clang/test/OpenMP/target_teams_distribute_parallel_for_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 -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
   8 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
   9 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
  10 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
  11 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  12 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
  13
  14 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  15 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  16 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  17 // RUN: %clang_cc1 -DCK1 -verify -Wno-vla -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  18 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  19 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  20 #ifdef CK1
  21
  22 template <typename T, int X, long long Y>
  23 struct SS{
  24   T a[X][Y];
  25
  26   int foo(void) {
  27
  28     #pragma omp target teams distribute parallel for collapse(2)
  29     for(int i = 0; i < X; i++) {
  30       for(int j = 0; j < Y; j++) {
  31         a[i][j] = (T)0;
  32       }
  33     }
  34
  35     // discard loop variables not needed here
  36
  37
  38     return a[0][0];
  39   }
  40 };
  41
  42 int teams_template_struct(void) {
  43   SS<int, 123, 456> V;
  44   return V.foo();
  45
  46 }
  47 #endif // CK1
  48
  49 // Test host codegen.
  50 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
  51 // 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
  52 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
  53 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
  54 // 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
  55 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
  56
  57 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  58 // 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
  59 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  60 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  61 // 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
  62 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  63 #ifdef CK2
  64
  65 template <typename T, int n, int m>
  66 int tmain(T argc) {
  67   T a[n][m];
  68   #pragma omp target teams distribute parallel for collapse(2)
  69   for(int i = 0; i < n; i++) {
  70     for(int j = 0; j < m; j++) {
  71       a[i][j] = (T)0;
  72     }
  73   }
  74   return 0;
  75 }
  76
  77 int main (int argc, char **argv) {
  78   int n = 100;
  79   int m = 2;
  80   int a[n][m];
  81   #pragma omp target teams distribute parallel for collapse(2)
  82   for(int i = 0; i < n; i++) {
  83     for(int j = 0; j < m; j++) {
  84       a[i][j] = 0;
  85     }
  86   }
  87   return tmain<int, 10, 2>(argc);
  88 }
  89
  90
  91
  92
  93
  94
  95
  96
  97 // discard loop variables not needed here
  98
  99
 100 #endif // CK2
 101 #endif // #ifndef HEADER
 102 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 103 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 104 // CHECK1-NEXT:  entry:
 105 // CHECK1-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 106 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(224352) [[V]])
 107 // CHECK1-NEXT:    ret i32 [[CALL]]
 108 //
 109 //
 110 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 111 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat {
 112 // CHECK1-NEXT:  entry:
 113 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 114 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 115 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 116 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 117 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 118 // CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 119 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 120 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 121 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 122 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 123 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 124 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 8
 125 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 126 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
 127 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 128 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 129 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 130 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 131 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 132 // CHECK1-NEXT:    store i32 3, ptr [[TMP5]], align 4
 133 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 134 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
 135 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 136 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 137 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 138 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 139 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 140 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 8
 141 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 142 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 8
 143 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 144 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 145 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 146 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 147 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 148 // CHECK1-NEXT:    store i64 56088, ptr [[TMP13]], align 8
 149 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 150 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
 151 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 152 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 153 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 154 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 155 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 156 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
 157 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 158 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 159 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 160 // CHECK1:       omp_offload.failed:
 161 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 162 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 163 // CHECK1:       omp_offload.cont:
 164 // CHECK1-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 165 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A3]], i64 0, i64 0
 166 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 0
 167 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
 168 // CHECK1-NEXT:    ret i32 [[TMP20]]
 169 //
 170 //
 171 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 172 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 173 // CHECK1-NEXT:  entry:
 174 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 175 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 176 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 177 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 178 // CHECK1-NEXT:    ret void
 179 //
 180 //
 181 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 182 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 183 // CHECK1-NEXT:  entry:
 184 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 185 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 186 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 187 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 188 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 189 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 190 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 191 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 192 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 193 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 194 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 195 // CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
 196 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 197 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 198 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 199 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 200 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 201 // CHECK1-NEXT:    store i32 56087, ptr [[DOTOMP_COMB_UB]], align 4
 202 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 203 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 204 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 205 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 206 // 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)
 207 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 208 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
 209 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 210 // CHECK1:       cond.true:
 211 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 212 // CHECK1:       cond.false:
 213 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 214 // CHECK1-NEXT:    br label [[COND_END]]
 215 // CHECK1:       cond.end:
 216 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 217 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 218 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 219 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 220 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 221 // CHECK1:       omp.inner.for.cond:
 222 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 223 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 224 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 225 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 226 // CHECK1:       omp.inner.for.body:
 227 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 228 // CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
 229 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 230 // CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
 231 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
 232 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 233 // CHECK1:       omp.inner.for.inc:
 234 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 235 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 236 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
 237 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 238 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 239 // CHECK1:       omp.inner.for.end:
 240 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 241 // CHECK1:       omp.loop.exit:
 242 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 243 // CHECK1-NEXT:    ret void
 244 //
 245 //
 246 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined
 247 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 248 // CHECK1-NEXT:  entry:
 249 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 250 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 251 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 252 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 253 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 254 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 255 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 256 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 257 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 258 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 259 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 260 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 261 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 262 // CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
 263 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 264 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 265 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 266 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 267 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 268 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 269 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 270 // CHECK1-NEXT:    store i32 56087, ptr [[DOTOMP_UB]], align 4
 271 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 272 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
 273 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 274 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
 275 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 276 // CHECK1-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
 277 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 278 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 279 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 280 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 281 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 282 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 283 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 56087
 284 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 285 // CHECK1:       cond.true:
 286 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 287 // CHECK1:       cond.false:
 288 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 289 // CHECK1-NEXT:    br label [[COND_END]]
 290 // CHECK1:       cond.end:
 291 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 292 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 293 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 294 // CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 295 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 296 // CHECK1:       omp.inner.for.cond:
 297 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 298 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 299 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 300 // CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 301 // CHECK1:       omp.inner.for.body:
 302 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 303 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
 304 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 305 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 306 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 307 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 308 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 309 // CHECK1-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 456
 310 // CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 456
 311 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
 312 // CHECK1-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
 313 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
 314 // CHECK1-NEXT:    store i32 [[ADD7]], ptr [[J]], align 4
 315 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 316 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
 317 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
 318 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i64 0, i64 [[IDXPROM]]
 319 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
 320 // CHECK1-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
 321 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
 322 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX9]], align 4
 323 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 324 // CHECK1:       omp.body.continue:
 325 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 326 // CHECK1:       omp.inner.for.inc:
 327 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 328 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
 329 // CHECK1-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
 330 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 331 // CHECK1:       omp.inner.for.end:
 332 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 333 // CHECK1:       omp.loop.exit:
 334 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
 335 // CHECK1-NEXT:    ret void
 336 //
 337 //
 338 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 339 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 340 // CHECK3-NEXT:  entry:
 341 // CHECK3-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 342 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(224352) [[V]])
 343 // CHECK3-NEXT:    ret i32 [[CALL]]
 344 //
 345 //
 346 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 347 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
 348 // CHECK3-NEXT:  entry:
 349 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 350 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
 351 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
 352 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
 353 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 354 // CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 355 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 356 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 357 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 358 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 359 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 360 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 4
 361 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 362 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
 363 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 364 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
 365 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 366 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 367 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 368 // CHECK3-NEXT:    store i32 3, ptr [[TMP5]], align 4
 369 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 370 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
 371 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 372 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
 373 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 374 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
 375 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 376 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 4
 377 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 378 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 4
 379 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 380 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
 381 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 382 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 383 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 384 // CHECK3-NEXT:    store i64 56088, ptr [[TMP13]], align 8
 385 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 386 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
 387 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 388 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 389 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 390 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 391 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 392 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
 393 // CHECK3-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, ptr [[KERNEL_ARGS]])
 394 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 395 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 396 // CHECK3:       omp_offload.failed:
 397 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 398 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 399 // CHECK3:       omp_offload.cont:
 400 // CHECK3-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 401 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A3]], i32 0, i32 0
 402 // CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 0
 403 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
 404 // CHECK3-NEXT:    ret i32 [[TMP20]]
 405 //
 406 //
 407 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 408 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 409 // CHECK3-NEXT:  entry:
 410 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 411 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 412 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 413 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 414 // CHECK3-NEXT:    ret void
 415 //
 416 //
 417 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 418 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 419 // CHECK3-NEXT:  entry:
 420 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 421 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 422 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 423 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 424 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 425 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 426 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 427 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 428 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 429 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 430 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 431 // CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
 432 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 433 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 434 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 435 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 436 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 437 // CHECK3-NEXT:    store i32 56087, ptr [[DOTOMP_COMB_UB]], align 4
 438 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 439 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 440 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 441 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 442 // 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)
 443 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 444 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
 445 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 446 // CHECK3:       cond.true:
 447 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 448 // CHECK3:       cond.false:
 449 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 450 // CHECK3-NEXT:    br label [[COND_END]]
 451 // CHECK3:       cond.end:
 452 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 453 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 454 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 455 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 456 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 457 // CHECK3:       omp.inner.for.cond:
 458 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 459 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 460 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 461 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 462 // CHECK3:       omp.inner.for.body:
 463 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 464 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 465 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
 466 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 467 // CHECK3:       omp.inner.for.inc:
 468 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 469 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 470 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
 471 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 472 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 473 // CHECK3:       omp.inner.for.end:
 474 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 475 // CHECK3:       omp.loop.exit:
 476 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 477 // CHECK3-NEXT:    ret void
 478 //
 479 //
 480 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined
 481 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 482 // CHECK3-NEXT:  entry:
 483 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 484 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 485 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 486 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 487 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 488 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 489 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 490 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 491 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 492 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 493 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 494 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 495 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 496 // CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
 497 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 498 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 499 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 500 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 501 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 502 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 503 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 504 // CHECK3-NEXT:    store i32 56087, ptr [[DOTOMP_UB]], align 4
 505 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 506 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 507 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
 508 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
 509 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 510 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 511 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 512 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 513 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 514 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 515 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 56087
 516 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 517 // CHECK3:       cond.true:
 518 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 519 // CHECK3:       cond.false:
 520 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 521 // CHECK3-NEXT:    br label [[COND_END]]
 522 // CHECK3:       cond.end:
 523 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 524 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 525 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 526 // CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 527 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 528 // CHECK3:       omp.inner.for.cond:
 529 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 530 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 531 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 532 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 533 // CHECK3:       omp.inner.for.body:
 534 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 535 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
 536 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 537 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 538 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 539 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 540 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 541 // CHECK3-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 456
 542 // CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
 543 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
 544 // CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 545 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
 546 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
 547 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 548 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
 549 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i32 0, i32 [[TMP13]]
 550 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
 551 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP14]]
 552 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
 553 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 554 // CHECK3:       omp.body.continue:
 555 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 556 // CHECK3:       omp.inner.for.inc:
 557 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 558 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
 559 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
 560 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 561 // CHECK3:       omp.inner.for.end:
 562 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 563 // CHECK3:       omp.loop.exit:
 564 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
 565 // CHECK3-NEXT:    ret void
 566 //
 567 //
 568 // CHECK9-LABEL: define {{[^@]+}}@main
 569 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 570 // CHECK9-NEXT:  entry:
 571 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 572 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 573 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
 574 // CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
 575 // CHECK9-NEXT:    [[M:%.*]] = alloca i32, align 4
 576 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
 577 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
 578 // CHECK9-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
 579 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 580 // CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
 581 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 8
 582 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 8
 583 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 8
 584 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
 585 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 586 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 587 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 588 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 589 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
 590 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 591 // CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 592 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 593 // CHECK9-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
 594 // CHECK9-NEXT:    store i32 100, ptr [[N]], align 4
 595 // CHECK9-NEXT:    store i32 2, ptr [[M]], align 4
 596 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 597 // CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
 598 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[M]], align 4
 599 // CHECK9-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
 600 // CHECK9-NEXT:    [[TMP4:%.*]] = call ptr @llvm.stacksave.p0()
 601 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[SAVED_STACK]], align 8
 602 // CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
 603 // CHECK9-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 4
 604 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
 605 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[__VLA_EXPR1]], align 8
 606 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N]], align 4
 607 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED]], align 4
 608 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED]], align 8
 609 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[M]], align 4
 610 // CHECK9-NEXT:    store i32 [[TMP8]], ptr [[M_CASTED]], align 4
 611 // CHECK9-NEXT:    [[TMP9:%.*]] = load i64, ptr [[M_CASTED]], align 8
 612 // CHECK9-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
 613 // CHECK9-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
 614 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 40, i1 false)
 615 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 616 // CHECK9-NEXT:    store i64 [[TMP7]], ptr [[TMP12]], align 8
 617 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 618 // CHECK9-NEXT:    store i64 [[TMP7]], ptr [[TMP13]], align 8
 619 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 620 // CHECK9-NEXT:    store ptr null, ptr [[TMP14]], align 8
 621 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 622 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[TMP15]], align 8
 623 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 624 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[TMP16]], align 8
 625 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
 626 // CHECK9-NEXT:    store ptr null, ptr [[TMP17]], align 8
 627 // CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 628 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP18]], align 8
 629 // CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 630 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP19]], align 8
 631 // CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
 632 // CHECK9-NEXT:    store ptr null, ptr [[TMP20]], align 8
 633 // CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 634 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP21]], align 8
 635 // CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 636 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP22]], align 8
 637 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
 638 // CHECK9-NEXT:    store ptr null, ptr [[TMP23]], align 8
 639 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 640 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP24]], align 8
 641 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 642 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP25]], align 8
 643 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 4
 644 // CHECK9-NEXT:    store i64 [[TMP11]], ptr [[TMP26]], align 8
 645 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
 646 // CHECK9-NEXT:    store ptr null, ptr [[TMP27]], align 8
 647 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 648 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 649 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
 650 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, ptr [[N]], align 4
 651 // CHECK9-NEXT:    store i32 [[TMP31]], ptr [[DOTCAPTURE_EXPR_]], align 4
 652 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, ptr [[M]], align 4
 653 // CHECK9-NEXT:    store i32 [[TMP32]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 654 // CHECK9-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 655 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP33]], 0
 656 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 657 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 658 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 659 // CHECK9-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[TMP34]], 0
 660 // CHECK9-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
 661 // CHECK9-NEXT:    [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
 662 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
 663 // CHECK9-NEXT:    [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
 664 // CHECK9-NEXT:    store i64 [[SUB7]], ptr [[DOTCAPTURE_EXPR_3]], align 8
 665 // CHECK9-NEXT:    [[TMP35:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8
 666 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP35]], 1
 667 // CHECK9-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 668 // CHECK9-NEXT:    store i32 3, ptr [[TMP36]], align 4
 669 // CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 670 // CHECK9-NEXT:    store i32 5, ptr [[TMP37]], align 4
 671 // CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 672 // CHECK9-NEXT:    store ptr [[TMP28]], ptr [[TMP38]], align 8
 673 // CHECK9-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 674 // CHECK9-NEXT:    store ptr [[TMP29]], ptr [[TMP39]], align 8
 675 // CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 676 // CHECK9-NEXT:    store ptr [[TMP30]], ptr [[TMP40]], align 8
 677 // CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 678 // CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP41]], align 8
 679 // CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 680 // CHECK9-NEXT:    store ptr null, ptr [[TMP42]], align 8
 681 // CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 682 // CHECK9-NEXT:    store ptr null, ptr [[TMP43]], align 8
 683 // CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 684 // CHECK9-NEXT:    store i64 [[ADD]], ptr [[TMP44]], align 8
 685 // CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 686 // CHECK9-NEXT:    store i64 0, ptr [[TMP45]], align 8
 687 // CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 688 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP46]], align 4
 689 // CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 690 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4
 691 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 692 // CHECK9-NEXT:    store i32 0, ptr [[TMP48]], align 4
 693 // CHECK9-NEXT:    [[TMP49:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.region_id, ptr [[KERNEL_ARGS]])
 694 // CHECK9-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
 695 // CHECK9-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 696 // CHECK9:       omp_offload.failed:
 697 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81(i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP1]], i64 [[TMP3]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
 698 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 699 // CHECK9:       omp_offload.cont:
 700 // CHECK9-NEXT:    [[TMP51:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
 701 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef signext [[TMP51]])
 702 // CHECK9-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
 703 // CHECK9-NEXT:    [[TMP52:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
 704 // CHECK9-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP52]])
 705 // CHECK9-NEXT:    [[TMP53:%.*]] = load i32, ptr [[RETVAL]], align 4
 706 // CHECK9-NEXT:    ret i32 [[TMP53]]
 707 //
 708 //
 709 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
 710 // 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]+]] {
 711 // CHECK9-NEXT:  entry:
 712 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 713 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
 714 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 715 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 716 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 717 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 718 // CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
 719 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 720 // CHECK9-NEXT:    store i64 [[M]], ptr [[M_ADDR]], align 8
 721 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 722 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 723 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 724 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 725 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 726 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 727 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
 728 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
 729 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, ptr [[N_CASTED]], align 8
 730 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[M_ADDR]], align 4
 731 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[M_CASTED]], align 4
 732 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, ptr [[M_CASTED]], align 8
 733 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined, i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
 734 // CHECK9-NEXT:    ret void
 735 //
 736 //
 737 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined
 738 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
 739 // CHECK9-NEXT:  entry:
 740 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 741 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 742 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 743 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
 744 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 745 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 746 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 747 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 748 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 749 // CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
 750 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 751 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
 752 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
 753 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 754 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
 755 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
 756 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
 757 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 758 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 759 // CHECK9-NEXT:    [[I11:%.*]] = alloca i32, align 4
 760 // CHECK9-NEXT:    [[J12:%.*]] = alloca i32, align 4
 761 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 762 // CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
 763 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 764 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 765 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 766 // CHECK9-NEXT:    store i64 [[M]], ptr [[M_ADDR]], align 8
 767 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 768 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 769 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 770 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 771 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 772 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 773 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
 774 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
 775 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[M_ADDR]], align 4
 776 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_4]], align 4
 777 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 778 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
 779 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 780 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 781 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 782 // CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP6]], 0
 783 // CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 784 // CHECK9-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
 785 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
 786 // CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
 787 // CHECK9-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
 788 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
 789 // CHECK9-NEXT:    store i32 0, ptr [[J]], align 4
 790 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 791 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
 792 // CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
 793 // CHECK9:       land.lhs.true:
 794 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 795 // CHECK9-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP8]]
 796 // CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
 797 // CHECK9:       omp.precond.then:
 798 // CHECK9-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
 799 // CHECK9-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 800 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[DOTOMP_COMB_UB]], align 8
 801 // CHECK9-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
 802 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 803 // CHECK9-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 804 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
 805 // CHECK9-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
 806 // CHECK9-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 807 // CHECK9-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 808 // CHECK9-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP12]], [[TMP13]]
 809 // CHECK9-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 810 // CHECK9:       cond.true:
 811 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 812 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 813 // CHECK9:       cond.false:
 814 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 815 // CHECK9-NEXT:    br label [[COND_END]]
 816 // CHECK9:       cond.end:
 817 // CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
 818 // CHECK9-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
 819 // CHECK9-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
 820 // CHECK9-NEXT:    store i64 [[TMP16]], ptr [[DOTOMP_IV]], align 8
 821 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 822 // CHECK9:       omp.inner.for.cond:
 823 // CHECK9-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 824 // CHECK9-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 825 // CHECK9-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP17]], [[TMP18]]
 826 // CHECK9-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 827 // CHECK9:       omp.inner.for.body:
 828 // CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
 829 // CHECK9-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 830 // CHECK9-NEXT:    [[TMP21:%.*]] = load i32, ptr [[N_ADDR]], align 4
 831 // CHECK9-NEXT:    store i32 [[TMP21]], ptr [[N_CASTED]], align 4
 832 // CHECK9-NEXT:    [[TMP22:%.*]] = load i64, ptr [[N_CASTED]], align 8
 833 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, ptr [[M_ADDR]], align 4
 834 // CHECK9-NEXT:    store i32 [[TMP23]], ptr [[M_CASTED]], align 4
 835 // CHECK9-NEXT:    [[TMP24:%.*]] = load i64, ptr [[M_CASTED]], align 8
 836 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined.omp_outlined, i64 [[TMP19]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP24]], i64 [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
 837 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 838 // CHECK9:       omp.inner.for.inc:
 839 // CHECK9-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 840 // CHECK9-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
 841 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP25]], [[TMP26]]
 842 // CHECK9-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
 843 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 844 // CHECK9:       omp.inner.for.end:
 845 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 846 // CHECK9:       omp.loop.exit:
 847 // CHECK9-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 848 // CHECK9-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
 849 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP28]])
 850 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 851 // CHECK9:       omp.precond.end:
 852 // CHECK9-NEXT:    ret void
 853 //
 854 //
 855 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined.omp_outlined
 856 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
 857 // CHECK9-NEXT:  entry:
 858 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 859 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 860 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 861 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 862 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 863 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
 864 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 865 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 866 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 867 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 868 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 869 // CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
 870 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 871 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
 872 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
 873 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 874 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
 875 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 876 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 877 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 878 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 879 // CHECK9-NEXT:    [[I11:%.*]] = alloca i32, align 4
 880 // CHECK9-NEXT:    [[J12:%.*]] = alloca i32, align 4
 881 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 882 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 883 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 884 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 885 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 886 // CHECK9-NEXT:    store i64 [[M]], ptr [[M_ADDR]], align 8
 887 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 888 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 889 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 890 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 891 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 892 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 893 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
 894 // CHECK9-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
 895 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[M_ADDR]], align 4
 896 // CHECK9-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_4]], align 4
 897 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 898 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
 899 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 900 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 901 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 902 // CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP6]], 0
 903 // CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 904 // CHECK9-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
 905 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
 906 // CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
 907 // CHECK9-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
 908 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
 909 // CHECK9-NEXT:    store i32 0, ptr [[J]], align 4
 910 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 911 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
 912 // CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
 913 // CHECK9:       land.lhs.true:
 914 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 915 // CHECK9-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP8]]
 916 // CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
 917 // CHECK9:       omp.precond.then:
 918 // CHECK9-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
 919 // CHECK9-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 920 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[DOTOMP_UB]], align 8
 921 // CHECK9-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 922 // CHECK9-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 923 // CHECK9-NEXT:    store i64 [[TMP10]], ptr [[DOTOMP_LB]], align 8
 924 // CHECK9-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_UB]], align 8
 925 // CHECK9-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
 926 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 927 // CHECK9-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 928 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
 929 // CHECK9-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP13]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
 930 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 931 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 932 // CHECK9-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
 933 // CHECK9-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 934 // CHECK9:       cond.true:
 935 // CHECK9-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 936 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 937 // CHECK9:       cond.false:
 938 // CHECK9-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 939 // CHECK9-NEXT:    br label [[COND_END]]
 940 // CHECK9:       cond.end:
 941 // CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
 942 // CHECK9-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
 943 // CHECK9-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
 944 // CHECK9-NEXT:    store i64 [[TMP18]], ptr [[DOTOMP_IV]], align 8
 945 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 946 // CHECK9:       omp.inner.for.cond:
 947 // CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 948 // CHECK9-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 949 // CHECK9-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
 950 // CHECK9-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 951 // CHECK9:       omp.inner.for.body:
 952 // CHECK9-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 953 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 954 // CHECK9-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP22]], 0
 955 // CHECK9-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
 956 // CHECK9-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
 957 // CHECK9-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
 958 // CHECK9-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP21]], [[CONV18]]
 959 // CHECK9-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
 960 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
 961 // CHECK9-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
 962 // CHECK9-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
 963 // CHECK9-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 964 // CHECK9-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 965 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 966 // CHECK9-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP25]], 0
 967 // CHECK9-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
 968 // CHECK9-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
 969 // CHECK9-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
 970 // CHECK9-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP24]], [[CONV25]]
 971 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 972 // CHECK9-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP26]], 0
 973 // CHECK9-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
 974 // CHECK9-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
 975 // CHECK9-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
 976 // CHECK9-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
 977 // CHECK9-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP23]], [[MUL31]]
 978 // CHECK9-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
 979 // CHECK9-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
 980 // CHECK9-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
 981 // CHECK9-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
 982 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[I11]], align 4
 983 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP27]] to i64
 984 // CHECK9-NEXT:    [[TMP28:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP1]]
 985 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 [[TMP28]]
 986 // CHECK9-NEXT:    [[TMP29:%.*]] = load i32, ptr [[J12]], align 4
 987 // CHECK9-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP29]] to i64
 988 // CHECK9-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i64 [[IDXPROM36]]
 989 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX37]], align 4
 990 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 991 // CHECK9:       omp.body.continue:
 992 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 993 // CHECK9:       omp.inner.for.inc:
 994 // CHECK9-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 995 // CHECK9-NEXT:    [[ADD38:%.*]] = add nsw i64 [[TMP30]], 1
 996 // CHECK9-NEXT:    store i64 [[ADD38]], ptr [[DOTOMP_IV]], align 8
 997 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 998 // CHECK9:       omp.inner.for.end:
 999 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1000 // CHECK9:       omp.loop.exit:
1001 // CHECK9-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1002 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
1003 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP32]])
1004 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
1005 // CHECK9:       omp.precond.end:
1006 // CHECK9-NEXT:    ret void
1007 //
1008 //
1009 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1010 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1011 // CHECK9-NEXT:  entry:
1012 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1013 // CHECK9-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1014 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
1015 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
1016 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
1017 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1018 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1019 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1020 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1021 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1022 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP0]], align 8
1023 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1024 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
1025 // CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1026 // CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
1027 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1028 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1029 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1030 // CHECK9-NEXT:    store i32 3, ptr [[TMP5]], align 4
1031 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1032 // CHECK9-NEXT:    store i32 1, ptr [[TMP6]], align 4
1033 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1034 // CHECK9-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
1035 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1036 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
1037 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1038 // CHECK9-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP9]], align 8
1039 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1040 // CHECK9-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP10]], align 8
1041 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1042 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
1043 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1044 // CHECK9-NEXT:    store ptr null, ptr [[TMP12]], align 8
1045 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1046 // CHECK9-NEXT:    store i64 20, ptr [[TMP13]], align 8
1047 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1048 // CHECK9-NEXT:    store i64 0, ptr [[TMP14]], align 8
1049 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1050 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
1051 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1052 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
1053 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1054 // CHECK9-NEXT:    store i32 0, ptr [[TMP17]], align 4
1055 // CHECK9-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.region_id, ptr [[KERNEL_ARGS]])
1056 // CHECK9-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1057 // CHECK9-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1058 // CHECK9:       omp_offload.failed:
1059 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68(ptr [[A]]) #[[ATTR3]]
1060 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1061 // CHECK9:       omp_offload.cont:
1062 // CHECK9-NEXT:    ret i32 0
1063 //
1064 //
1065 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68
1066 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1067 // CHECK9-NEXT:  entry:
1068 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1069 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1070 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1071 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined, ptr [[TMP0]])
1072 // CHECK9-NEXT:    ret void
1073 //
1074 //
1075 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined
1076 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1077 // CHECK9-NEXT:  entry:
1078 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1079 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1080 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1081 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1082 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1083 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1084 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1085 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1086 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1087 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1088 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1089 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
1090 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1091 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1092 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1093 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1094 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1095 // CHECK9-NEXT:    store i32 19, ptr [[DOTOMP_COMB_UB]], align 4
1096 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1097 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1098 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1099 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1100 // 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)
1101 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1102 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
1103 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1104 // CHECK9:       cond.true:
1105 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1106 // CHECK9:       cond.false:
1107 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1108 // CHECK9-NEXT:    br label [[COND_END]]
1109 // CHECK9:       cond.end:
1110 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1111 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1112 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1113 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1114 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1115 // CHECK9:       omp.inner.for.cond:
1116 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1117 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1118 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1119 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1120 // CHECK9:       omp.inner.for.body:
1121 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1122 // CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1123 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1124 // CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1125 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
1126 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1127 // CHECK9:       omp.inner.for.inc:
1128 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1129 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1130 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1131 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1132 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1133 // CHECK9:       omp.inner.for.end:
1134 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1135 // CHECK9:       omp.loop.exit:
1136 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1137 // CHECK9-NEXT:    ret void
1138 //
1139 //
1140 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined.omp_outlined
1141 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1142 // CHECK9-NEXT:  entry:
1143 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1144 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1145 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1146 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1147 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1148 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1149 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1150 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1151 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1152 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1153 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1154 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1155 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1156 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
1157 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1158 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1159 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1160 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1161 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1162 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1163 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1164 // CHECK9-NEXT:    store i32 19, ptr [[DOTOMP_UB]], align 4
1165 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1166 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1167 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1168 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
1169 // CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1170 // CHECK9-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
1171 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1172 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1173 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1174 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
1175 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1176 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1177 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
1178 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1179 // CHECK9:       cond.true:
1180 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1181 // CHECK9:       cond.false:
1182 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1183 // CHECK9-NEXT:    br label [[COND_END]]
1184 // CHECK9:       cond.end:
1185 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1186 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1187 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1188 // CHECK9-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
1189 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1190 // CHECK9:       omp.inner.for.cond:
1191 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1192 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1193 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1194 // CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1195 // CHECK9:       omp.inner.for.body:
1196 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1197 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
1198 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1199 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1200 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1201 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1202 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1203 // CHECK9-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 2
1204 // CHECK9-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 2
1205 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
1206 // CHECK9-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
1207 // CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
1208 // CHECK9-NEXT:    store i32 [[ADD7]], ptr [[J]], align 4
1209 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
1210 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
1211 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1212 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
1213 // CHECK9-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
1214 // CHECK9-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
1215 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX9]], align 4
1216 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1217 // CHECK9:       omp.body.continue:
1218 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1219 // CHECK9:       omp.inner.for.inc:
1220 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1221 // CHECK9-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
1222 // CHECK9-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
1223 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1224 // CHECK9:       omp.inner.for.end:
1225 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1226 // CHECK9:       omp.loop.exit:
1227 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
1228 // CHECK9-NEXT:    ret void
1229 //
1230 //
1231 // CHECK11-LABEL: define {{[^@]+}}@main
1232 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1233 // CHECK11-NEXT:  entry:
1234 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1235 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1236 // CHECK11-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 4
1237 // CHECK11-NEXT:    [[N:%.*]] = alloca i32, align 4
1238 // CHECK11-NEXT:    [[M:%.*]] = alloca i32, align 4
1239 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
1240 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1241 // CHECK11-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
1242 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
1243 // CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
1244 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 4
1245 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 4
1246 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 4
1247 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
1248 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1249 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1250 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1251 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1252 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
1253 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1254 // CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
1255 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1256 // CHECK11-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 4
1257 // CHECK11-NEXT:    store i32 100, ptr [[N]], align 4
1258 // CHECK11-NEXT:    store i32 2, ptr [[M]], align 4
1259 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
1260 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[M]], align 4
1261 // CHECK11-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
1262 // CHECK11-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 4
1263 // CHECK11-NEXT:    [[TMP3:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
1264 // CHECK11-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP3]], align 4
1265 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[__VLA_EXPR0]], align 4
1266 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[__VLA_EXPR1]], align 4
1267 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N]], align 4
1268 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
1269 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_CASTED]], align 4
1270 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[M]], align 4
1271 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[M_CASTED]], align 4
1272 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[M_CASTED]], align 4
1273 // CHECK11-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
1274 // CHECK11-NEXT:    [[TMP9:%.*]] = mul nuw i32 [[TMP8]], 4
1275 // CHECK11-NEXT:    [[TMP10:%.*]] = sext i32 [[TMP9]] to i64
1276 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 40, i1 false)
1277 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1278 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP11]], align 4
1279 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1280 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP12]], align 4
1281 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1282 // CHECK11-NEXT:    store ptr null, ptr [[TMP13]], align 4
1283 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1284 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[TMP14]], align 4
1285 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1286 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[TMP15]], align 4
1287 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1288 // CHECK11-NEXT:    store ptr null, ptr [[TMP16]], align 4
1289 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1290 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP17]], align 4
1291 // CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1292 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP18]], align 4
1293 // CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1294 // CHECK11-NEXT:    store ptr null, ptr [[TMP19]], align 4
1295 // CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1296 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP20]], align 4
1297 // CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1298 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP21]], align 4
1299 // CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
1300 // CHECK11-NEXT:    store ptr null, ptr [[TMP22]], align 4
1301 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1302 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP23]], align 4
1303 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1304 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP24]], align 4
1305 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 4
1306 // CHECK11-NEXT:    store i64 [[TMP10]], ptr [[TMP25]], align 4
1307 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
1308 // CHECK11-NEXT:    store ptr null, ptr [[TMP26]], align 4
1309 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1310 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1311 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1312 // CHECK11-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N]], align 4
1313 // CHECK11-NEXT:    store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR_]], align 4
1314 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32, ptr [[M]], align 4
1315 // CHECK11-NEXT:    store i32 [[TMP31]], ptr [[DOTCAPTURE_EXPR_2]], align 4
1316 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1317 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
1318 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1319 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
1320 // CHECK11-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1321 // CHECK11-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[TMP33]], 0
1322 // CHECK11-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
1323 // CHECK11-NEXT:    [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
1324 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
1325 // CHECK11-NEXT:    [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
1326 // CHECK11-NEXT:    store i64 [[SUB7]], ptr [[DOTCAPTURE_EXPR_3]], align 8
1327 // CHECK11-NEXT:    [[TMP34:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8
1328 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP34]], 1
1329 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1330 // CHECK11-NEXT:    store i32 3, ptr [[TMP35]], align 4
1331 // CHECK11-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1332 // CHECK11-NEXT:    store i32 5, ptr [[TMP36]], align 4
1333 // CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1334 // CHECK11-NEXT:    store ptr [[TMP27]], ptr [[TMP37]], align 4
1335 // CHECK11-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1336 // CHECK11-NEXT:    store ptr [[TMP28]], ptr [[TMP38]], align 4
1337 // CHECK11-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1338 // CHECK11-NEXT:    store ptr [[TMP29]], ptr [[TMP39]], align 4
1339 // CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1340 // CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP40]], align 4
1341 // CHECK11-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1342 // CHECK11-NEXT:    store ptr null, ptr [[TMP41]], align 4
1343 // CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1344 // CHECK11-NEXT:    store ptr null, ptr [[TMP42]], align 4
1345 // CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1346 // CHECK11-NEXT:    store i64 [[ADD]], ptr [[TMP43]], align 8
1347 // CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1348 // CHECK11-NEXT:    store i64 0, ptr [[TMP44]], align 8
1349 // CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1350 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP45]], align 4
1351 // CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1352 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP46]], align 4
1353 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1354 // CHECK11-NEXT:    store i32 0, ptr [[TMP47]], align 4
1355 // CHECK11-NEXT:    [[TMP48:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.region_id, ptr [[KERNEL_ARGS]])
1356 // CHECK11-NEXT:    [[TMP49:%.*]] = icmp ne i32 [[TMP48]], 0
1357 // CHECK11-NEXT:    br i1 [[TMP49]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1358 // CHECK11:       omp_offload.failed:
1359 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81(i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP0]], i32 [[TMP1]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
1360 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1361 // CHECK11:       omp_offload.cont:
1362 // CHECK11-NEXT:    [[TMP50:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
1363 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef [[TMP50]])
1364 // CHECK11-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
1365 // CHECK11-NEXT:    [[TMP51:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
1366 // CHECK11-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP51]])
1367 // CHECK11-NEXT:    [[TMP52:%.*]] = load i32, ptr [[RETVAL]], align 4
1368 // CHECK11-NEXT:    ret i32 [[TMP52]]
1369 //
1370 //
1371 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
1372 // 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]+]] {
1373 // CHECK11-NEXT:  entry:
1374 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1375 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
1376 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1377 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1378 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1379 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
1380 // CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
1381 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
1382 // CHECK11-NEXT:    store i32 [[M]], ptr [[M_ADDR]], align 4
1383 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1384 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1385 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1386 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1387 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1388 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1389 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
1390 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[N_CASTED]], align 4
1391 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N_CASTED]], align 4
1392 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[M_ADDR]], align 4
1393 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[M_CASTED]], align 4
1394 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[M_CASTED]], align 4
1395 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined, i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
1396 // CHECK11-NEXT:    ret void
1397 //
1398 //
1399 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined
1400 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1401 // CHECK11-NEXT:  entry:
1402 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1403 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1404 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1405 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
1406 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1407 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1408 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1409 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1410 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1411 // CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
1412 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1413 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1414 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
1415 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1416 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1417 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
1418 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
1419 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1420 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1421 // CHECK11-NEXT:    [[I11:%.*]] = alloca i32, align 4
1422 // CHECK11-NEXT:    [[J12:%.*]] = alloca i32, align 4
1423 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
1424 // CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
1425 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1426 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1427 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
1428 // CHECK11-NEXT:    store i32 [[M]], ptr [[M_ADDR]], align 4
1429 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1430 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1431 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1432 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1433 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1434 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1435 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
1436 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1437 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[M_ADDR]], align 4
1438 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_4]], align 4
1439 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1440 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
1441 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1442 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
1443 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1444 // CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP6]], 0
1445 // CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1446 // CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
1447 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
1448 // CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
1449 // CHECK11-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
1450 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
1451 // CHECK11-NEXT:    store i32 0, ptr [[J]], align 4
1452 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1453 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
1454 // CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
1455 // CHECK11:       land.lhs.true:
1456 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1457 // CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP8]]
1458 // CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
1459 // CHECK11:       omp.precond.then:
1460 // CHECK11-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
1461 // CHECK11-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1462 // CHECK11-NEXT:    store i64 [[TMP9]], ptr [[DOTOMP_COMB_UB]], align 8
1463 // CHECK11-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
1464 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1465 // CHECK11-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1466 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
1467 // CHECK11-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
1468 // CHECK11-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1469 // CHECK11-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1470 // CHECK11-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP12]], [[TMP13]]
1471 // CHECK11-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1472 // CHECK11:       cond.true:
1473 // CHECK11-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1474 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1475 // CHECK11:       cond.false:
1476 // CHECK11-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1477 // CHECK11-NEXT:    br label [[COND_END]]
1478 // CHECK11:       cond.end:
1479 // CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
1480 // CHECK11-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
1481 // CHECK11-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
1482 // CHECK11-NEXT:    store i64 [[TMP16]], ptr [[DOTOMP_IV]], align 8
1483 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1484 // CHECK11:       omp.inner.for.cond:
1485 // CHECK11-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1486 // CHECK11-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1487 // CHECK11-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP17]], [[TMP18]]
1488 // CHECK11-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1489 // CHECK11:       omp.inner.for.body:
1490 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
1491 // CHECK11-NEXT:    [[TMP20:%.*]] = trunc i64 [[TMP19]] to i32
1492 // CHECK11-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1493 // CHECK11-NEXT:    [[TMP22:%.*]] = trunc i64 [[TMP21]] to i32
1494 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, ptr [[N_ADDR]], align 4
1495 // CHECK11-NEXT:    store i32 [[TMP23]], ptr [[N_CASTED]], align 4
1496 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, ptr [[N_CASTED]], align 4
1497 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[M_ADDR]], align 4
1498 // CHECK11-NEXT:    store i32 [[TMP25]], ptr [[M_CASTED]], align 4
1499 // CHECK11-NEXT:    [[TMP26:%.*]] = load i32, ptr [[M_CASTED]], align 4
1500 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined.omp_outlined, i32 [[TMP20]], i32 [[TMP22]], i32 [[TMP24]], i32 [[TMP26]], i32 [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
1501 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1502 // CHECK11:       omp.inner.for.inc:
1503 // CHECK11-NEXT:    [[TMP27:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1504 // CHECK11-NEXT:    [[TMP28:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
1505 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP27]], [[TMP28]]
1506 // CHECK11-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
1507 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1508 // CHECK11:       omp.inner.for.end:
1509 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1510 // CHECK11:       omp.loop.exit:
1511 // CHECK11-NEXT:    [[TMP29:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1512 // CHECK11-NEXT:    [[TMP30:%.*]] = load i32, ptr [[TMP29]], align 4
1513 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP30]])
1514 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
1515 // CHECK11:       omp.precond.end:
1516 // CHECK11-NEXT:    ret void
1517 //
1518 //
1519 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.omp_outlined.omp_outlined
1520 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1521 // CHECK11-NEXT:  entry:
1522 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1523 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1524 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1525 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1526 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1527 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
1528 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1529 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1530 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1531 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1532 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1533 // CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
1534 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1535 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1536 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
1537 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1538 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1539 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
1540 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
1541 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1542 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1543 // CHECK11-NEXT:    [[I13:%.*]] = alloca i32, align 4
1544 // CHECK11-NEXT:    [[J14:%.*]] = alloca i32, align 4
1545 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1546 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1547 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1548 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1549 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
1550 // CHECK11-NEXT:    store i32 [[M]], ptr [[M_ADDR]], align 4
1551 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1552 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1553 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1554 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1555 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1556 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1557 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[N_ADDR]], align 4
1558 // CHECK11-NEXT:    store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR_]], align 4
1559 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[M_ADDR]], align 4
1560 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR_4]], align 4
1561 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1562 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
1563 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1564 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
1565 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1566 // CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP6]], 0
1567 // CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1568 // CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
1569 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
1570 // CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
1571 // CHECK11-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
1572 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
1573 // CHECK11-NEXT:    store i32 0, ptr [[J]], align 4
1574 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1575 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
1576 // CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
1577 // CHECK11:       land.lhs.true:
1578 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1579 // CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP8]]
1580 // CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
1581 // CHECK11:       omp.precond.then:
1582 // CHECK11-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
1583 // CHECK11-NEXT:    [[TMP9:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1584 // CHECK11-NEXT:    store i64 [[TMP9]], ptr [[DOTOMP_UB]], align 8
1585 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1586 // CHECK11-NEXT:    [[CONV11:%.*]] = zext i32 [[TMP10]] to i64
1587 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1588 // CHECK11-NEXT:    [[CONV12:%.*]] = zext i32 [[TMP11]] to i64
1589 // CHECK11-NEXT:    store i64 [[CONV11]], ptr [[DOTOMP_LB]], align 8
1590 // CHECK11-NEXT:    store i64 [[CONV12]], ptr [[DOTOMP_UB]], align 8
1591 // CHECK11-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
1592 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1593 // CHECK11-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1594 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
1595 // CHECK11-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP13]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
1596 // CHECK11-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
1597 // CHECK11-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1598 // CHECK11-NEXT:    [[CMP15:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
1599 // CHECK11-NEXT:    br i1 [[CMP15]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1600 // CHECK11:       cond.true:
1601 // CHECK11-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1602 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1603 // CHECK11:       cond.false:
1604 // CHECK11-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
1605 // CHECK11-NEXT:    br label [[COND_END]]
1606 // CHECK11:       cond.end:
1607 // CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
1608 // CHECK11-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
1609 // CHECK11-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
1610 // CHECK11-NEXT:    store i64 [[TMP18]], ptr [[DOTOMP_IV]], align 8
1611 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1612 // CHECK11:       omp.inner.for.cond:
1613 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1614 // CHECK11-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
1615 // CHECK11-NEXT:    [[CMP16:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
1616 // CHECK11-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1617 // CHECK11:       omp.inner.for.body:
1618 // CHECK11-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1619 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1620 // CHECK11-NEXT:    [[SUB17:%.*]] = sub nsw i32 [[TMP22]], 0
1621 // CHECK11-NEXT:    [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
1622 // CHECK11-NEXT:    [[MUL19:%.*]] = mul nsw i32 1, [[DIV18]]
1623 // CHECK11-NEXT:    [[CONV20:%.*]] = sext i32 [[MUL19]] to i64
1624 // CHECK11-NEXT:    [[DIV21:%.*]] = sdiv i64 [[TMP21]], [[CONV20]]
1625 // CHECK11-NEXT:    [[MUL22:%.*]] = mul nsw i64 [[DIV21]], 1
1626 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL22]]
1627 // CHECK11-NEXT:    [[CONV23:%.*]] = trunc i64 [[ADD]] to i32
1628 // CHECK11-NEXT:    store i32 [[CONV23]], ptr [[I13]], align 4
1629 // CHECK11-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1630 // CHECK11-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1631 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1632 // CHECK11-NEXT:    [[SUB24:%.*]] = sub nsw i32 [[TMP25]], 0
1633 // CHECK11-NEXT:    [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
1634 // CHECK11-NEXT:    [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
1635 // CHECK11-NEXT:    [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
1636 // CHECK11-NEXT:    [[DIV28:%.*]] = sdiv i64 [[TMP24]], [[CONV27]]
1637 // CHECK11-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1638 // CHECK11-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP26]], 0
1639 // CHECK11-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
1640 // CHECK11-NEXT:    [[MUL31:%.*]] = mul nsw i32 1, [[DIV30]]
1641 // CHECK11-NEXT:    [[CONV32:%.*]] = sext i32 [[MUL31]] to i64
1642 // CHECK11-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[DIV28]], [[CONV32]]
1643 // CHECK11-NEXT:    [[SUB34:%.*]] = sub nsw i64 [[TMP23]], [[MUL33]]
1644 // CHECK11-NEXT:    [[MUL35:%.*]] = mul nsw i64 [[SUB34]], 1
1645 // CHECK11-NEXT:    [[ADD36:%.*]] = add nsw i64 0, [[MUL35]]
1646 // CHECK11-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
1647 // CHECK11-NEXT:    store i32 [[CONV37]], ptr [[J14]], align 4
1648 // CHECK11-NEXT:    [[TMP27:%.*]] = load i32, ptr [[I13]], align 4
1649 // CHECK11-NEXT:    [[TMP28:%.*]] = mul nsw i32 [[TMP27]], [[TMP1]]
1650 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 [[TMP28]]
1651 // CHECK11-NEXT:    [[TMP29:%.*]] = load i32, ptr [[J14]], align 4
1652 // CHECK11-NEXT:    [[ARRAYIDX38:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 [[TMP29]]
1653 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX38]], align 4
1654 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1655 // CHECK11:       omp.body.continue:
1656 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1657 // CHECK11:       omp.inner.for.inc:
1658 // CHECK11-NEXT:    [[TMP30:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1659 // CHECK11-NEXT:    [[ADD39:%.*]] = add nsw i64 [[TMP30]], 1
1660 // CHECK11-NEXT:    store i64 [[ADD39]], ptr [[DOTOMP_IV]], align 8
1661 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1662 // CHECK11:       omp.inner.for.end:
1663 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1664 // CHECK11:       omp.loop.exit:
1665 // CHECK11-NEXT:    [[TMP31:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1666 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 4
1667 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP32]])
1668 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
1669 // CHECK11:       omp.precond.end:
1670 // CHECK11-NEXT:    ret void
1671 //
1672 //
1673 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1674 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1675 // CHECK11-NEXT:  entry:
1676 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1677 // CHECK11-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1678 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
1679 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
1680 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
1681 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1682 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1683 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1684 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1685 // CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1686 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP0]], align 4
1687 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1688 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
1689 // CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1690 // CHECK11-NEXT:    store ptr null, ptr [[TMP2]], align 4
1691 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1692 // CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1693 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1694 // CHECK11-NEXT:    store i32 3, ptr [[TMP5]], align 4
1695 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1696 // CHECK11-NEXT:    store i32 1, ptr [[TMP6]], align 4
1697 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1698 // CHECK11-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
1699 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1700 // CHECK11-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
1701 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1702 // CHECK11-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP9]], align 4
1703 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1704 // CHECK11-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP10]], align 4
1705 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1706 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
1707 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1708 // CHECK11-NEXT:    store ptr null, ptr [[TMP12]], align 4
1709 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1710 // CHECK11-NEXT:    store i64 20, ptr [[TMP13]], align 8
1711 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1712 // CHECK11-NEXT:    store i64 0, ptr [[TMP14]], align 8
1713 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1714 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
1715 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1716 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
1717 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1718 // CHECK11-NEXT:    store i32 0, ptr [[TMP17]], align 4
1719 // CHECK11-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.region_id, ptr [[KERNEL_ARGS]])
1720 // CHECK11-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1721 // CHECK11-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1722 // CHECK11:       omp_offload.failed:
1723 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68(ptr [[A]]) #[[ATTR3]]
1724 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1725 // CHECK11:       omp_offload.cont:
1726 // CHECK11-NEXT:    ret i32 0
1727 //
1728 //
1729 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68
1730 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1731 // CHECK11-NEXT:  entry:
1732 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1733 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1734 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1735 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined, ptr [[TMP0]])
1736 // CHECK11-NEXT:    ret void
1737 //
1738 //
1739 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined
1740 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1741 // CHECK11-NEXT:  entry:
1742 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1743 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1744 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1745 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1746 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1747 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1748 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1749 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1750 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1751 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1752 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1753 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1754 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1755 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1756 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1757 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1758 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1759 // CHECK11-NEXT:    store i32 19, ptr [[DOTOMP_COMB_UB]], align 4
1760 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1761 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1762 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1763 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1764 // 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)
1765 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1766 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
1767 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1768 // CHECK11:       cond.true:
1769 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1770 // CHECK11:       cond.false:
1771 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1772 // CHECK11-NEXT:    br label [[COND_END]]
1773 // CHECK11:       cond.end:
1774 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1775 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1776 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1777 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1778 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1779 // CHECK11:       omp.inner.for.cond:
1780 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1781 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1782 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1783 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1784 // CHECK11:       omp.inner.for.body:
1785 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1786 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1787 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
1788 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1789 // CHECK11:       omp.inner.for.inc:
1790 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1791 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1792 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1793 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1794 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1795 // CHECK11:       omp.inner.for.end:
1796 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1797 // CHECK11:       omp.loop.exit:
1798 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1799 // CHECK11-NEXT:    ret void
1800 //
1801 //
1802 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l68.omp_outlined.omp_outlined
1803 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1804 // CHECK11-NEXT:  entry:
1805 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1806 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1807 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1808 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1809 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1810 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1811 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1812 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1813 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1814 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1815 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1816 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1817 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1818 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1819 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1820 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1821 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1822 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1823 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1824 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1825 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1826 // CHECK11-NEXT:    store i32 19, ptr [[DOTOMP_UB]], align 4
1827 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1828 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1829 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
1830 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
1831 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1832 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1833 // CHECK11-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1834 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
1835 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1836 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1837 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
1838 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1839 // CHECK11:       cond.true:
1840 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1841 // CHECK11:       cond.false:
1842 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1843 // CHECK11-NEXT:    br label [[COND_END]]
1844 // CHECK11:       cond.end:
1845 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1846 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1847 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1848 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
1849 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1850 // CHECK11:       omp.inner.for.cond:
1851 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1852 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1853 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1854 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1855 // CHECK11:       omp.inner.for.body:
1856 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1857 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
1858 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1859 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1860 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1861 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1862 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1863 // CHECK11-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 2
1864 // CHECK11-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
1865 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
1866 // CHECK11-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
1867 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
1868 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
1869 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
1870 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i32 0, i32 [[TMP13]]
1871 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
1872 // CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP14]]
1873 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
1874 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1875 // CHECK11:       omp.body.continue:
1876 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1877 // CHECK11:       omp.inner.for.inc:
1878 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1879 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
1880 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
1881 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1882 // CHECK11:       omp.inner.for.end:
1883 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1884 // CHECK11:       omp.loop.exit:
1885 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
1886 // CHECK11-NEXT:    ret void
1887 //