clang/test/OpenMP/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
  29     #pragma omp teams distribute parallel for collapse(2)
  30     for(int i = 0; i < X; i++) {
  31       for(int j = 0; j < Y; j++) {
  32         a[i][j] = (T)0;
  33       }
  34     }
  35
  36     // discard loop variables not needed here
  37
  38
  39     return a[0][0];
  40   }
  41 };
  42
  43 int teams_template_struct(void) {
  44   SS<int, 123, 456> V;
  45   return V.foo();
  46
  47 }
  48 #endif // CK1
  49
  50 // Test host codegen.
  51 // RUN: %clang_cc1 -DCK2 -verify -Wno-vla -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
  52 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  53 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
  54 // 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
  55 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  56 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
  57
  58 // 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}}"
  59 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  60 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  61 // 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}}"
  62 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  63 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify -Wno-vla %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  64 #ifdef CK2
  65
  66 template <typename T, int n, int m>
  67 int tmain(T argc) {
  68   T a[n][m];
  69   #pragma omp target
  70   #pragma omp teams distribute parallel for collapse(2)
  71   for(int i = 0; i < n; i++) {
  72     for(int j = 0; j < m; j++) {
  73       a[i][j] = (T)0;
  74     }
  75   }
  76   return 0;
  77 }
  78
  79 int main (int argc, char **argv) {
  80   int n = 100;
  81   int m = 2;
  82   int a[n][m];
  83   #pragma omp target
  84   #pragma omp teams distribute parallel for collapse(2)
  85   for(int i = 0; i < n; i++) {
  86     for(int j = 0; j < m; j++) {
  87       a[i][j] = 0;
  88     }
  89   }
  90   return tmain<int, 10, 2>(argc);
  91 }
  92
  93
  94
  95
  96
  97
  98
  99
 100 // discard loop variables not needed here
 101
 102
 103 #endif // CK2
 104 #endif // #ifndef HEADER
 105 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 106 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
 107 // CHECK1-NEXT:  entry:
 108 // CHECK1-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 109 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(224352) [[V]])
 110 // CHECK1-NEXT:    ret i32 [[CALL]]
 111 //
 112 //
 113 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 114 // CHECK1-SAME: (ptr noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat {
 115 // CHECK1-NEXT:  entry:
 116 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 117 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
 118 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
 119 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
 120 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 121 // CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 122 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 123 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 124 // CHECK1-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 125 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 126 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 127 // CHECK1-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 8
 128 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 129 // CHECK1-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
 130 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 131 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 132 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 133 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 134 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 135 // CHECK1-NEXT:    store i32 3, ptr [[TMP5]], align 4
 136 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 137 // CHECK1-NEXT:    store i32 1, ptr [[TMP6]], align 4
 138 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 139 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
 140 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 141 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
 142 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 143 // CHECK1-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 8
 144 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 145 // CHECK1-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 8
 146 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 147 // CHECK1-NEXT:    store ptr null, ptr [[TMP11]], align 8
 148 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 149 // CHECK1-NEXT:    store ptr null, ptr [[TMP12]], align 8
 150 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 151 // CHECK1-NEXT:    store i64 56088, ptr [[TMP13]], align 8
 152 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 153 // CHECK1-NEXT:    store i64 0, ptr [[TMP14]], align 8
 154 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 155 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 156 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 157 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 158 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 159 // CHECK1-NEXT:    store i32 0, ptr [[TMP17]], align 4
 160 // CHECK1-NEXT:    [[TMP18:%.*]] = call i32 @__tgt_target_kernel(ptr @[[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]])
 161 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 162 // CHECK1-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 163 // CHECK1:       omp_offload.failed:
 164 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 165 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 166 // CHECK1:       omp_offload.cont:
 167 // CHECK1-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 168 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A3]], i64 0, i64 0
 169 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 0
 170 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
 171 // CHECK1-NEXT:    ret i32 [[TMP20]]
 172 //
 173 //
 174 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 175 // CHECK1-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 176 // CHECK1-NEXT:  entry:
 177 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 178 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 179 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 180 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined, ptr [[TMP0]])
 181 // CHECK1-NEXT:    ret void
 182 //
 183 //
 184 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 185 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 186 // CHECK1-NEXT:  entry:
 187 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 188 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 189 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 190 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 191 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 192 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 193 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 194 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 195 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 196 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 197 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 198 // CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
 199 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 200 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 201 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 202 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 203 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 204 // CHECK1-NEXT:    store i32 56087, ptr [[DOTOMP_COMB_UB]], align 4
 205 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 206 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 207 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 208 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 209 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 210 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 211 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
 212 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 213 // CHECK1:       cond.true:
 214 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 215 // CHECK1:       cond.false:
 216 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 217 // CHECK1-NEXT:    br label [[COND_END]]
 218 // CHECK1:       cond.end:
 219 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 220 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 221 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 222 // CHECK1-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 223 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 224 // CHECK1:       omp.inner.for.cond:
 225 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 226 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 227 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 228 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 229 // CHECK1:       omp.inner.for.body:
 230 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 231 // CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
 232 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 233 // CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
 234 // 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]])
 235 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 236 // CHECK1:       omp.inner.for.inc:
 237 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 238 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 239 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
 240 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 241 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 242 // CHECK1:       omp.inner.for.end:
 243 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 244 // CHECK1:       omp.loop.exit:
 245 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 246 // CHECK1-NEXT:    ret void
 247 //
 248 //
 249 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined
 250 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 251 // CHECK1-NEXT:  entry:
 252 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 253 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 254 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 255 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 256 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
 257 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 258 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 259 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 260 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 261 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 262 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 263 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 264 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 265 // CHECK1-NEXT:    [[J:%.*]] = alloca i32, align 4
 266 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 267 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 268 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 269 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 270 // CHECK1-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
 271 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
 272 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 273 // CHECK1-NEXT:    store i32 56087, ptr [[DOTOMP_UB]], align 4
 274 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 275 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
 276 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 277 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
 278 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 279 // CHECK1-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
 280 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 281 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 282 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 283 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 284 // 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)
 285 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 286 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 56087
 287 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 288 // CHECK1:       cond.true:
 289 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 290 // CHECK1:       cond.false:
 291 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 292 // CHECK1-NEXT:    br label [[COND_END]]
 293 // CHECK1:       cond.end:
 294 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 295 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 296 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 297 // CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 298 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 299 // CHECK1:       omp.inner.for.cond:
 300 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 301 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 302 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 303 // CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 304 // CHECK1:       omp.inner.for.body:
 305 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 306 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
 307 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 308 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 309 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 310 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 311 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 312 // CHECK1-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 456
 313 // CHECK1-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 456
 314 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
 315 // CHECK1-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
 316 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
 317 // CHECK1-NEXT:    store i32 [[ADD7]], ptr [[J]], align 4
 318 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 319 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
 320 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
 321 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i64 0, i64 [[IDXPROM]]
 322 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
 323 // CHECK1-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
 324 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
 325 // CHECK1-NEXT:    store i32 0, ptr [[ARRAYIDX9]], align 4
 326 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 327 // CHECK1:       omp.body.continue:
 328 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 329 // CHECK1:       omp.inner.for.inc:
 330 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 331 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
 332 // CHECK1-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
 333 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 334 // CHECK1:       omp.inner.for.end:
 335 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 336 // CHECK1:       omp.loop.exit:
 337 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
 338 // CHECK1-NEXT:    ret void
 339 //
 340 //
 341 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
 342 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
 343 // CHECK3-NEXT:  entry:
 344 // CHECK3-NEXT:    [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
 345 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(ptr noundef nonnull align 4 dereferenceable(224352) [[V]])
 346 // CHECK3-NEXT:    ret i32 [[CALL]]
 347 //
 348 //
 349 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
 350 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
 351 // CHECK3-NEXT:  entry:
 352 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 353 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
 354 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
 355 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
 356 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 357 // CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 358 // CHECK3-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 359 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 360 // CHECK3-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 361 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[THIS1]], i32 0, i32 0
 362 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 363 // CHECK3-NEXT:    store ptr [[THIS1]], ptr [[TMP0]], align 4
 364 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 365 // CHECK3-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
 366 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 367 // CHECK3-NEXT:    store ptr null, ptr [[TMP2]], align 4
 368 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 369 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 370 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 371 // CHECK3-NEXT:    store i32 3, ptr [[TMP5]], align 4
 372 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 373 // CHECK3-NEXT:    store i32 1, ptr [[TMP6]], align 4
 374 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 375 // CHECK3-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
 376 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 377 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
 378 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 379 // CHECK3-NEXT:    store ptr @.offload_sizes, ptr [[TMP9]], align 4
 380 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 381 // CHECK3-NEXT:    store ptr @.offload_maptypes, ptr [[TMP10]], align 4
 382 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 383 // CHECK3-NEXT:    store ptr null, ptr [[TMP11]], align 4
 384 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 385 // CHECK3-NEXT:    store ptr null, ptr [[TMP12]], align 4
 386 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 387 // CHECK3-NEXT:    store i64 56088, ptr [[TMP13]], align 8
 388 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 389 // CHECK3-NEXT:    store i64 0, ptr [[TMP14]], align 8
 390 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 391 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
 392 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 393 // CHECK3-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
 394 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 395 // CHECK3-NEXT:    store i32 0, ptr [[TMP17]], align 4
 396 // 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]])
 397 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 398 // CHECK3-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 399 // CHECK3:       omp_offload.failed:
 400 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(ptr [[THIS1]]) #[[ATTR2:[0-9]+]]
 401 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 402 // CHECK3:       omp_offload.cont:
 403 // CHECK3-NEXT:    [[A3:%.*]] = getelementptr inbounds nuw [[STRUCT_SS]], ptr [[THIS1]], i32 0, i32 0
 404 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A3]], i32 0, i32 0
 405 // CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 0
 406 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
 407 // CHECK3-NEXT:    ret i32 [[TMP20]]
 408 //
 409 //
 410 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
 411 // CHECK3-SAME: (ptr noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
 412 // CHECK3-NEXT:  entry:
 413 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 414 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 415 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 416 // 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]])
 417 // CHECK3-NEXT:    ret void
 418 //
 419 //
 420 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined
 421 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 422 // CHECK3-NEXT:  entry:
 423 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 424 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 425 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 426 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 427 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 428 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 429 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 430 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 431 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 432 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 433 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 434 // CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
 435 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 436 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 437 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 438 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 439 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 440 // CHECK3-NEXT:    store i32 56087, ptr [[DOTOMP_COMB_UB]], align 4
 441 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 442 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 443 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 444 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 445 // 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)
 446 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 447 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
 448 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 449 // CHECK3:       cond.true:
 450 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 451 // CHECK3:       cond.false:
 452 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 453 // CHECK3-NEXT:    br label [[COND_END]]
 454 // CHECK3:       cond.end:
 455 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
 456 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 457 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 458 // CHECK3-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
 459 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 460 // CHECK3:       omp.inner.for.cond:
 461 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 462 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 463 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
 464 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 465 // CHECK3:       omp.inner.for.body:
 466 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 467 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 468 // 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]])
 469 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 470 // CHECK3:       omp.inner.for.inc:
 471 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 472 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 473 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
 474 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 475 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 476 // CHECK3:       omp.inner.for.end:
 477 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 478 // CHECK3:       omp.loop.exit:
 479 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
 480 // CHECK3-NEXT:    ret void
 481 //
 482 //
 483 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.omp_outlined.omp_outlined
 484 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef [[THIS:%.*]]) #[[ATTR1]] {
 485 // CHECK3-NEXT:  entry:
 486 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 487 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 488 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
 489 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
 490 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
 491 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 492 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 493 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 494 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 495 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 496 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 497 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 498 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
 499 // CHECK3-NEXT:    [[J:%.*]] = alloca i32, align 4
 500 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 501 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 502 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 503 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 504 // CHECK3-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
 505 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
 506 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 507 // CHECK3-NEXT:    store i32 56087, ptr [[DOTOMP_UB]], align 4
 508 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
 509 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
 510 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
 511 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
 512 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 513 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 514 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 515 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 516 // 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)
 517 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 518 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 56087
 519 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 520 // CHECK3:       cond.true:
 521 // CHECK3-NEXT:    br label [[COND_END:%.*]]
 522 // CHECK3:       cond.false:
 523 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 524 // CHECK3-NEXT:    br label [[COND_END]]
 525 // CHECK3:       cond.end:
 526 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 527 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 528 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 529 // CHECK3-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 530 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 531 // CHECK3:       omp.inner.for.cond:
 532 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 533 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 534 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 535 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 536 // CHECK3:       omp.inner.for.body:
 537 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 538 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 456
 539 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 540 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 541 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 542 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 543 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 544 // CHECK3-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 456
 545 // CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
 546 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
 547 // CHECK3-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
 548 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
 549 // CHECK3-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
 550 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds nuw [[STRUCT_SS:%.*]], ptr [[TMP0]], i32 0, i32 0
 551 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
 552 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], ptr [[A]], i32 0, i32 [[TMP13]]
 553 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
 554 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP14]]
 555 // CHECK3-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
 556 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 557 // CHECK3:       omp.body.continue:
 558 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 559 // CHECK3:       omp.inner.for.inc:
 560 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 561 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
 562 // CHECK3-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
 563 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
 564 // CHECK3:       omp.inner.for.end:
 565 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 566 // CHECK3:       omp.loop.exit:
 567 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
 568 // CHECK3-NEXT:    ret void
 569 //
 570 //
 571 // CHECK9-LABEL: define {{[^@]+}}@main
 572 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
 573 // CHECK9-NEXT:  entry:
 574 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
 575 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
 576 // CHECK9-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
 577 // CHECK9-NEXT:    [[N:%.*]] = alloca i32, align 4
 578 // CHECK9-NEXT:    [[M:%.*]] = alloca i32, align 4
 579 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
 580 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
 581 // CHECK9-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
 582 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
 583 // CHECK9-NEXT:    [[M_CASTED:%.*]] = alloca i64, align 8
 584 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 8
 585 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 8
 586 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 8
 587 // CHECK9-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
 588 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 589 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 590 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 591 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
 592 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
 593 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 594 // CHECK9-NEXT:    store i32 0, ptr [[RETVAL]], align 4
 595 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
 596 // CHECK9-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
 597 // CHECK9-NEXT:    store i32 100, ptr [[N]], align 4
 598 // CHECK9-NEXT:    store i32 2, ptr [[M]], align 4
 599 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
 600 // CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
 601 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[M]], align 4
 602 // CHECK9-NEXT:    [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
 603 // CHECK9-NEXT:    [[TMP4:%.*]] = call ptr @llvm.stacksave.p0()
 604 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[SAVED_STACK]], align 8
 605 // CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
 606 // CHECK9-NEXT:    [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 4
 607 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[__VLA_EXPR0]], align 8
 608 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[__VLA_EXPR1]], align 8
 609 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[N]], align 4
 610 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[N_CASTED]], align 4
 611 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, ptr [[N_CASTED]], align 8
 612 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[M]], align 4
 613 // CHECK9-NEXT:    store i32 [[TMP8]], ptr [[M_CASTED]], align 4
 614 // CHECK9-NEXT:    [[TMP9:%.*]] = load i64, ptr [[M_CASTED]], align 8
 615 // CHECK9-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
 616 // CHECK9-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
 617 // CHECK9-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[DOTOFFLOAD_SIZES]], ptr align 8 @.offload_sizes, i64 40, i1 false)
 618 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 619 // CHECK9-NEXT:    store i64 [[TMP7]], ptr [[TMP12]], align 8
 620 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 621 // CHECK9-NEXT:    store i64 [[TMP7]], ptr [[TMP13]], align 8
 622 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
 623 // CHECK9-NEXT:    store ptr null, ptr [[TMP14]], align 8
 624 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
 625 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[TMP15]], align 8
 626 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
 627 // CHECK9-NEXT:    store i64 [[TMP9]], ptr [[TMP16]], align 8
 628 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
 629 // CHECK9-NEXT:    store ptr null, ptr [[TMP17]], align 8
 630 // CHECK9-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
 631 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP18]], align 8
 632 // CHECK9-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
 633 // CHECK9-NEXT:    store i64 [[TMP1]], ptr [[TMP19]], align 8
 634 // CHECK9-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
 635 // CHECK9-NEXT:    store ptr null, ptr [[TMP20]], align 8
 636 // CHECK9-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
 637 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP21]], align 8
 638 // CHECK9-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
 639 // CHECK9-NEXT:    store i64 [[TMP3]], ptr [[TMP22]], align 8
 640 // CHECK9-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
 641 // CHECK9-NEXT:    store ptr null, ptr [[TMP23]], align 8
 642 // CHECK9-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
 643 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP24]], align 8
 644 // CHECK9-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
 645 // CHECK9-NEXT:    store ptr [[VLA]], ptr [[TMP25]], align 8
 646 // CHECK9-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 4
 647 // CHECK9-NEXT:    store i64 [[TMP11]], ptr [[TMP26]], align 8
 648 // CHECK9-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
 649 // CHECK9-NEXT:    store ptr null, ptr [[TMP27]], align 8
 650 // CHECK9-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 651 // CHECK9-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 652 // CHECK9-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
 653 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, ptr [[N]], align 4
 654 // CHECK9-NEXT:    store i32 [[TMP31]], ptr [[DOTCAPTURE_EXPR_]], align 4
 655 // CHECK9-NEXT:    [[TMP32:%.*]] = load i32, ptr [[M]], align 4
 656 // CHECK9-NEXT:    store i32 [[TMP32]], ptr [[DOTCAPTURE_EXPR_2]], align 4
 657 // CHECK9-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 658 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP33]], 0
 659 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 660 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 661 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
 662 // CHECK9-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[TMP34]], 0
 663 // CHECK9-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
 664 // CHECK9-NEXT:    [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
 665 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
 666 // CHECK9-NEXT:    [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
 667 // CHECK9-NEXT:    store i64 [[SUB7]], ptr [[DOTCAPTURE_EXPR_3]], align 8
 668 // CHECK9-NEXT:    [[TMP35:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8
 669 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP35]], 1
 670 // CHECK9-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 671 // CHECK9-NEXT:    store i32 3, ptr [[TMP36]], align 4
 672 // CHECK9-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 673 // CHECK9-NEXT:    store i32 5, ptr [[TMP37]], align 4
 674 // CHECK9-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 675 // CHECK9-NEXT:    store ptr [[TMP28]], ptr [[TMP38]], align 8
 676 // CHECK9-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 677 // CHECK9-NEXT:    store ptr [[TMP29]], ptr [[TMP39]], align 8
 678 // CHECK9-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 679 // CHECK9-NEXT:    store ptr [[TMP30]], ptr [[TMP40]], align 8
 680 // CHECK9-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 681 // CHECK9-NEXT:    store ptr @.offload_maptypes, ptr [[TMP41]], align 8
 682 // CHECK9-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 683 // CHECK9-NEXT:    store ptr null, ptr [[TMP42]], align 8
 684 // CHECK9-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 685 // CHECK9-NEXT:    store ptr null, ptr [[TMP43]], align 8
 686 // CHECK9-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 687 // CHECK9-NEXT:    store i64 [[ADD]], ptr [[TMP44]], align 8
 688 // CHECK9-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 689 // CHECK9-NEXT:    store i64 0, ptr [[TMP45]], align 8
 690 // CHECK9-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 691 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP46]], align 4
 692 // CHECK9-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 693 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP47]], align 4
 694 // CHECK9-NEXT:    [[TMP48:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 695 // CHECK9-NEXT:    store i32 0, ptr [[TMP48]], align 4
 696 // 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_l83.region_id, ptr [[KERNEL_ARGS]])
 697 // CHECK9-NEXT:    [[TMP50:%.*]] = icmp ne i32 [[TMP49]], 0
 698 // CHECK9-NEXT:    br i1 [[TMP50]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 699 // CHECK9:       omp_offload.failed:
 700 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83(i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP1]], i64 [[TMP3]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
 701 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 702 // CHECK9:       omp_offload.cont:
 703 // CHECK9-NEXT:    [[TMP51:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
 704 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef signext [[TMP51]])
 705 // CHECK9-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
 706 // CHECK9-NEXT:    [[TMP52:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
 707 // CHECK9-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP52]])
 708 // CHECK9-NEXT:    [[TMP53:%.*]] = load i32, ptr [[RETVAL]], align 4
 709 // CHECK9-NEXT:    ret i32 [[TMP53]]
 710 //
 711 //
 712 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83
 713 // 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]+]] {
 714 // CHECK9-NEXT:  entry:
 715 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
 716 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca i64, align 8
 717 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 718 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 719 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 720 // CHECK9-NEXT:    store i64 [[N]], ptr [[N_ADDR]], align 8
 721 // CHECK9-NEXT:    store i64 [[M]], ptr [[M_ADDR]], align 8
 722 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 723 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 724 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 725 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 726 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 727 // CHECK9-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 728 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i64 [[TMP0]], i64 [[TMP1]], ptr [[TMP2]])
 729 // CHECK9-NEXT:    ret void
 730 //
 731 //
 732 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined
 733 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
 734 // CHECK9-NEXT:  entry:
 735 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 736 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 737 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
 738 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 8
 739 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 740 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 741 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 742 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 743 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 744 // CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
 745 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 746 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
 747 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
 748 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 749 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
 750 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
 751 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
 752 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 753 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 754 // CHECK9-NEXT:    [[I11:%.*]] = alloca i32, align 4
 755 // CHECK9-NEXT:    [[J12:%.*]] = alloca i32, align 4
 756 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 757 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 758 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
 759 // CHECK9-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 8
 760 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 761 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 762 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 763 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
 764 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 8
 765 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 766 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 767 // CHECK9-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 768 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
 769 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
 770 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
 771 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
 772 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 773 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
 774 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 775 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 776 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 777 // CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
 778 // CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 779 // CHECK9-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
 780 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
 781 // CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
 782 // CHECK9-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
 783 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
 784 // CHECK9-NEXT:    store i32 0, ptr [[J]], align 4
 785 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 786 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
 787 // CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
 788 // CHECK9:       land.lhs.true:
 789 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 790 // CHECK9-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
 791 // CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
 792 // CHECK9:       omp.precond.then:
 793 // CHECK9-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
 794 // CHECK9-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 795 // CHECK9-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_COMB_UB]], align 8
 796 // CHECK9-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
 797 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 798 // CHECK9-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 799 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
 800 // CHECK9-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP13]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
 801 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 802 // CHECK9-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 803 // CHECK9-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
 804 // CHECK9-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 805 // CHECK9:       cond.true:
 806 // CHECK9-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 807 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 808 // CHECK9:       cond.false:
 809 // CHECK9-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 810 // CHECK9-NEXT:    br label [[COND_END]]
 811 // CHECK9:       cond.end:
 812 // CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
 813 // CHECK9-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
 814 // CHECK9-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
 815 // CHECK9-NEXT:    store i64 [[TMP18]], ptr [[DOTOMP_IV]], align 8
 816 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 817 // CHECK9:       omp.inner.for.cond:
 818 // CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 819 // CHECK9-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 820 // CHECK9-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
 821 // CHECK9-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 822 // CHECK9:       omp.inner.for.body:
 823 // CHECK9-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
 824 // CHECK9-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 825 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined, i64 [[TMP21]], i64 [[TMP22]], ptr [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], i64 [[TMP3]], ptr [[TMP4]])
 826 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 827 // CHECK9:       omp.inner.for.inc:
 828 // CHECK9-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 829 // CHECK9-NEXT:    [[TMP24:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
 830 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP23]], [[TMP24]]
 831 // CHECK9-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
 832 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 833 // CHECK9:       omp.inner.for.end:
 834 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 835 // CHECK9:       omp.loop.exit:
 836 // CHECK9-NEXT:    [[TMP25:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 837 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32, ptr [[TMP25]], align 4
 838 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP26]])
 839 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 840 // CHECK9:       omp.precond.end:
 841 // CHECK9-NEXT:    ret void
 842 //
 843 //
 844 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined
 845 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
 846 // CHECK9-NEXT:  entry:
 847 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 848 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 849 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 850 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 851 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 8
 852 // CHECK9-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 8
 853 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
 854 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
 855 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 856 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 857 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 858 // CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
 859 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
 860 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
 861 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
 862 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
 863 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
 864 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 865 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 866 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 867 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 868 // CHECK9-NEXT:    [[I11:%.*]] = alloca i32, align 4
 869 // CHECK9-NEXT:    [[J12:%.*]] = alloca i32, align 4
 870 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 871 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 872 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 873 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 874 // CHECK9-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 8
 875 // CHECK9-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 8
 876 // CHECK9-NEXT:    store i64 [[VLA]], ptr [[VLA_ADDR]], align 8
 877 // CHECK9-NEXT:    store i64 [[VLA1]], ptr [[VLA_ADDR2]], align 8
 878 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 879 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 8
 880 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 8
 881 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR]], align 8
 882 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, ptr [[VLA_ADDR2]], align 8
 883 // CHECK9-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 884 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
 885 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
 886 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
 887 // CHECK9-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
 888 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 889 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
 890 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
 891 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
 892 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 893 // CHECK9-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
 894 // CHECK9-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
 895 // CHECK9-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
 896 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
 897 // CHECK9-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
 898 // CHECK9-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
 899 // CHECK9-NEXT:    store i32 0, ptr [[I]], align 4
 900 // CHECK9-NEXT:    store i32 0, ptr [[J]], align 4
 901 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
 902 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
 903 // CHECK9-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
 904 // CHECK9:       land.lhs.true:
 905 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 906 // CHECK9-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
 907 // CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
 908 // CHECK9:       omp.precond.then:
 909 // CHECK9-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
 910 // CHECK9-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 911 // CHECK9-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_UB]], align 8
 912 // CHECK9-NEXT:    [[TMP12:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 913 // CHECK9-NEXT:    [[TMP13:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 914 // CHECK9-NEXT:    store i64 [[TMP12]], ptr [[DOTOMP_LB]], align 8
 915 // CHECK9-NEXT:    store i64 [[TMP13]], ptr [[DOTOMP_UB]], align 8
 916 // CHECK9-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
 917 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 918 // CHECK9-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 919 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
 920 // CHECK9-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
 921 // CHECK9-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 922 // CHECK9-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 923 // CHECK9-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP16]], [[TMP17]]
 924 // CHECK9-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 925 // CHECK9:       cond.true:
 926 // CHECK9-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
 927 // CHECK9-NEXT:    br label [[COND_END:%.*]]
 928 // CHECK9:       cond.false:
 929 // CHECK9-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 930 // CHECK9-NEXT:    br label [[COND_END]]
 931 // CHECK9:       cond.end:
 932 // CHECK9-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP18]], [[COND_TRUE]] ], [ [[TMP19]], [[COND_FALSE]] ]
 933 // CHECK9-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
 934 // CHECK9-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
 935 // CHECK9-NEXT:    store i64 [[TMP20]], ptr [[DOTOMP_IV]], align 8
 936 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 937 // CHECK9:       omp.inner.for.cond:
 938 // CHECK9-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 939 // CHECK9-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 940 // CHECK9-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP21]], [[TMP22]]
 941 // CHECK9-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 942 // CHECK9:       omp.inner.for.body:
 943 // CHECK9-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 944 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 945 // CHECK9-NEXT:    [[SUB15:%.*]] = sub nsw i32 [[TMP24]], 0
 946 // CHECK9-NEXT:    [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
 947 // CHECK9-NEXT:    [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
 948 // CHECK9-NEXT:    [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
 949 // CHECK9-NEXT:    [[DIV19:%.*]] = sdiv i64 [[TMP23]], [[CONV18]]
 950 // CHECK9-NEXT:    [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
 951 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
 952 // CHECK9-NEXT:    [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
 953 // CHECK9-NEXT:    store i32 [[CONV21]], ptr [[I11]], align 4
 954 // CHECK9-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 955 // CHECK9-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 956 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 957 // CHECK9-NEXT:    [[SUB22:%.*]] = sub nsw i32 [[TMP27]], 0
 958 // CHECK9-NEXT:    [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
 959 // CHECK9-NEXT:    [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
 960 // CHECK9-NEXT:    [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
 961 // CHECK9-NEXT:    [[DIV26:%.*]] = sdiv i64 [[TMP26]], [[CONV25]]
 962 // CHECK9-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
 963 // CHECK9-NEXT:    [[SUB27:%.*]] = sub nsw i32 [[TMP28]], 0
 964 // CHECK9-NEXT:    [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
 965 // CHECK9-NEXT:    [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
 966 // CHECK9-NEXT:    [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
 967 // CHECK9-NEXT:    [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
 968 // CHECK9-NEXT:    [[SUB32:%.*]] = sub nsw i64 [[TMP25]], [[MUL31]]
 969 // CHECK9-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
 970 // CHECK9-NEXT:    [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
 971 // CHECK9-NEXT:    [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
 972 // CHECK9-NEXT:    store i32 [[CONV35]], ptr [[J12]], align 4
 973 // CHECK9-NEXT:    [[TMP29:%.*]] = load i32, ptr [[I11]], align 4
 974 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP29]] to i64
 975 // CHECK9-NEXT:    [[TMP30:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP3]]
 976 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i64 [[TMP30]]
 977 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J12]], align 4
 978 // CHECK9-NEXT:    [[IDXPROM36:%.*]] = sext i32 [[TMP31]] to i64
 979 // CHECK9-NEXT:    [[ARRAYIDX37:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i64 [[IDXPROM36]]
 980 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX37]], align 4
 981 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 982 // CHECK9:       omp.body.continue:
 983 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 984 // CHECK9:       omp.inner.for.inc:
 985 // CHECK9-NEXT:    [[TMP32:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 986 // CHECK9-NEXT:    [[ADD38:%.*]] = add nsw i64 [[TMP32]], 1
 987 // CHECK9-NEXT:    store i64 [[ADD38]], ptr [[DOTOMP_IV]], align 8
 988 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
 989 // CHECK9:       omp.inner.for.end:
 990 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 991 // CHECK9:       omp.loop.exit:
 992 // CHECK9-NEXT:    [[TMP33:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 993 // CHECK9-NEXT:    [[TMP34:%.*]] = load i32, ptr [[TMP33]], align 4
 994 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP34]])
 995 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
 996 // CHECK9:       omp.precond.end:
 997 // CHECK9-NEXT:    ret void
 998 //
 999 //
1000 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1001 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1002 // CHECK9-NEXT:  entry:
1003 // CHECK9-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1004 // CHECK9-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1005 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
1006 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
1007 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
1008 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1009 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1010 // CHECK9-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1011 // CHECK9-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1012 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1013 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP0]], align 8
1014 // CHECK9-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1015 // CHECK9-NEXT:    store ptr [[A]], ptr [[TMP1]], align 8
1016 // CHECK9-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1017 // CHECK9-NEXT:    store ptr null, ptr [[TMP2]], align 8
1018 // CHECK9-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1019 // CHECK9-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1020 // CHECK9-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1021 // CHECK9-NEXT:    store i32 3, ptr [[TMP5]], align 4
1022 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1023 // CHECK9-NEXT:    store i32 1, ptr [[TMP6]], align 4
1024 // CHECK9-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1025 // CHECK9-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 8
1026 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1027 // CHECK9-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 8
1028 // CHECK9-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1029 // CHECK9-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP9]], align 8
1030 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1031 // CHECK9-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP10]], align 8
1032 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1033 // CHECK9-NEXT:    store ptr null, ptr [[TMP11]], align 8
1034 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1035 // CHECK9-NEXT:    store ptr null, ptr [[TMP12]], align 8
1036 // CHECK9-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1037 // CHECK9-NEXT:    store i64 20, ptr [[TMP13]], align 8
1038 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1039 // CHECK9-NEXT:    store i64 0, ptr [[TMP14]], align 8
1040 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1041 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
1042 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1043 // CHECK9-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
1044 // CHECK9-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1045 // CHECK9-NEXT:    store i32 0, ptr [[TMP17]], align 4
1046 // 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__l69.region_id, ptr [[KERNEL_ARGS]])
1047 // CHECK9-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1048 // CHECK9-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1049 // CHECK9:       omp_offload.failed:
1050 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69(ptr [[A]]) #[[ATTR3]]
1051 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1052 // CHECK9:       omp_offload.cont:
1053 // CHECK9-NEXT:    ret i32 0
1054 //
1055 //
1056 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69
1057 // CHECK9-SAME: (ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1058 // CHECK9-NEXT:  entry:
1059 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1060 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1061 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1062 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
1063 // CHECK9-NEXT:    ret void
1064 //
1065 //
1066 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined
1067 // CHECK9-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1068 // CHECK9-NEXT:  entry:
1069 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1070 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1071 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1072 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1073 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1074 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1075 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1076 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1077 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1078 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1079 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1080 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
1081 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1082 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1083 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1084 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1085 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1086 // CHECK9-NEXT:    store i32 19, ptr [[DOTOMP_COMB_UB]], align 4
1087 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1088 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1089 // CHECK9-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1090 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1091 // 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)
1092 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1093 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
1094 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1095 // CHECK9:       cond.true:
1096 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1097 // CHECK9:       cond.false:
1098 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1099 // CHECK9-NEXT:    br label [[COND_END]]
1100 // CHECK9:       cond.end:
1101 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1102 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1103 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1104 // CHECK9-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1105 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1106 // CHECK9:       omp.inner.for.cond:
1107 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1108 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1109 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1110 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1111 // CHECK9:       omp.inner.for.body:
1112 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1113 // CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1114 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1115 // CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1116 // CHECK9-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined, i64 [[TMP9]], i64 [[TMP11]], ptr [[TMP0]])
1117 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1118 // CHECK9:       omp.inner.for.inc:
1119 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1120 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1121 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1122 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1123 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1124 // CHECK9:       omp.inner.for.end:
1125 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1126 // CHECK9:       omp.loop.exit:
1127 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1128 // CHECK9-NEXT:    ret void
1129 //
1130 //
1131 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined
1132 // 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]] {
1133 // CHECK9-NEXT:  entry:
1134 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1135 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1136 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1137 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1138 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
1139 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1140 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1141 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1142 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1143 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1144 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1145 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1146 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1147 // CHECK9-NEXT:    [[J:%.*]] = alloca i32, align 4
1148 // CHECK9-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1149 // CHECK9-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1150 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1151 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1152 // CHECK9-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
1153 // CHECK9-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
1154 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1155 // CHECK9-NEXT:    store i32 19, ptr [[DOTOMP_UB]], align 4
1156 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1157 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1158 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1159 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
1160 // CHECK9-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1161 // CHECK9-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
1162 // CHECK9-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1163 // CHECK9-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1164 // CHECK9-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1165 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
1166 // 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)
1167 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1168 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
1169 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1170 // CHECK9:       cond.true:
1171 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1172 // CHECK9:       cond.false:
1173 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1174 // CHECK9-NEXT:    br label [[COND_END]]
1175 // CHECK9:       cond.end:
1176 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1177 // CHECK9-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1178 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1179 // CHECK9-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
1180 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1181 // CHECK9:       omp.inner.for.cond:
1182 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1183 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1184 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1185 // CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1186 // CHECK9:       omp.inner.for.body:
1187 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1188 // CHECK9-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
1189 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1190 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1191 // CHECK9-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1192 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1193 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1194 // CHECK9-NEXT:    [[DIV4:%.*]] = sdiv i32 [[TMP12]], 2
1195 // CHECK9-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[DIV4]], 2
1196 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL5]]
1197 // CHECK9-NEXT:    [[MUL6:%.*]] = mul nsw i32 [[SUB]], 1
1198 // CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 0, [[MUL6]]
1199 // CHECK9-NEXT:    store i32 [[ADD7]], ptr [[J]], align 4
1200 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
1201 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
1202 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i64 0, i64 [[IDXPROM]]
1203 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
1204 // CHECK9-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP14]] to i64
1205 // CHECK9-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM8]]
1206 // CHECK9-NEXT:    store i32 0, ptr [[ARRAYIDX9]], align 4
1207 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1208 // CHECK9:       omp.body.continue:
1209 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1210 // CHECK9:       omp.inner.for.inc:
1211 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1212 // CHECK9-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP15]], 1
1213 // CHECK9-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
1214 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1215 // CHECK9:       omp.inner.for.end:
1216 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1217 // CHECK9:       omp.loop.exit:
1218 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
1219 // CHECK9-NEXT:    ret void
1220 //
1221 //
1222 // CHECK11-LABEL: define {{[^@]+}}@main
1223 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1224 // CHECK11-NEXT:  entry:
1225 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1226 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1227 // CHECK11-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 4
1228 // CHECK11-NEXT:    [[N:%.*]] = alloca i32, align 4
1229 // CHECK11-NEXT:    [[M:%.*]] = alloca i32, align 4
1230 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 4
1231 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1232 // CHECK11-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
1233 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
1234 // CHECK11-NEXT:    [[M_CASTED:%.*]] = alloca i32, align 4
1235 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 4
1236 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 4
1237 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 4
1238 // CHECK11-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
1239 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1240 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1241 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1242 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1243 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
1244 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1245 // CHECK11-NEXT:    store i32 0, ptr [[RETVAL]], align 4
1246 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1247 // CHECK11-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 4
1248 // CHECK11-NEXT:    store i32 100, ptr [[N]], align 4
1249 // CHECK11-NEXT:    store i32 2, ptr [[M]], align 4
1250 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[N]], align 4
1251 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[M]], align 4
1252 // CHECK11-NEXT:    [[TMP2:%.*]] = call ptr @llvm.stacksave.p0()
1253 // CHECK11-NEXT:    store ptr [[TMP2]], ptr [[SAVED_STACK]], align 4
1254 // CHECK11-NEXT:    [[TMP3:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
1255 // CHECK11-NEXT:    [[VLA:%.*]] = alloca i32, i32 [[TMP3]], align 4
1256 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[__VLA_EXPR0]], align 4
1257 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[__VLA_EXPR1]], align 4
1258 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[N]], align 4
1259 // CHECK11-NEXT:    store i32 [[TMP4]], ptr [[N_CASTED]], align 4
1260 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[N_CASTED]], align 4
1261 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[M]], align 4
1262 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[M_CASTED]], align 4
1263 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[M_CASTED]], align 4
1264 // CHECK11-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
1265 // CHECK11-NEXT:    [[TMP9:%.*]] = mul nuw i32 [[TMP8]], 4
1266 // CHECK11-NEXT:    [[TMP10:%.*]] = sext i32 [[TMP9]] to i64
1267 // CHECK11-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[DOTOFFLOAD_SIZES]], ptr align 4 @.offload_sizes, i32 40, i1 false)
1268 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1269 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP11]], align 4
1270 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1271 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[TMP12]], align 4
1272 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1273 // CHECK11-NEXT:    store ptr null, ptr [[TMP13]], align 4
1274 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1275 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[TMP14]], align 4
1276 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1277 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[TMP15]], align 4
1278 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1279 // CHECK11-NEXT:    store ptr null, ptr [[TMP16]], align 4
1280 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1281 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP17]], align 4
1282 // CHECK11-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1283 // CHECK11-NEXT:    store i32 [[TMP0]], ptr [[TMP18]], align 4
1284 // CHECK11-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1285 // CHECK11-NEXT:    store ptr null, ptr [[TMP19]], align 4
1286 // CHECK11-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1287 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP20]], align 4
1288 // CHECK11-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1289 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[TMP21]], align 4
1290 // CHECK11-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
1291 // CHECK11-NEXT:    store ptr null, ptr [[TMP22]], align 4
1292 // CHECK11-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1293 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP23]], align 4
1294 // CHECK11-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1295 // CHECK11-NEXT:    store ptr [[VLA]], ptr [[TMP24]], align 4
1296 // CHECK11-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 4
1297 // CHECK11-NEXT:    store i64 [[TMP10]], ptr [[TMP25]], align 4
1298 // CHECK11-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
1299 // CHECK11-NEXT:    store ptr null, ptr [[TMP26]], align 4
1300 // CHECK11-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1301 // CHECK11-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1302 // CHECK11-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i64], ptr [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1303 // CHECK11-NEXT:    [[TMP30:%.*]] = load i32, ptr [[N]], align 4
1304 // CHECK11-NEXT:    store i32 [[TMP30]], ptr [[DOTCAPTURE_EXPR_]], align 4
1305 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32, ptr [[M]], align 4
1306 // CHECK11-NEXT:    store i32 [[TMP31]], ptr [[DOTCAPTURE_EXPR_2]], align 4
1307 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1308 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
1309 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1310 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
1311 // CHECK11-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2]], align 4
1312 // CHECK11-NEXT:    [[SUB4:%.*]] = sub nsw i32 [[TMP33]], 0
1313 // CHECK11-NEXT:    [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
1314 // CHECK11-NEXT:    [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
1315 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
1316 // CHECK11-NEXT:    [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
1317 // CHECK11-NEXT:    store i64 [[SUB7]], ptr [[DOTCAPTURE_EXPR_3]], align 8
1318 // CHECK11-NEXT:    [[TMP34:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_3]], align 8
1319 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP34]], 1
1320 // CHECK11-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1321 // CHECK11-NEXT:    store i32 3, ptr [[TMP35]], align 4
1322 // CHECK11-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1323 // CHECK11-NEXT:    store i32 5, ptr [[TMP36]], align 4
1324 // CHECK11-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1325 // CHECK11-NEXT:    store ptr [[TMP27]], ptr [[TMP37]], align 4
1326 // CHECK11-NEXT:    [[TMP38:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1327 // CHECK11-NEXT:    store ptr [[TMP28]], ptr [[TMP38]], align 4
1328 // CHECK11-NEXT:    [[TMP39:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1329 // CHECK11-NEXT:    store ptr [[TMP29]], ptr [[TMP39]], align 4
1330 // CHECK11-NEXT:    [[TMP40:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1331 // CHECK11-NEXT:    store ptr @.offload_maptypes, ptr [[TMP40]], align 4
1332 // CHECK11-NEXT:    [[TMP41:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1333 // CHECK11-NEXT:    store ptr null, ptr [[TMP41]], align 4
1334 // CHECK11-NEXT:    [[TMP42:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1335 // CHECK11-NEXT:    store ptr null, ptr [[TMP42]], align 4
1336 // CHECK11-NEXT:    [[TMP43:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1337 // CHECK11-NEXT:    store i64 [[ADD]], ptr [[TMP43]], align 8
1338 // CHECK11-NEXT:    [[TMP44:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1339 // CHECK11-NEXT:    store i64 0, ptr [[TMP44]], align 8
1340 // CHECK11-NEXT:    [[TMP45:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1341 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP45]], align 4
1342 // CHECK11-NEXT:    [[TMP46:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1343 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP46]], align 4
1344 // CHECK11-NEXT:    [[TMP47:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1345 // CHECK11-NEXT:    store i32 0, ptr [[TMP47]], align 4
1346 // 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_l83.region_id, ptr [[KERNEL_ARGS]])
1347 // CHECK11-NEXT:    [[TMP49:%.*]] = icmp ne i32 [[TMP48]], 0
1348 // CHECK11-NEXT:    br i1 [[TMP49]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1349 // CHECK11:       omp_offload.failed:
1350 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83(i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP0]], i32 [[TMP1]], ptr [[VLA]]) #[[ATTR3:[0-9]+]]
1351 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1352 // CHECK11:       omp_offload.cont:
1353 // CHECK11-NEXT:    [[TMP50:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
1354 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef [[TMP50]])
1355 // CHECK11-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
1356 // CHECK11-NEXT:    [[TMP51:%.*]] = load ptr, ptr [[SAVED_STACK]], align 4
1357 // CHECK11-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP51]])
1358 // CHECK11-NEXT:    [[TMP52:%.*]] = load i32, ptr [[RETVAL]], align 4
1359 // CHECK11-NEXT:    ret i32 [[TMP52]]
1360 //
1361 //
1362 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83
1363 // 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]+]] {
1364 // CHECK11-NEXT:  entry:
1365 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1366 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca i32, align 4
1367 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1368 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1369 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1370 // CHECK11-NEXT:    store i32 [[N]], ptr [[N_ADDR]], align 4
1371 // CHECK11-NEXT:    store i32 [[M]], ptr [[M_ADDR]], align 4
1372 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1373 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1374 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1375 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1376 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1377 // CHECK11-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1378 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined, ptr [[N_ADDR]], ptr [[M_ADDR]], i32 [[TMP0]], i32 [[TMP1]], ptr [[TMP2]])
1379 // CHECK11-NEXT:    ret void
1380 //
1381 //
1382 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined
1383 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1384 // CHECK11-NEXT:  entry:
1385 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1386 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1387 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1388 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 4
1389 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1390 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1391 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1392 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1393 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1394 // CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
1395 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1396 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1397 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
1398 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1399 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1400 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
1401 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
1402 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1403 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1404 // CHECK11-NEXT:    [[I11:%.*]] = alloca i32, align 4
1405 // CHECK11-NEXT:    [[J12:%.*]] = alloca i32, align 4
1406 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1407 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1408 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1409 // CHECK11-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 4
1410 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1411 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1412 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1413 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1414 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4
1415 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1416 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1417 // CHECK11-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1418 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
1419 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
1420 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
1421 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
1422 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1423 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
1424 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1425 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
1426 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1427 // CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
1428 // CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1429 // CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
1430 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
1431 // CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
1432 // CHECK11-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
1433 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
1434 // CHECK11-NEXT:    store i32 0, ptr [[J]], align 4
1435 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1436 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
1437 // CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
1438 // CHECK11:       land.lhs.true:
1439 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1440 // CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
1441 // CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
1442 // CHECK11:       omp.precond.then:
1443 // CHECK11-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
1444 // CHECK11-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1445 // CHECK11-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_COMB_UB]], align 8
1446 // CHECK11-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
1447 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1448 // CHECK11-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1449 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
1450 // CHECK11-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP13]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
1451 // CHECK11-NEXT:    [[TMP14:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1452 // CHECK11-NEXT:    [[TMP15:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1453 // CHECK11-NEXT:    [[CMP13:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
1454 // CHECK11-NEXT:    br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1455 // CHECK11:       cond.true:
1456 // CHECK11-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1457 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1458 // CHECK11:       cond.false:
1459 // CHECK11-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1460 // CHECK11-NEXT:    br label [[COND_END]]
1461 // CHECK11:       cond.end:
1462 // CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
1463 // CHECK11-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
1464 // CHECK11-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
1465 // CHECK11-NEXT:    store i64 [[TMP18]], ptr [[DOTOMP_IV]], align 8
1466 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1467 // CHECK11:       omp.inner.for.cond:
1468 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1469 // CHECK11-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1470 // CHECK11-NEXT:    [[CMP14:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
1471 // CHECK11-NEXT:    br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1472 // CHECK11:       omp.inner.for.body:
1473 // CHECK11-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
1474 // CHECK11-NEXT:    [[TMP22:%.*]] = trunc i64 [[TMP21]] to i32
1475 // CHECK11-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
1476 // CHECK11-NEXT:    [[TMP24:%.*]] = trunc i64 [[TMP23]] to i32
1477 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined, i32 [[TMP22]], i32 [[TMP24]], ptr [[TMP0]], ptr [[TMP1]], i32 [[TMP2]], i32 [[TMP3]], ptr [[TMP4]])
1478 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1479 // CHECK11:       omp.inner.for.inc:
1480 // CHECK11-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1481 // CHECK11-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
1482 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP25]], [[TMP26]]
1483 // CHECK11-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
1484 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1485 // CHECK11:       omp.inner.for.end:
1486 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1487 // CHECK11:       omp.loop.exit:
1488 // CHECK11-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1489 // CHECK11-NEXT:    [[TMP28:%.*]] = load i32, ptr [[TMP27]], align 4
1490 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP28]])
1491 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
1492 // CHECK11:       omp.precond.end:
1493 // CHECK11-NEXT:    ret void
1494 //
1495 //
1496 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l83.omp_outlined.omp_outlined
1497 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1498 // CHECK11-NEXT:  entry:
1499 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1500 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1501 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1502 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1503 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca ptr, align 4
1504 // CHECK11-NEXT:    [[M_ADDR:%.*]] = alloca ptr, align 4
1505 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
1506 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
1507 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1508 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1509 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1510 // CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
1511 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1512 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1513 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
1514 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1515 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1516 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
1517 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
1518 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1519 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1520 // CHECK11-NEXT:    [[I13:%.*]] = alloca i32, align 4
1521 // CHECK11-NEXT:    [[J14:%.*]] = alloca i32, align 4
1522 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1523 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1524 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1525 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1526 // CHECK11-NEXT:    store ptr [[N]], ptr [[N_ADDR]], align 4
1527 // CHECK11-NEXT:    store ptr [[M]], ptr [[M_ADDR]], align 4
1528 // CHECK11-NEXT:    store i32 [[VLA]], ptr [[VLA_ADDR]], align 4
1529 // CHECK11-NEXT:    store i32 [[VLA1]], ptr [[VLA_ADDR2]], align 4
1530 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1531 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[N_ADDR]], align 4
1532 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4
1533 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[VLA_ADDR]], align 4
1534 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[VLA_ADDR2]], align 4
1535 // CHECK11-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1536 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
1537 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTCAPTURE_EXPR_]], align 4
1538 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP1]], align 4
1539 // CHECK11-NEXT:    store i32 [[TMP6]], ptr [[DOTCAPTURE_EXPR_4]], align 4
1540 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1541 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP7]], 0
1542 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1543 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[DIV]] to i64
1544 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1545 // CHECK11-NEXT:    [[SUB6:%.*]] = sub nsw i32 [[TMP8]], 0
1546 // CHECK11-NEXT:    [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1547 // CHECK11-NEXT:    [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
1548 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
1549 // CHECK11-NEXT:    [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
1550 // CHECK11-NEXT:    store i64 [[SUB9]], ptr [[DOTCAPTURE_EXPR_5]], align 8
1551 // CHECK11-NEXT:    store i32 0, ptr [[I]], align 4
1552 // CHECK11-NEXT:    store i32 0, ptr [[J]], align 4
1553 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_]], align 4
1554 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
1555 // CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
1556 // CHECK11:       land.lhs.true:
1557 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1558 // CHECK11-NEXT:    [[CMP10:%.*]] = icmp slt i32 0, [[TMP10]]
1559 // CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
1560 // CHECK11:       omp.precond.then:
1561 // CHECK11-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
1562 // CHECK11-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1563 // CHECK11-NEXT:    store i64 [[TMP11]], ptr [[DOTOMP_UB]], align 8
1564 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1565 // CHECK11-NEXT:    [[CONV11:%.*]] = zext i32 [[TMP12]] to i64
1566 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1567 // CHECK11-NEXT:    [[CONV12:%.*]] = zext i32 [[TMP13]] to i64
1568 // CHECK11-NEXT:    store i64 [[CONV11]], ptr [[DOTOMP_LB]], align 8
1569 // CHECK11-NEXT:    store i64 [[CONV12]], ptr [[DOTOMP_UB]], align 8
1570 // CHECK11-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
1571 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1572 // CHECK11-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1573 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP14]], align 4
1574 // CHECK11-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB2:[0-9]+]], i32 [[TMP15]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
1575 // CHECK11-NEXT:    [[TMP16:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
1576 // CHECK11-NEXT:    [[TMP17:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1577 // CHECK11-NEXT:    [[CMP15:%.*]] = icmp sgt i64 [[TMP16]], [[TMP17]]
1578 // CHECK11-NEXT:    br i1 [[CMP15]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1579 // CHECK11:       cond.true:
1580 // CHECK11-NEXT:    [[TMP18:%.*]] = load i64, ptr [[DOTCAPTURE_EXPR_5]], align 8
1581 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1582 // CHECK11:       cond.false:
1583 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
1584 // CHECK11-NEXT:    br label [[COND_END]]
1585 // CHECK11:       cond.end:
1586 // CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP18]], [[COND_TRUE]] ], [ [[TMP19]], [[COND_FALSE]] ]
1587 // CHECK11-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
1588 // CHECK11-NEXT:    [[TMP20:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
1589 // CHECK11-NEXT:    store i64 [[TMP20]], ptr [[DOTOMP_IV]], align 8
1590 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1591 // CHECK11:       omp.inner.for.cond:
1592 // CHECK11-NEXT:    [[TMP21:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1593 // CHECK11-NEXT:    [[TMP22:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
1594 // CHECK11-NEXT:    [[CMP16:%.*]] = icmp sle i64 [[TMP21]], [[TMP22]]
1595 // CHECK11-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1596 // CHECK11:       omp.inner.for.body:
1597 // CHECK11-NEXT:    [[TMP23:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1598 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1599 // CHECK11-NEXT:    [[SUB17:%.*]] = sub nsw i32 [[TMP24]], 0
1600 // CHECK11-NEXT:    [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
1601 // CHECK11-NEXT:    [[MUL19:%.*]] = mul nsw i32 1, [[DIV18]]
1602 // CHECK11-NEXT:    [[CONV20:%.*]] = sext i32 [[MUL19]] to i64
1603 // CHECK11-NEXT:    [[DIV21:%.*]] = sdiv i64 [[TMP23]], [[CONV20]]
1604 // CHECK11-NEXT:    [[MUL22:%.*]] = mul nsw i64 [[DIV21]], 1
1605 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL22]]
1606 // CHECK11-NEXT:    [[CONV23:%.*]] = trunc i64 [[ADD]] to i32
1607 // CHECK11-NEXT:    store i32 [[CONV23]], ptr [[I13]], align 4
1608 // CHECK11-NEXT:    [[TMP25:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1609 // CHECK11-NEXT:    [[TMP26:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1610 // CHECK11-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1611 // CHECK11-NEXT:    [[SUB24:%.*]] = sub nsw i32 [[TMP27]], 0
1612 // CHECK11-NEXT:    [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
1613 // CHECK11-NEXT:    [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
1614 // CHECK11-NEXT:    [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
1615 // CHECK11-NEXT:    [[DIV28:%.*]] = sdiv i64 [[TMP26]], [[CONV27]]
1616 // CHECK11-NEXT:    [[TMP28:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_4]], align 4
1617 // CHECK11-NEXT:    [[SUB29:%.*]] = sub nsw i32 [[TMP28]], 0
1618 // CHECK11-NEXT:    [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
1619 // CHECK11-NEXT:    [[MUL31:%.*]] = mul nsw i32 1, [[DIV30]]
1620 // CHECK11-NEXT:    [[CONV32:%.*]] = sext i32 [[MUL31]] to i64
1621 // CHECK11-NEXT:    [[MUL33:%.*]] = mul nsw i64 [[DIV28]], [[CONV32]]
1622 // CHECK11-NEXT:    [[SUB34:%.*]] = sub nsw i64 [[TMP25]], [[MUL33]]
1623 // CHECK11-NEXT:    [[MUL35:%.*]] = mul nsw i64 [[SUB34]], 1
1624 // CHECK11-NEXT:    [[ADD36:%.*]] = add nsw i64 0, [[MUL35]]
1625 // CHECK11-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
1626 // CHECK11-NEXT:    store i32 [[CONV37]], ptr [[J14]], align 4
1627 // CHECK11-NEXT:    [[TMP29:%.*]] = load i32, ptr [[I13]], align 4
1628 // CHECK11-NEXT:    [[TMP30:%.*]] = mul nsw i32 [[TMP29]], [[TMP3]]
1629 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 [[TMP30]]
1630 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32, ptr [[J14]], align 4
1631 // CHECK11-NEXT:    [[ARRAYIDX38:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 [[TMP31]]
1632 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX38]], align 4
1633 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1634 // CHECK11:       omp.body.continue:
1635 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1636 // CHECK11:       omp.inner.for.inc:
1637 // CHECK11-NEXT:    [[TMP32:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
1638 // CHECK11-NEXT:    [[ADD39:%.*]] = add nsw i64 [[TMP32]], 1
1639 // CHECK11-NEXT:    store i64 [[ADD39]], ptr [[DOTOMP_IV]], align 8
1640 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1641 // CHECK11:       omp.inner.for.end:
1642 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1643 // CHECK11:       omp.loop.exit:
1644 // CHECK11-NEXT:    [[TMP33:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1645 // CHECK11-NEXT:    [[TMP34:%.*]] = load i32, ptr [[TMP33]], align 4
1646 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP34]])
1647 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
1648 // CHECK11:       omp.precond.end:
1649 // CHECK11-NEXT:    ret void
1650 //
1651 //
1652 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1653 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1654 // CHECK11-NEXT:  entry:
1655 // CHECK11-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
1656 // CHECK11-NEXT:    [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1657 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
1658 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
1659 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
1660 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1661 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1662 // CHECK11-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1663 // CHECK11-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
1664 // CHECK11-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1665 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP0]], align 4
1666 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1667 // CHECK11-NEXT:    store ptr [[A]], ptr [[TMP1]], align 4
1668 // CHECK11-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1669 // CHECK11-NEXT:    store ptr null, ptr [[TMP2]], align 4
1670 // CHECK11-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1671 // CHECK11-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1672 // CHECK11-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1673 // CHECK11-NEXT:    store i32 3, ptr [[TMP5]], align 4
1674 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1675 // CHECK11-NEXT:    store i32 1, ptr [[TMP6]], align 4
1676 // CHECK11-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1677 // CHECK11-NEXT:    store ptr [[TMP3]], ptr [[TMP7]], align 4
1678 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1679 // CHECK11-NEXT:    store ptr [[TMP4]], ptr [[TMP8]], align 4
1680 // CHECK11-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1681 // CHECK11-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP9]], align 4
1682 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1683 // CHECK11-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP10]], align 4
1684 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1685 // CHECK11-NEXT:    store ptr null, ptr [[TMP11]], align 4
1686 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1687 // CHECK11-NEXT:    store ptr null, ptr [[TMP12]], align 4
1688 // CHECK11-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1689 // CHECK11-NEXT:    store i64 20, ptr [[TMP13]], align 8
1690 // CHECK11-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1691 // CHECK11-NEXT:    store i64 0, ptr [[TMP14]], align 8
1692 // CHECK11-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1693 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP15]], align 4
1694 // CHECK11-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1695 // CHECK11-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP16]], align 4
1696 // CHECK11-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1697 // CHECK11-NEXT:    store i32 0, ptr [[TMP17]], align 4
1698 // 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__l69.region_id, ptr [[KERNEL_ARGS]])
1699 // CHECK11-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1700 // CHECK11-NEXT:    br i1 [[TMP19]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1701 // CHECK11:       omp_offload.failed:
1702 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69(ptr [[A]]) #[[ATTR3]]
1703 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1704 // CHECK11:       omp_offload.cont:
1705 // CHECK11-NEXT:    ret i32 0
1706 //
1707 //
1708 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69
1709 // CHECK11-SAME: (ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1710 // CHECK11-NEXT:  entry:
1711 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1712 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1713 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1714 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined, ptr [[TMP0]])
1715 // CHECK11-NEXT:    ret void
1716 //
1717 //
1718 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined
1719 // CHECK11-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1720 // CHECK11-NEXT:  entry:
1721 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1722 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1723 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1724 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1725 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1726 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1727 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1728 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1729 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1730 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1731 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1732 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1733 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1734 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1735 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1736 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1737 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1738 // CHECK11-NEXT:    store i32 19, ptr [[DOTOMP_COMB_UB]], align 4
1739 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1740 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1741 // CHECK11-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1742 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
1743 // 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)
1744 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1745 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
1746 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1747 // CHECK11:       cond.true:
1748 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1749 // CHECK11:       cond.false:
1750 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1751 // CHECK11-NEXT:    br label [[COND_END]]
1752 // CHECK11:       cond.end:
1753 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1754 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1755 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1756 // CHECK11-NEXT:    store i32 [[TMP5]], ptr [[DOTOMP_IV]], align 4
1757 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1758 // CHECK11:       omp.inner.for.cond:
1759 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1760 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1761 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1762 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1763 // CHECK11:       omp.inner.for.body:
1764 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1765 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1766 // CHECK11-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 3, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined, i32 [[TMP8]], i32 [[TMP9]], ptr [[TMP0]])
1767 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1768 // CHECK11:       omp.inner.for.inc:
1769 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1770 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1771 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1772 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1773 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1774 // CHECK11:       omp.inner.for.end:
1775 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1776 // CHECK11:       omp.loop.exit:
1777 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP2]])
1778 // CHECK11-NEXT:    ret void
1779 //
1780 //
1781 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l69.omp_outlined.omp_outlined
1782 // 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]] {
1783 // CHECK11-NEXT:  entry:
1784 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1785 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1786 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1787 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1788 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 4
1789 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1790 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1791 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1792 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1793 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1794 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1795 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1796 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
1797 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
1798 // CHECK11-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1799 // CHECK11-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1800 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1801 // CHECK11-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1802 // CHECK11-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 4
1803 // CHECK11-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 4
1804 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1805 // CHECK11-NEXT:    store i32 19, ptr [[DOTOMP_UB]], align 4
1806 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1807 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1808 // CHECK11-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
1809 // CHECK11-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
1810 // CHECK11-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1811 // CHECK11-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1812 // CHECK11-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1813 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
1814 // 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)
1815 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1816 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 19
1817 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1818 // CHECK11:       cond.true:
1819 // CHECK11-NEXT:    br label [[COND_END:%.*]]
1820 // CHECK11:       cond.false:
1821 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1822 // CHECK11-NEXT:    br label [[COND_END]]
1823 // CHECK11:       cond.end:
1824 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1825 // CHECK11-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1826 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1827 // CHECK11-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
1828 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1829 // CHECK11:       omp.inner.for.cond:
1830 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1831 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1832 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1833 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1834 // CHECK11:       omp.inner.for.body:
1835 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1836 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 2
1837 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1838 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1839 // CHECK11-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1840 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1841 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1842 // CHECK11-NEXT:    [[DIV3:%.*]] = sdiv i32 [[TMP12]], 2
1843 // CHECK11-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
1844 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL4]]
1845 // CHECK11-NEXT:    [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
1846 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
1847 // CHECK11-NEXT:    store i32 [[ADD6]], ptr [[J]], align 4
1848 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
1849 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], ptr [[TMP0]], i32 0, i32 [[TMP13]]
1850 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, ptr [[J]], align 4
1851 // CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], ptr [[ARRAYIDX]], i32 0, i32 [[TMP14]]
1852 // CHECK11-NEXT:    store i32 0, ptr [[ARRAYIDX7]], align 4
1853 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1854 // CHECK11:       omp.body.continue:
1855 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1856 // CHECK11:       omp.inner.for.inc:
1857 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1858 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
1859 // CHECK11-NEXT:    store i32 [[ADD8]], ptr [[DOTOMP_IV]], align 4
1860 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
1861 // CHECK11:       omp.inner.for.end:
1862 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1863 // CHECK11:       omp.loop.exit:
1864 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP4]])
1865 // CHECK11-NEXT:    ret void
1866 //