner.for.body:
1299 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1300 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP8]], 2
1301 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1302 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1303 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1304 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1305 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1306 // CHECK11-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP10]], 2
1307 // CHECK11-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
1308 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
1309 // CHECK11-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
1310 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
1311 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
1312 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
1313 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i32 0, i32 [[TMP11]]
1314 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[J]], align 4
1315 // CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP12]]
1316 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
1317 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1318 // CHECK11:       omp.body.continue:
1319 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1320 // CHECK11:       omp.inner.for.inc:
1321 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1322 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
1323 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
1324 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1325 // CHECK11:       omp.inner.for.end:
1326 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1327 // CHECK11:       omp.loop.exit:
1328 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1329 // CHECK11-NEXT:    ret void
1330 //