clang/test/OpenMP/parallel_for_lastprivate_conditional.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 // RUN: %clang_cc1 -verify -fopenmp -DOMP5 -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
   3 // RUN: %clang_cc1 -fopenmp -DOMP5 -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
   4 // RUN: %clang_cc1 -fopenmp -DOMP5 -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
   5
   6 // RUN: %clang_cc1 -verify -fopenmp-simd -DOMP5 -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
   7 // RUN: %clang_cc1 -fopenmp-simd -DOMP5 -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
   8 // RUN: %clang_cc1 -fopenmp-simd -DOMP5 -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
   9 // expected-no-diagnostics
  10
  11 #ifndef HEADER
  12 #define HEADER
  13
  14 int main() {
  15   int a = 0;
  16 #pragma omp parallel for lastprivate(conditional: a)
  17   for (int i = 0; i < 10; ++i) {
  18     if (i < 5) {
  19       a = 0;
  20 #pragma omp parallel reduction(+:a) num_threads(10)
  21       a += i;
  22 #pragma omp atomic
  23       a += i;
  24 #pragma omp parallel num_threads(10)
  25 #pragma omp atomic
  26       a += i;
  27     }
  28   }
  29   return 0;
  30 }
  31
  32
  33
  34
  35 #endif // HEADER
  36 // CHECK1-LABEL: define {{[^@]+}}@main
  37 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
  38 // CHECK1-NEXT:  entry:
  39 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
  40 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
  41 // CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
  42 // CHECK1-NEXT:    store i32 0, ptr [[A]], align 4
  43 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2:[0-9]+]], i32 1, ptr @main.omp_outlined, ptr [[A]])
  44 // CHECK1-NEXT:    ret i32 0
  45 //
  46 //
  47 // CHECK1-LABEL: define {{[^@]+}}@main.omp_outlined
  48 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
  49 // CHECK1-NEXT:  entry:
  50 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
  51 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
  52 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
  53 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
  54 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
  55 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
  56 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
  57 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
  58 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
  59 // CHECK1-NEXT:    [[A1:%.*]] = alloca [[STRUCT_LASPRIVATE_CONDITIONAL:%.*]], align 4
  60 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
  61 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
  62 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
  63 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
  64 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
  65 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
  66 // CHECK1-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
  67 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
  68 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
  69 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], ptr [[A1]], i32 0, i32 1
  70 // CHECK1-NEXT:    store i8 0, ptr [[TMP1]], align 4
  71 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], ptr [[A1]], i32 0, i32 0
  72 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
  73 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
  74 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
  75 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
  76 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
  77 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
  78 // CHECK1:       cond.true:
  79 // CHECK1-NEXT:    br label [[COND_END:%.*]]
  80 // CHECK1:       cond.false:
  81 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
  82 // CHECK1-NEXT:    br label [[COND_END]]
  83 // CHECK1:       cond.end:
  84 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
  85 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
  86 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
  87 // CHECK1-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
  88 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
  89 // CHECK1:       omp.inner.for.cond:
  90 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
  91 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
  92 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
  93 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
  94 // CHECK1:       omp.inner.for.body:
  95 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
  96 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
  97 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
  98 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
  99 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[I]], align 4
 100 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp slt i32 [[TMP11]], 5
 101 // CHECK1-NEXT:    br i1 [[CMP3]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
 102 // CHECK1:       if.then:
 103 // CHECK1-NEXT:    store i32 0, ptr [[TMP2]], align 4
 104 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 105 // CHECK1-NEXT:    call void @__kmpc_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 106 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr @.{{pl_cond[.].+[.|,]}} align 4
 107 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp sle i32 [[TMP13]], [[TMP12]]
 108 // CHECK1-NEXT:    br i1 [[TMP14]], label [[LP_COND_THEN:%.*]], label [[LP_COND_EXIT:%.*]]
 109 // CHECK1:       lp_cond_then:
 110 // CHECK1-NEXT:    store i32 [[TMP12]], ptr @.{{pl_cond[.].+[.|,]}} align 4
 111 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP2]], align 4
 112 // CHECK1-NEXT:    store i32 [[TMP15]], ptr @{{pl_cond[.].+[.|,]}} align 4
 113 // CHECK1-NEXT:    br label [[LP_COND_EXIT]]
 114 // CHECK1:       lp_cond_exit:
 115 // CHECK1-NEXT:    call void @__kmpc_end_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 116 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB2]], i32 [[TMP4]], i32 10)
 117 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 2, ptr @main.omp_outlined.omp_outlined, ptr [[TMP2]], ptr [[I]])
 118 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 119 // CHECK1-NEXT:    call void @__kmpc_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 120 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr @.{{pl_cond[.].+[.|,]}} align 4
 121 // CHECK1-NEXT:    [[TMP18:%.*]] = icmp sle i32 [[TMP17]], [[TMP16]]
 122 // CHECK1-NEXT:    br i1 [[TMP18]], label [[LP_COND_THEN4:%.*]], label [[LP_COND_EXIT5:%.*]]
 123 // CHECK1:       lp_cond_then4:
 124 // CHECK1-NEXT:    store i32 [[TMP16]], ptr @.{{pl_cond[.].+[.|,]}} align 4
 125 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP2]], align 4
 126 // CHECK1-NEXT:    store i32 [[TMP19]], ptr @{{pl_cond[.].+[.|,]}} align 4
 127 // CHECK1-NEXT:    br label [[LP_COND_EXIT5]]
 128 // CHECK1:       lp_cond_exit5:
 129 // CHECK1-NEXT:    call void @__kmpc_end_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 130 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, ptr [[I]], align 4
 131 // CHECK1-NEXT:    [[TMP21:%.*]] = atomicrmw add ptr [[TMP2]], i32 [[TMP20]] monotonic, align 4
 132 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 133 // CHECK1-NEXT:    call void @__kmpc_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 134 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr @.{{pl_cond[.].+[.|,]}} align 4
 135 // CHECK1-NEXT:    [[TMP24:%.*]] = icmp sle i32 [[TMP23]], [[TMP22]]
 136 // CHECK1-NEXT:    br i1 [[TMP24]], label [[LP_COND_THEN6:%.*]], label [[LP_COND_EXIT7:%.*]]
 137 // CHECK1:       lp_cond_then6:
 138 // CHECK1-NEXT:    store i32 [[TMP22]], ptr @.{{pl_cond[.].+[.|,]}} align 4
 139 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP2]], align 4
 140 // CHECK1-NEXT:    store i32 [[TMP25]], ptr @{{pl_cond[.].+[.|,]}} align 4
 141 // CHECK1-NEXT:    br label [[LP_COND_EXIT7]]
 142 // CHECK1:       lp_cond_exit7:
 143 // CHECK1-NEXT:    call void @__kmpc_end_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 144 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(ptr @[[GLOB2]], i32 [[TMP4]], i32 10)
 145 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 2, ptr @main.omp_outlined.omp_outlined.1, ptr [[TMP2]], ptr [[I]])
 146 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], ptr [[A1]], i32 0, i32 1
 147 // CHECK1-NEXT:    [[TMP27:%.*]] = load i8, ptr [[TMP26]], align 4
 148 // CHECK1-NEXT:    [[TMP28:%.*]] = icmp ne i8 [[TMP27]], 0
 149 // CHECK1-NEXT:    br i1 [[TMP28]], label [[LPC_THEN:%.*]], label [[LPC_DONE:%.*]]
 150 // CHECK1:       lpc.then:
 151 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 152 // CHECK1-NEXT:    call void @__kmpc_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 153 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, ptr @.{{pl_cond[.].+[.|,]}} align 4
 154 // CHECK1-NEXT:    [[TMP31:%.*]] = icmp sle i32 [[TMP30]], [[TMP29]]
 155 // CHECK1-NEXT:    br i1 [[TMP31]], label [[LP_COND_THEN8:%.*]], label [[LP_COND_EXIT9:%.*]]
 156 // CHECK1:       lp_cond_then8:
 157 // CHECK1-NEXT:    store i32 [[TMP29]], ptr @.{{pl_cond[.].+[.|,]}} align 4
 158 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, ptr [[TMP2]], align 4
 159 // CHECK1-NEXT:    store i32 [[TMP32]], ptr @{{pl_cond[.].+[.|,]}} align 4
 160 // CHECK1-NEXT:    br label [[LP_COND_EXIT9]]
 161 // CHECK1:       lp_cond_exit9:
 162 // CHECK1-NEXT:    call void @__kmpc_end_critical(ptr @[[GLOB2]], i32 [[TMP4]], ptr @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
 163 // CHECK1-NEXT:    br label [[LPC_DONE]]
 164 // CHECK1:       lpc.done:
 165 // CHECK1-NEXT:    br label [[IF_END]]
 166 // CHECK1:       if.end:
 167 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 168 // CHECK1:       omp.body.continue:
 169 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 170 // CHECK1:       omp.inner.for.inc:
 171 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 172 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP33]], 1
 173 // CHECK1-NEXT:    store i32 [[ADD10]], ptr [[DOTOMP_IV]], align 4
 174 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 175 // CHECK1:       omp.inner.for.end:
 176 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 177 // CHECK1:       omp.loop.exit:
 178 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP4]])
 179 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 180 // CHECK1-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
 181 // CHECK1-NEXT:    call void @__kmpc_barrier(ptr @[[GLOB4:[0-9]+]], i32 [[TMP4]])
 182 // CHECK1-NEXT:    br i1 [[TMP35]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 183 // CHECK1:       .omp.lastprivate.then:
 184 // CHECK1-NEXT:    [[TMP36:%.*]] = load i32, ptr @{{pl_cond[.].+[.|,]}} align 4
 185 // CHECK1-NEXT:    store i32 [[TMP36]], ptr [[TMP2]], align 4
 186 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, ptr [[TMP2]], align 4
 187 // CHECK1-NEXT:    store i32 [[TMP37]], ptr [[TMP0]], align 4
 188 // CHECK1-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 189 // CHECK1:       .omp.lastprivate.done:
 190 // CHECK1-NEXT:    ret void
 191 //
 192 //
 193 // CHECK1-LABEL: define {{[^@]+}}@main.omp_outlined.omp_outlined
 194 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR1]] {
 195 // CHECK1-NEXT:  entry:
 196 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 197 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 198 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 199 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca ptr, align 8
 200 // CHECK1-NEXT:    [[A1:%.*]] = alloca i32, align 4
 201 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 202 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 203 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 204 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 205 // CHECK1-NEXT:    store ptr [[I]], ptr [[I_ADDR]], align 8
 206 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 207 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[I_ADDR]], align 8
 208 // CHECK1-NEXT:    store i32 0, ptr [[A1]], align 4
 209 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 210 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[A1]], align 4
 211 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], [[TMP2]]
 212 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[A1]], align 4
 213 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 214 // CHECK1-NEXT:    store ptr [[A1]], ptr [[TMP4]], align 8
 215 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 216 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 217 // CHECK1-NEXT:    [[TMP7:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP6]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @main.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 218 // CHECK1-NEXT:    switch i32 [[TMP7]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 219 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 220 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 221 // CHECK1-NEXT:    ]
 222 // CHECK1:       .omp.reduction.case1:
 223 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP0]], align 4
 224 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[A1]], align 4
 225 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
 226 // CHECK1-NEXT:    store i32 [[ADD2]], ptr [[TMP0]], align 4
 227 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP6]], ptr @.gomp_critical_user_.reduction.var)
 228 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 229 // CHECK1:       .omp.reduction.case2:
 230 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[A1]], align 4
 231 // CHECK1-NEXT:    [[TMP11:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP10]] monotonic, align 4
 232 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 233 // CHECK1:       .omp.reduction.default:
 234 // CHECK1-NEXT:    ret void
 235 //
 236 //
 237 // CHECK1-LABEL: define {{[^@]+}}@main.omp_outlined.omp_outlined.omp.reduction.reduction_func
 238 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
 239 // CHECK1-NEXT:  entry:
 240 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 241 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 242 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 243 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 244 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 245 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 246 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 247 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 248 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 249 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 250 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
 251 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 4
 252 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], [[TMP9]]
 253 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
 254 // CHECK1-NEXT:    ret void
 255 //
 256 //
 257 // CHECK1-LABEL: define {{[^@]+}}@main.omp_outlined.omp_outlined.1
 258 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[A:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR1]] {
 259 // CHECK1-NEXT:  entry:
 260 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 261 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 262 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca ptr, align 8
 263 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca ptr, align 8
 264 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 265 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 266 // CHECK1-NEXT:    store ptr [[A]], ptr [[A_ADDR]], align 8
 267 // CHECK1-NEXT:    store ptr [[I]], ptr [[I_ADDR]], align 8
 268 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[A_ADDR]], align 8
 269 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[I_ADDR]], align 8
 270 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
 271 // CHECK1-NEXT:    [[TMP3:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP2]] monotonic, align 4
 272 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL:%.*]], ptr [[TMP0]], i32 0, i32 1
 273 // CHECK1-NEXT:    store atomic volatile i8 1, ptr [[TMP4]] unordered, align 1
 274 // CHECK1-NEXT:    ret void
 275 //