clang/test/OpenMP/distribute_parallel_for_reduction_task_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 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-linux -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
   3 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -emit-pch -o %t %s
   4 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
   5
   6 // RUN: %clang_cc1 -triple x86_64-unknown-linux -verify -fopenmp-simd -x c++ -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
   7 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
   8 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -debug-info-kind=limited -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
   9 // expected-no-diagnostics
  10 #ifndef HEADER
  11 #define HEADER
  12
  13 int main(int argc, char **argv) {
  14 #pragma omp target teams
  15 #pragma omp distribute parallel for reduction(task, +: argc, argv[0:10][0:argc])
  16   for (long long i = 0; i < 10; ++i) {
  17 #pragma omp task in_reduction(+: argc, argv[0:10][0:argc])
  18     ;
  19   }
  20 }
  21
  22
  23
  24 // Init firstprivate copy of argc
  25
  26 // Init firstprivate copy of argv[0:10][0:argc]
  27
  28 // Register task reduction.
  29
  30
  31
  32
  33
  34
  35
  36
  37
  38 #endif
  39 // CHECK1-LABEL: define {{[^@]+}}@main
  40 // CHECK1-SAME: (i32 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
  41 // CHECK1-NEXT:  entry:
  42 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
  43 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
  44 // CHECK1-NEXT:    [[ARGC_CASTED:%.*]] = alloca i64, align 8
  45 // CHECK1-NEXT:    store i32 [[ARGC]], ptr [[ARGC_ADDR]], align 4
  46 // CHECK1-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
  47 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
  48 // CHECK1-NEXT:    store i32 [[TMP0]], ptr [[ARGC_CASTED]], align 4
  49 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[ARGC_CASTED]], align 8
  50 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
  51 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14(i64 [[TMP1]], ptr [[TMP2]]) #[[ATTR2:[0-9]+]]
  52 // CHECK1-NEXT:    ret i32 0
  53 //
  54 //
  55 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14
  56 // CHECK1-SAME: (i64 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
  57 // CHECK1-NEXT:  entry:
  58 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i64, align 8
  59 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
  60 // CHECK1-NEXT:    [[ARGC_CASTED:%.*]] = alloca i64, align 8
  61 // CHECK1-NEXT:    store i64 [[ARGC]], ptr [[ARGC_ADDR]], align 8
  62 // CHECK1-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
  63 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, ptr [[ARGC_ADDR]], align 4
  64 // CHECK1-NEXT:    store i32 [[TMP0]], ptr [[ARGC_CASTED]], align 4
  65 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[ARGC_CASTED]], align 8
  66 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
  67 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB2:[0-9]+]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14.omp_outlined, i64 [[TMP1]], ptr [[TMP2]])
  68 // CHECK1-NEXT:    ret void
  69 //
  70 //
  71 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14.omp_outlined
  72 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR1]] {
  73 // CHECK1-NEXT:  entry:
  74 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
  75 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
  76 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i64, align 8
  77 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
  78 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
  79 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
  80 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
  81 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
  82 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
  83 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
  84 // CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
  85 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
  86 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
  87 // CHECK1-NEXT:    store i64 [[ARGC]], ptr [[ARGC_ADDR]], align 8
  88 // CHECK1-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
  89 // CHECK1-NEXT:    store i64 0, ptr [[DOTOMP_COMB_LB]], align 8
  90 // CHECK1-NEXT:    store i64 9, ptr [[DOTOMP_COMB_UB]], align 8
  91 // CHECK1-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
  92 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
  93 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
  94 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
  95 // CHECK1-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
  96 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
  97 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP2]], 9
  98 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
  99 // CHECK1:       cond.true:
 100 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 101 // CHECK1:       cond.false:
 102 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 103 // CHECK1-NEXT:    br label [[COND_END]]
 104 // CHECK1:       cond.end:
 105 // CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 106 // CHECK1-NEXT:    store i64 [[COND]], ptr [[DOTOMP_COMB_UB]], align 8
 107 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
 108 // CHECK1-NEXT:    store i64 [[TMP4]], ptr [[DOTOMP_IV]], align 8
 109 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 110 // CHECK1:       omp.inner.for.cond:
 111 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 112 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 113 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i64 [[TMP5]], [[TMP6]]
 114 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 115 // CHECK1:       omp.inner.for.body:
 116 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, ptr [[DOTOMP_COMB_LB]], align 8
 117 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, ptr [[DOTOMP_COMB_UB]], align 8
 118 // CHECK1-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
 119 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14.omp_outlined.omp_outlined, i64 [[TMP7]], i64 [[TMP8]], ptr [[ARGC_ADDR]], ptr [[TMP9]])
 120 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 121 // CHECK1:       omp.inner.for.inc:
 122 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 123 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, ptr [[DOTOMP_STRIDE]], align 8
 124 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP10]], [[TMP11]]
 125 // CHECK1-NEXT:    store i64 [[ADD]], ptr [[DOTOMP_IV]], align 8
 126 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 127 // CHECK1:       omp.inner.for.end:
 128 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 129 // CHECK1:       omp.loop.exit:
 130 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 131 // CHECK1-NEXT:    ret void
 132 //
 133 //
 134 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14.omp_outlined.omp_outlined
 135 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], ptr noundef [[ARGV:%.*]]) #[[ATTR1]] {
 136 // CHECK1-NEXT:  entry:
 137 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 138 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 139 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 140 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 141 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca ptr, align 8
 142 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca ptr, align 8
 143 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
 144 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
 145 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
 146 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
 147 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
 148 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 149 // CHECK1-NEXT:    [[ARGC1:%.*]] = alloca i32, align 4
 150 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca ptr, align 8
 151 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
 152 // CHECK1-NEXT:    [[_TMP5:%.*]] = alloca ptr, align 8
 153 // CHECK1-NEXT:    [[_TMP6:%.*]] = alloca ptr, align 8
 154 // CHECK1-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
 155 // CHECK1-NEXT:    [[DOTTASK_RED_:%.*]] = alloca ptr, align 8
 156 // CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
 157 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
 158 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x ptr], align 8
 159 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
 160 // CHECK1-NEXT:    [[_TMP28:%.*]] = alloca i8, align 1
 161 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 162 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 163 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 164 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 165 // CHECK1-NEXT:    store ptr [[ARGC]], ptr [[ARGC_ADDR]], align 8
 166 // CHECK1-NEXT:    store ptr [[ARGV]], ptr [[ARGV_ADDR]], align 8
 167 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[ARGC_ADDR]], align 8
 168 // CHECK1-NEXT:    store i64 0, ptr [[DOTOMP_LB]], align 8
 169 // CHECK1-NEXT:    store i64 9, ptr [[DOTOMP_UB]], align 8
 170 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 171 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 172 // CHECK1-NEXT:    store i64 [[TMP1]], ptr [[DOTOMP_LB]], align 8
 173 // CHECK1-NEXT:    store i64 [[TMP2]], ptr [[DOTOMP_UB]], align 8
 174 // CHECK1-NEXT:    store i64 1, ptr [[DOTOMP_STRIDE]], align 8
 175 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 176 // CHECK1-NEXT:    store i32 0, ptr [[ARGC1]], align 4
 177 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
 178 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds ptr, ptr [[TMP3]], i64 0
 179 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[ARRAYIDX]], align 8
 180 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i64 0
 181 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
 182 // CHECK1-NEXT:    [[TMP6:%.*]] = sext i32 [[TMP5]] to i64
 183 // CHECK1-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP6]]
 184 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
 185 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds ptr, ptr [[TMP7]], i64 9
 186 // CHECK1-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[ARRAYIDX3]], align 8
 187 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i64 [[LB_ADD_LEN]]
 188 // CHECK1-NEXT:    [[TMP9:%.*]] = ptrtoint ptr [[ARRAYIDX4]] to i64
 189 // CHECK1-NEXT:    [[TMP10:%.*]] = ptrtoint ptr [[ARRAYIDX2]] to i64
 190 // CHECK1-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP9]], [[TMP10]]
 191 // CHECK1-NEXT:    [[TMP12:%.*]] = sdiv exact i64 [[TMP11]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
 192 // CHECK1-NEXT:    [[TMP13:%.*]] = add nuw i64 [[TMP12]], 1
 193 // CHECK1-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
 194 // CHECK1-NEXT:    [[TMP15:%.*]] = call ptr @llvm.stacksave.p0()
 195 // CHECK1-NEXT:    store ptr [[TMP15]], ptr [[SAVED_STACK]], align 8
 196 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP13]], align 16
 197 // CHECK1-NEXT:    store i64 [[TMP13]], ptr [[__VLA_EXPR0]], align 8
 198 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[VLA]], i64 [[TMP13]]
 199 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[VLA]], [[TMP16]]
 200 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 201 // CHECK1:       omp.arrayinit.body:
 202 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 203 // CHECK1-NEXT:    store i8 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
 204 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 205 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP16]]
 206 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 207 // CHECK1:       omp.arrayinit.done:
 208 // CHECK1-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
 209 // CHECK1-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[TMP17]], align 8
 210 // CHECK1-NEXT:    [[TMP19:%.*]] = ptrtoint ptr [[TMP18]] to i64
 211 // CHECK1-NEXT:    [[TMP20:%.*]] = ptrtoint ptr [[ARRAYIDX2]] to i64
 212 // CHECK1-NEXT:    [[TMP21:%.*]] = sub i64 [[TMP19]], [[TMP20]]
 213 // CHECK1-NEXT:    [[TMP22:%.*]] = sdiv exact i64 [[TMP21]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
 214 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr i8, ptr [[VLA]], i64 [[TMP22]]
 215 // CHECK1-NEXT:    store ptr [[_TMP6]], ptr [[_TMP5]], align 8
 216 // CHECK1-NEXT:    store ptr [[TMP23]], ptr [[_TMP6]], align 8
 217 // CHECK1-NEXT:    [[DOTRD_INPUT_GEP_:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], ptr [[DOTRD_INPUT_]], i64 0, i64 0
 218 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], ptr [[DOTRD_INPUT_GEP_]], i32 0, i32 0
 219 // CHECK1-NEXT:    store ptr [[ARGC1]], ptr [[TMP24]], align 8
 220 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_]], i32 0, i32 1
 221 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[TMP25]], align 8
 222 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_]], i32 0, i32 2
 223 // CHECK1-NEXT:    store i64 4, ptr [[TMP26]], align 8
 224 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_]], i32 0, i32 3
 225 // CHECK1-NEXT:    store ptr @.red_init., ptr [[TMP27]], align 8
 226 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_]], i32 0, i32 4
 227 // CHECK1-NEXT:    store ptr null, ptr [[TMP28]], align 8
 228 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_]], i32 0, i32 5
 229 // CHECK1-NEXT:    store ptr @.red_comb., ptr [[TMP29]], align 8
 230 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_]], i32 0, i32 6
 231 // CHECK1-NEXT:    call void @llvm.memset.p0.i64(ptr align 8 [[TMP30]], i8 0, i64 4, i1 false)
 232 // CHECK1-NEXT:    [[DOTRD_INPUT_GEP_7:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], ptr [[DOTRD_INPUT_]], i64 0, i64 1
 233 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_7]], i32 0, i32 0
 234 // CHECK1-NEXT:    [[TMP32:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
 235 // CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds ptr, ptr [[TMP32]], i64 0
 236 // CHECK1-NEXT:    [[TMP33:%.*]] = load ptr, ptr [[ARRAYIDX8]], align 8
 237 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i8, ptr [[TMP33]], i64 0
 238 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, ptr [[TMP0]], align 4
 239 // CHECK1-NEXT:    [[TMP35:%.*]] = sext i32 [[TMP34]] to i64
 240 // CHECK1-NEXT:    [[LB_ADD_LEN10:%.*]] = add nsw i64 -1, [[TMP35]]
 241 // CHECK1-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[ARGV_ADDR]], align 8
 242 // CHECK1-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds ptr, ptr [[TMP36]], i64 9
 243 // CHECK1-NEXT:    [[TMP37:%.*]] = load ptr, ptr [[ARRAYIDX11]], align 8
 244 // CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds i8, ptr [[TMP37]], i64 [[LB_ADD_LEN10]]
 245 // CHECK1-NEXT:    store ptr [[VLA]], ptr [[TMP31]], align 8
 246 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
 247 // CHECK1-NEXT:    store ptr [[ARRAYIDX9]], ptr [[TMP38]], align 8
 248 // CHECK1-NEXT:    [[TMP39:%.*]] = ptrtoint ptr [[ARRAYIDX12]] to i64
 249 // CHECK1-NEXT:    [[TMP40:%.*]] = ptrtoint ptr [[ARRAYIDX9]] to i64
 250 // CHECK1-NEXT:    [[TMP41:%.*]] = sub i64 [[TMP39]], [[TMP40]]
 251 // CHECK1-NEXT:    [[TMP42:%.*]] = sdiv exact i64 [[TMP41]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
 252 // CHECK1-NEXT:    [[TMP43:%.*]] = add nuw i64 [[TMP42]], 1
 253 // CHECK1-NEXT:    [[TMP44:%.*]] = mul nuw i64 [[TMP43]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
 254 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
 255 // CHECK1-NEXT:    store i64 [[TMP44]], ptr [[TMP45]], align 8
 256 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
 257 // CHECK1-NEXT:    store ptr @.red_init..1, ptr [[TMP46]], align 8
 258 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
 259 // CHECK1-NEXT:    store ptr null, ptr [[TMP47]], align 8
 260 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
 261 // CHECK1-NEXT:    store ptr @.red_comb..2, ptr [[TMP48]], align 8
 262 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], ptr [[DOTRD_INPUT_GEP_7]], i32 0, i32 6
 263 // CHECK1-NEXT:    store i32 1, ptr [[TMP49]], align 8
 264 // CHECK1-NEXT:    [[TMP50:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 265 // CHECK1-NEXT:    [[TMP51:%.*]] = load i32, ptr [[TMP50]], align 4
 266 // CHECK1-NEXT:    [[TMP52:%.*]] = call ptr @__kmpc_taskred_modifier_init(ptr @[[GLOB2]], i32 [[TMP51]], i32 1, i32 2, ptr [[DOTRD_INPUT_]])
 267 // CHECK1-NEXT:    store ptr [[TMP52]], ptr [[DOTTASK_RED_]], align 8
 268 // CHECK1-NEXT:    [[TMP53:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 269 // CHECK1-NEXT:    [[TMP54:%.*]] = load i32, ptr [[TMP53]], align 4
 270 // CHECK1-NEXT:    call void @__kmpc_for_static_init_8(ptr @[[GLOB3:[0-9]+]], i32 [[TMP54]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i64 1, i64 1)
 271 // CHECK1-NEXT:    [[TMP55:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 272 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP55]], 9
 273 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 274 // CHECK1:       cond.true:
 275 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 276 // CHECK1:       cond.false:
 277 // CHECK1-NEXT:    [[TMP56:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 278 // CHECK1-NEXT:    br label [[COND_END]]
 279 // CHECK1:       cond.end:
 280 // CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP56]], [[COND_FALSE]] ]
 281 // CHECK1-NEXT:    store i64 [[COND]], ptr [[DOTOMP_UB]], align 8
 282 // CHECK1-NEXT:    [[TMP57:%.*]] = load i64, ptr [[DOTOMP_LB]], align 8
 283 // CHECK1-NEXT:    store i64 [[TMP57]], ptr [[DOTOMP_IV]], align 8
 284 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 285 // CHECK1:       omp.inner.for.cond:
 286 // CHECK1-NEXT:    [[TMP58:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 287 // CHECK1-NEXT:    [[TMP59:%.*]] = load i64, ptr [[DOTOMP_UB]], align 8
 288 // CHECK1-NEXT:    [[CMP13:%.*]] = icmp sle i64 [[TMP58]], [[TMP59]]
 289 // CHECK1-NEXT:    br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
 290 // CHECK1:       omp.inner.for.cond.cleanup:
 291 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
 292 // CHECK1:       omp.inner.for.body:
 293 // CHECK1-NEXT:    [[TMP60:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 294 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP60]], 1
 295 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i64 0, [[MUL]]
 296 // CHECK1-NEXT:    store i64 [[ADD]], ptr [[I]], align 8
 297 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 0
 298 // CHECK1-NEXT:    store ptr [[DOTTASK_RED_]], ptr [[TMP61]], align 8
 299 // CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 1
 300 // CHECK1-NEXT:    store ptr [[ARGC1]], ptr [[TMP62]], align 8
 301 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 2
 302 // CHECK1-NEXT:    [[TMP64:%.*]] = load ptr, ptr [[_TMP5]], align 8
 303 // CHECK1-NEXT:    store ptr [[TMP64]], ptr [[TMP63]], align 8
 304 // CHECK1-NEXT:    [[TMP65:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 305 // CHECK1-NEXT:    [[TMP66:%.*]] = load i32, ptr [[TMP65]], align 4
 306 // CHECK1-NEXT:    [[TMP67:%.*]] = call ptr @__kmpc_omp_task_alloc(ptr @[[GLOB2]], i32 [[TMP66]], i32 1, i64 48, i64 24, ptr @.omp_task_entry.)
 307 // CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP67]], i32 0, i32 0
 308 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP68]], i32 0, i32 0
 309 // CHECK1-NEXT:    [[TMP70:%.*]] = load ptr, ptr [[TMP69]], align 8
 310 // CHECK1-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 8 [[TMP70]], ptr align 8 [[AGG_CAPTURED]], i64 24, i1 false)
 311 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP67]], i32 0, i32 1
 312 // CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP71]], i32 0, i32 0
 313 // CHECK1-NEXT:    [[TMP73:%.*]] = load ptr, ptr [[DOTTASK_RED_]], align 8
 314 // CHECK1-NEXT:    store ptr [[TMP73]], ptr [[TMP72]], align 8
 315 // CHECK1-NEXT:    [[TMP74:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 316 // CHECK1-NEXT:    [[TMP75:%.*]] = load i32, ptr [[TMP74]], align 4
 317 // CHECK1-NEXT:    [[TMP76:%.*]] = call i32 @__kmpc_omp_task(ptr @[[GLOB2]], i32 [[TMP75]], ptr [[TMP67]])
 318 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 319 // CHECK1:       omp.body.continue:
 320 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 321 // CHECK1:       omp.inner.for.inc:
 322 // CHECK1-NEXT:    [[TMP77:%.*]] = load i64, ptr [[DOTOMP_IV]], align 8
 323 // CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i64 [[TMP77]], 1
 324 // CHECK1-NEXT:    store i64 [[ADD14]], ptr [[DOTOMP_IV]], align 8
 325 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 326 // CHECK1:       omp.inner.for.end:
 327 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 328 // CHECK1:       omp.loop.exit:
 329 // CHECK1-NEXT:    [[TMP78:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 330 // CHECK1-NEXT:    [[TMP79:%.*]] = load i32, ptr [[TMP78]], align 4
 331 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP79]])
 332 // CHECK1-NEXT:    [[TMP80:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 333 // CHECK1-NEXT:    [[TMP81:%.*]] = load i32, ptr [[TMP80]], align 4
 334 // CHECK1-NEXT:    call void @__kmpc_task_reduction_modifier_fini(ptr @[[GLOB2]], i32 [[TMP81]], i32 1)
 335 // CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 336 // CHECK1-NEXT:    store ptr [[ARGC1]], ptr [[TMP82]], align 8
 337 // CHECK1-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
 338 // CHECK1-NEXT:    store ptr [[VLA]], ptr [[TMP83]], align 8
 339 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [3 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
 340 // CHECK1-NEXT:    [[TMP85:%.*]] = inttoptr i64 [[TMP13]] to ptr
 341 // CHECK1-NEXT:    store ptr [[TMP85]], ptr [[TMP84]], align 8
 342 // CHECK1-NEXT:    [[TMP86:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 343 // CHECK1-NEXT:    [[TMP87:%.*]] = load i32, ptr [[TMP86]], align 4
 344 // CHECK1-NEXT:    [[TMP88:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB4:[0-9]+]], i32 [[TMP87]], i32 2, i64 24, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 345 // CHECK1-NEXT:    switch i32 [[TMP88]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 346 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 347 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 348 // CHECK1-NEXT:    ]
 349 // CHECK1:       .omp.reduction.case1:
 350 // CHECK1-NEXT:    [[TMP89:%.*]] = load i32, ptr [[TMP0]], align 4
 351 // CHECK1-NEXT:    [[TMP90:%.*]] = load i32, ptr [[ARGC1]], align 4
 352 // CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP89]], [[TMP90]]
 353 // CHECK1-NEXT:    store i32 [[ADD15]], ptr [[TMP0]], align 4
 354 // CHECK1-NEXT:    [[TMP91:%.*]] = getelementptr i8, ptr [[ARRAYIDX2]], i64 [[TMP13]]
 355 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAYIDX2]], [[TMP91]]
 356 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE22:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 357 // CHECK1:       omp.arraycpy.body:
 358 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 359 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi ptr [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 360 // CHECK1-NEXT:    [[TMP92:%.*]] = load i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
 361 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP92]] to i32
 362 // CHECK1-NEXT:    [[TMP93:%.*]] = load i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
 363 // CHECK1-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP93]] to i32
 364 // CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV]], [[CONV17]]
 365 // CHECK1-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
 366 // CHECK1-NEXT:    store i8 [[CONV19]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
 367 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
 368 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 369 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE21:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP91]]
 370 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_BODY]]
 371 // CHECK1:       omp.arraycpy.done22:
 372 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB4]], i32 [[TMP87]], ptr @.gomp_critical_user_.reduction.var)
 373 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 374 // CHECK1:       .omp.reduction.case2:
 375 // CHECK1-NEXT:    [[TMP94:%.*]] = load i32, ptr [[ARGC1]], align 4
 376 // CHECK1-NEXT:    [[TMP95:%.*]] = atomicrmw add ptr [[TMP0]], i32 [[TMP94]] monotonic, align 4
 377 // CHECK1-NEXT:    [[TMP96:%.*]] = getelementptr i8, ptr [[ARRAYIDX2]], i64 [[TMP13]]
 378 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY23:%.*]] = icmp eq ptr [[ARRAYIDX2]], [[TMP96]]
 379 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY23]], label [[OMP_ARRAYCPY_DONE36:%.*]], label [[OMP_ARRAYCPY_BODY24:%.*]]
 380 // CHECK1:       omp.arraycpy.body24:
 381 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST25:%.*]] = phi ptr [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT34:%.*]], [[ATOMIC_EXIT:%.*]] ]
 382 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi ptr [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT33:%.*]], [[ATOMIC_EXIT]] ]
 383 // CHECK1-NEXT:    [[TMP97:%.*]] = load i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
 384 // CHECK1-NEXT:    [[CONV27:%.*]] = sext i8 [[TMP97]] to i32
 385 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST26]] monotonic, align 1
 386 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
 387 // CHECK1:       atomic_cont:
 388 // CHECK1-NEXT:    [[TMP98:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY24]] ], [ [[TMP103:%.*]], [[ATOMIC_CONT]] ]
 389 // CHECK1-NEXT:    store i8 [[TMP98]], ptr [[_TMP28]], align 1
 390 // CHECK1-NEXT:    [[TMP99:%.*]] = load i8, ptr [[_TMP28]], align 1
 391 // CHECK1-NEXT:    [[CONV29:%.*]] = sext i8 [[TMP99]] to i32
 392 // CHECK1-NEXT:    [[TMP100:%.*]] = load i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
 393 // CHECK1-NEXT:    [[CONV30:%.*]] = sext i8 [[TMP100]] to i32
 394 // CHECK1-NEXT:    [[ADD31:%.*]] = add nsw i32 [[CONV29]], [[CONV30]]
 395 // CHECK1-NEXT:    [[CONV32:%.*]] = trunc i32 [[ADD31]] to i8
 396 // CHECK1-NEXT:    store i8 [[CONV32]], ptr [[ATOMIC_TEMP]], align 1
 397 // CHECK1-NEXT:    [[TMP101:%.*]] = load i8, ptr [[ATOMIC_TEMP]], align 1
 398 // CHECK1-NEXT:    [[TMP102:%.*]] = cmpxchg ptr [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i8 [[TMP98]], i8 [[TMP101]] monotonic monotonic, align 1
 399 // CHECK1-NEXT:    [[TMP103]] = extractvalue { i8, i1 } [[TMP102]], 0
 400 // CHECK1-NEXT:    [[TMP104:%.*]] = extractvalue { i8, i1 } [[TMP102]], 1
 401 // CHECK1-NEXT:    br i1 [[TMP104]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
 402 // CHECK1:       atomic_exit:
 403 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT33]] = getelementptr i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
 404 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT34]] = getelementptr i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST25]], i32 1
 405 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE35:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT33]], [[TMP96]]
 406 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_DONE36]], label [[OMP_ARRAYCPY_BODY24]]
 407 // CHECK1:       omp.arraycpy.done36:
 408 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 409 // CHECK1:       .omp.reduction.default:
 410 // CHECK1-NEXT:    [[TMP105:%.*]] = load ptr, ptr [[SAVED_STACK]], align 8
 411 // CHECK1-NEXT:    call void @llvm.stackrestore.p0(ptr [[TMP105]])
 412 // CHECK1-NEXT:    ret void
 413 //
 414 //
 415 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
 416 // CHECK1-SAME: (ptr noalias noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] {
 417 // CHECK1-NEXT:  entry:
 418 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 419 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 420 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 421 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 422 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 423 // CHECK1-NEXT:    store i32 0, ptr [[TMP2]], align 4
 424 // CHECK1-NEXT:    ret void
 425 //
 426 //
 427 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
 428 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
 429 // CHECK1-NEXT:  entry:
 430 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 431 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 432 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 433 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 434 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 435 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 436 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP2]], align 4
 437 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP3]], align 4
 438 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP4]], [[TMP5]]
 439 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP2]], align 4
 440 // CHECK1-NEXT:    ret void
 441 //
 442 //
 443 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
 444 // CHECK1-SAME: (ptr noalias noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR4]] {
 445 // CHECK1-NEXT:  entry:
 446 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 447 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 448 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 449 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 450 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 451 // CHECK1-NEXT:    [[TMP3:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 452 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP3]], align 8
 453 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP4]]
 454 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[TMP2]], [[TMP5]]
 455 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 456 // CHECK1:       omp.arrayinit.body:
 457 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 458 // CHECK1-NEXT:    store i8 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
 459 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 460 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
 461 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 462 // CHECK1:       omp.arrayinit.done:
 463 // CHECK1-NEXT:    ret void
 464 //
 465 //
 466 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
 467 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
 468 // CHECK1-NEXT:  entry:
 469 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 470 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 471 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 472 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 473 // CHECK1-NEXT:    [[TMP2:%.*]] = call align 8 ptr @llvm.threadlocal.address.p0(ptr align 8 @{{reduction_size[.].+[.]}})
 474 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, ptr [[TMP2]], align 8
 475 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 476 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 477 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]]
 478 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP4]], [[TMP6]]
 479 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 480 // CHECK1:       omp.arraycpy.body:
 481 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 482 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP4]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 483 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
 484 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP7]] to i32
 485 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
 486 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP8]] to i32
 487 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
 488 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
 489 // CHECK1-NEXT:    store i8 [[CONV3]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
 490 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 491 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 492 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
 493 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
 494 // CHECK1:       omp.arraycpy.done4:
 495 // CHECK1-NEXT:    ret void
 496 //
 497 //
 498 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
 499 // CHECK1-SAME: (ptr noalias noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR7:[0-9]+]] {
 500 // CHECK1-NEXT:  entry:
 501 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 502 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 503 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 504 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 505 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 506 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], ptr [[TMP2]], i32 0, i32 0
 507 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 508 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[TMP4]], align 8
 509 // CHECK1-NEXT:    ret void
 510 //
 511 //
 512 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
 513 // CHECK1-SAME: (i32 noundef [[TMP0:%.*]], ptr noalias noundef [[TMP1:%.*]]) #[[ATTR4]] {
 514 // CHECK1-NEXT:  entry:
 515 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
 516 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca ptr, align 8
 517 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca ptr, align 8
 518 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca ptr, align 8
 519 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca ptr, align 8
 520 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca ptr, align 8
 521 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca ptr, align 8
 522 // CHECK1-NEXT:    [[TMP_I:%.*]] = alloca ptr, align 8
 523 // CHECK1-NEXT:    [[TMP4_I:%.*]] = alloca ptr, align 8
 524 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
 525 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 526 // CHECK1-NEXT:    store i32 [[TMP0]], ptr [[DOTADDR]], align 4
 527 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 528 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTADDR]], align 4
 529 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 530 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], ptr [[TMP3]], i32 0, i32 0
 531 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], ptr [[TMP4]], i32 0, i32 2
 532 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], ptr [[TMP4]], i32 0, i32 0
 533 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 534 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], ptr [[TMP3]], i32 0, i32 1
 535 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
 536 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
 537 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
 538 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
 539 // CHECK1-NEXT:    store i32 [[TMP2]], ptr [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
 540 // CHECK1-NEXT:    store ptr [[TMP5]], ptr [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
 541 // CHECK1-NEXT:    store ptr [[TMP8]], ptr [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
 542 // CHECK1-NEXT:    store ptr @.omp_task_privates_map., ptr [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
 543 // CHECK1-NEXT:    store ptr [[TMP3]], ptr [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
 544 // CHECK1-NEXT:    store ptr [[TMP7]], ptr [[__CONTEXT_ADDR_I]], align 8, !noalias !12
 545 // CHECK1-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[__CONTEXT_ADDR_I]], align 8, !noalias !12
 546 // CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
 547 // CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
 548 // CHECK1-NEXT:    call void [[TMP10]](ptr [[TMP11]], ptr [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR2]]
 549 // CHECK1-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
 550 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], ptr [[TMP9]], i32 0, i32 1
 551 // CHECK1-NEXT:    [[TMP14:%.*]] = load ptr, ptr [[TMP13]], align 8
 552 // CHECK1-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[TMP12]], align 8
 553 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
 554 // CHECK1-NEXT:    [[TMP17:%.*]] = call ptr @__kmpc_task_reduction_get_th_data(i32 [[TMP16]], ptr [[TMP15]], ptr [[TMP14]])
 555 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[TMP9]], i32 0, i32 2
 556 // CHECK1-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[TMP18]], align 8
 557 // CHECK1-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[TMP19]], align 8
 558 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[TMP9]], i32 0, i32 1
 559 // CHECK1-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[TMP21]], align 8
 560 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
 561 // CHECK1-NEXT:    [[TMP24:%.*]] = sext i32 [[TMP23]] to i64
 562 // CHECK1-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP24]]
 563 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[TMP9]], i32 0, i32 2
 564 // CHECK1-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[TMP25]], align 8
 565 // CHECK1-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds ptr, ptr [[TMP26]], i64 9
 566 // CHECK1-NEXT:    [[TMP27:%.*]] = load ptr, ptr [[ARRAYIDX2_I]], align 8
 567 // CHECK1-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, ptr [[TMP27]], i64 [[LB_ADD_LEN_I]]
 568 // CHECK1-NEXT:    [[TMP28:%.*]] = ptrtoint ptr [[ARRAYIDX3_I]] to i64
 569 // CHECK1-NEXT:    [[TMP29:%.*]] = ptrtoint ptr [[TMP20]] to i64
 570 // CHECK1-NEXT:    [[TMP30:%.*]] = sub i64 [[TMP28]], [[TMP29]]
 571 // CHECK1-NEXT:    [[TMP31:%.*]] = add nuw i64 [[TMP30]], 1
 572 // CHECK1-NEXT:    [[TMP32:%.*]] = mul nuw i64 [[TMP31]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
 573 // CHECK1-NEXT:    store i64 [[TMP31]], ptr @{{reduction_size[.].+[.]}}, align 8, !noalias !12
 574 // CHECK1-NEXT:    [[TMP33:%.*]] = load ptr, ptr [[TMP12]], align 8
 575 // CHECK1-NEXT:    [[TMP34:%.*]] = call ptr @__kmpc_task_reduction_get_th_data(i32 [[TMP16]], ptr [[TMP33]], ptr [[TMP20]])
 576 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_ANON]], ptr [[TMP9]], i32 0, i32 2
 577 // CHECK1-NEXT:    [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
 578 // CHECK1-NEXT:    [[TMP37:%.*]] = load ptr, ptr [[TMP36]], align 8
 579 // CHECK1-NEXT:    [[TMP38:%.*]] = ptrtoint ptr [[TMP37]] to i64
 580 // CHECK1-NEXT:    [[TMP39:%.*]] = ptrtoint ptr [[TMP20]] to i64
 581 // CHECK1-NEXT:    [[TMP40:%.*]] = sub i64 [[TMP38]], [[TMP39]]
 582 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr i8, ptr [[TMP34]], i64 [[TMP40]]
 583 // CHECK1-NEXT:    store ptr [[TMP4_I]], ptr [[TMP_I]], align 8, !noalias !12
 584 // CHECK1-NEXT:    store ptr [[TMP41]], ptr [[TMP4_I]], align 8, !noalias !12
 585 // CHECK1-NEXT:    ret i32 0
 586 //
 587 //
 588 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14.omp_outlined.omp_outlined.omp.reduction.reduction_func
 589 // CHECK1-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR4]] {
 590 // CHECK1-NEXT:  entry:
 591 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 592 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 593 // CHECK1-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 594 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 595 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 596 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 597 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [3 x ptr], ptr [[TMP3]], i64 0, i64 0
 598 // CHECK1-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 599 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x ptr], ptr [[TMP2]], i64 0, i64 0
 600 // CHECK1-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 601 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [3 x ptr], ptr [[TMP3]], i64 0, i64 1
 602 // CHECK1-NEXT:    [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
 603 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x ptr], ptr [[TMP2]], i64 0, i64 1
 604 // CHECK1-NEXT:    [[TMP11:%.*]] = load ptr, ptr [[TMP10]], align 8
 605 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x ptr], ptr [[TMP2]], i64 0, i64 2
 606 // CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP12]], align 8
 607 // CHECK1-NEXT:    [[TMP14:%.*]] = ptrtoint ptr [[TMP13]] to i64
 608 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[TMP7]], align 4
 609 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[TMP5]], align 4
 610 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
 611 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[TMP7]], align 4
 612 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP11]], i64 [[TMP14]]
 613 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP11]], [[TMP17]]
 614 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 615 // CHECK1:       omp.arraycpy.body:
 616 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP9]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 617 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP11]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 618 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
 619 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP18]] to i32
 620 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
 621 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP19]] to i32
 622 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
 623 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
 624 // CHECK1-NEXT:    store i8 [[CONV4]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
 625 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 626 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 627 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP17]]
 628 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
 629 // CHECK1:       omp.arraycpy.done5:
 630 // CHECK1-NEXT:    ret void
 631 //