clang/test/OpenMP/teams_generic_loop_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 -triple x86_64-pc-linux-gnu -fopenmp -emit-llvm %s -o - | FileCheck %s --check-prefix=IR
   3
   4 // Check same results after serialization round-trip
   5 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -emit-pch -o %t %s
   6 // RUN: %clang_cc1 -verify -triple x86_64-pc-linux-gnu -fopenmp -include-pch %t -emit-llvm %s -o - | FileCheck %s --check-prefix=IR-PCH
   7
   8 // expected-no-diagnostics
   9
  10 #ifndef HEADER
  11 #define HEADER
  12 int foo() {
  13   int i;
  14   int j;
  15   int sum[10][10];
  16
  17   #pragma omp teams loop reduction(+:sum) collapse(2) bind(parallel) \
  18       order(concurrent) lastprivate(j)
  19   for(i=0; i<10; i++)
  20     for(j=0; j<10; j++)
  21       sum[i][j] += i;
  22
  23   return 0;
  24 }
  25 #endif
  26 // IR-LABEL: define {{[^@]+}}@_Z3foov
  27 // IR-SAME: () #[[ATTR0:[0-9]+]] {
  28 // IR-NEXT:  entry:
  29 // IR-NEXT:    [[I:%.*]] = alloca i32, align 4
  30 // IR-NEXT:    [[J:%.*]] = alloca i32, align 4
  31 // IR-NEXT:    [[SUM:%.*]] = alloca [10 x [10 x i32]], align 16
  32 // IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4:[0-9]+]], i32 2, ptr @_Z3foov.omp_outlined, ptr [[J]], ptr [[SUM]])
  33 // IR-NEXT:    ret i32 0
  34 //
  35 //
  36 // IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined
  37 // IR-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1:[0-9]+]] {
  38 // IR-NEXT:  entry:
  39 // IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
  40 // IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
  41 // IR-NEXT:    [[J_ADDR:%.*]] = alloca ptr, align 8
  42 // IR-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
  43 // IR-NEXT:    [[SUM1:%.*]] = alloca [10 x [10 x i32]], align 16
  44 // IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
  45 // IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
  46 // IR-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
  47 // IR-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
  48 // IR-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
  49 // IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
  50 // IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
  51 // IR-NEXT:    [[J3:%.*]] = alloca i32, align 4
  52 // IR-NEXT:    [[I:%.*]] = alloca i32, align 4
  53 // IR-NEXT:    [[J4:%.*]] = alloca i32, align 4
  54 // IR-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
  55 // IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
  56 // IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
  57 // IR-NEXT:    store ptr [[J]], ptr [[J_ADDR]], align 8
  58 // IR-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
  59 // IR-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[J_ADDR]], align 8
  60 // IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
  61 // IR-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i32 0, i32 0, i32 0
  62 // IR-NEXT:    [[TMP2:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
  63 // IR-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP2]]
  64 // IR-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
  65 // IR:       omp.arrayinit.body:
  66 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
  67 // IR-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
  68 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
  69 // IR-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP2]]
  70 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
  71 // IR:       omp.arrayinit.done:
  72 // IR-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
  73 // IR-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
  74 // IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
  75 // IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
  76 // IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
  77 // IR-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
  78 // IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
  79 // IR-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
  80 // IR-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
  81 // IR-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
  82 // IR:       cond.true:
  83 // IR-NEXT:    br label [[COND_END:%.*]]
  84 // IR:       cond.false:
  85 // IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
  86 // IR-NEXT:    br label [[COND_END]]
  87 // IR:       cond.end:
  88 // IR-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
  89 // IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
  90 // IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
  91 // IR-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
  92 // IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
  93 // IR:       omp.inner.for.cond:
  94 // IR-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
  95 // IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
  96 // IR-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
  97 // IR-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
  98 // IR:       omp.inner.for.body:
  99 // IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 100 // IR-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
 101 // IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 102 // IR-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
 103 // IR-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 4, ptr @_Z3foov.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]], ptr [[J3]], ptr [[SUM1]])
 104 // IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 105 // IR:       omp.inner.for.inc:
 106 // IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 107 // IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 108 // IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
 109 // IR-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 110 // IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
 111 // IR:       omp.inner.for.end:
 112 // IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 113 // IR:       omp.loop.exit:
 114 // IR-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 115 // IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
 116 // IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP17]])
 117 // IR-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 118 // IR-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 119 // IR-NEXT:    br i1 [[TMP19]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 120 // IR:       .omp.lastprivate.then:
 121 // IR-NEXT:    store i32 10, ptr [[J3]], align 4
 122 // IR-NEXT:    [[TMP20:%.*]] = load i32, ptr [[J3]], align 4
 123 // IR-NEXT:    store i32 [[TMP20]], ptr [[TMP0]], align 4
 124 // IR-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 125 // IR:       .omp.lastprivate.done:
 126 // IR-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 127 // IR-NEXT:    store ptr [[SUM1]], ptr [[TMP21]], align 8
 128 // IR-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 129 // IR-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
 130 // IR-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 131 // IR-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 132 // IR-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 133 // IR-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 134 // IR-NEXT:    ]
 135 // IR:       .omp.reduction.case1:
 136 // IR-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 137 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP25]]
 138 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 139 // IR:       omp.arraycpy.body:
 140 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 141 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 142 // IR-NEXT:    [[TMP26:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
 143 // IR-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 144 // IR-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
 145 // IR-NEXT:    store i32 [[ADD7]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
 146 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
 147 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 148 // IR-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP25]]
 149 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
 150 // IR:       omp.arraycpy.done10:
 151 // IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
 152 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 153 // IR:       .omp.reduction.case2:
 154 // IR-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 155 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq ptr [[TMP1]], [[TMP28]]
 156 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
 157 // IR:       omp.arraycpy.body12:
 158 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
 159 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
 160 // IR-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], align 4
 161 // IR-NEXT:    [[TMP30:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 [[TMP29]] monotonic, align 4
 162 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
 163 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
 164 // IR-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP28]]
 165 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY12]]
 166 // IR:       omp.arraycpy.done18:
 167 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 168 // IR:       .omp.reduction.default:
 169 // IR-NEXT:    ret void
 170 //
 171 //
 172 // IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined
 173 // IR-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) [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1]] {
 174 // IR-NEXT:  entry:
 175 // IR-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 176 // IR-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 177 // IR-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 178 // IR-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 179 // IR-NEXT:    [[J_ADDR:%.*]] = alloca ptr, align 8
 180 // IR-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
 181 // IR-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 182 // IR-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 183 // IR-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 184 // IR-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 185 // IR-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 186 // IR-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 187 // IR-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 188 // IR-NEXT:    [[J3:%.*]] = alloca i32, align 4
 189 // IR-NEXT:    [[SUM4:%.*]] = alloca [10 x [10 x i32]], align 16
 190 // IR-NEXT:    [[I:%.*]] = alloca i32, align 4
 191 // IR-NEXT:    [[J5:%.*]] = alloca i32, align 4
 192 // IR-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 193 // IR-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 194 // IR-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 195 // IR-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 196 // IR-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 197 // IR-NEXT:    store ptr [[J]], ptr [[J_ADDR]], align 8
 198 // IR-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
 199 // IR-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[J_ADDR]], align 8
 200 // IR-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
 201 // IR-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 202 // IR-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
 203 // IR-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 204 // IR-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP2]] to i32
 205 // IR-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 206 // IR-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP3]] to i32
 207 // IR-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 208 // IR-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
 209 // IR-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 210 // IR-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 211 // IR-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i32 0, i32 0, i32 0
 212 // IR-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
 213 // IR-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP4]]
 214 // IR-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 215 // IR:       omp.arrayinit.body:
 216 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 217 // IR-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 218 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 219 // IR-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
 220 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 221 // IR:       omp.arrayinit.done:
 222 // IR-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 223 // IR-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 224 // IR-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP6]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 225 // IR-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 226 // IR-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 99
 227 // IR-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 228 // IR:       cond.true:
 229 // IR-NEXT:    br label [[COND_END:%.*]]
 230 // IR:       cond.false:
 231 // IR-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 232 // IR-NEXT:    br label [[COND_END]]
 233 // IR:       cond.end:
 234 // IR-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
 235 // IR-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 236 // IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 237 // IR-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
 238 // IR-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 239 // IR:       omp.inner.for.cond:
 240 // IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3:![0-9]+]]
 241 // IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP3]]
 242 // IR-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 243 // IR-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 244 // IR:       omp.inner.for.body:
 245 // IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 246 // IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP12]], 10
 247 // IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 248 // IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 249 // IR-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 250 // IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 251 // IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 252 // IR-NEXT:    [[DIV7:%.*]] = sdiv i32 [[TMP14]], 10
 253 // IR-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 10
 254 // IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP13]], [[MUL8]]
 255 // IR-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[SUB]], 1
 256 // IR-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
 257 // IR-NEXT:    store i32 [[ADD10]], ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
 258 // IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 259 // IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 260 // IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 261 // IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i64 0, i64 [[IDXPROM]]
 262 // IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
 263 // IR-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP17]] to i64
 264 // IR-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM11]]
 265 // IR-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
 266 // IR-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP18]], [[TMP15]]
 267 // IR-NEXT:    store i32 [[ADD13]], ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
 268 // IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 269 // IR:       omp.body.continue:
 270 // IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 271 // IR:       omp.inner.for.inc:
 272 // IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 273 // IR-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP19]], 1
 274 // IR-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 275 // IR-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
 276 // IR:       omp.inner.for.end:
 277 // IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 278 // IR:       omp.loop.exit:
 279 // IR-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 280 // IR-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 281 // IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 282 // IR-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 283 // IR-NEXT:    store ptr [[SUM4]], ptr [[TMP22]], align 8
 284 // IR-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 285 // IR-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 286 // IR-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 287 // IR-NEXT:    switch i32 [[TMP25]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 288 // IR-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 289 // IR-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 290 // IR-NEXT:    ]
 291 // IR:       .omp.reduction.case1:
 292 // IR-NEXT:    [[TMP26:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 293 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP26]]
 294 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 295 // IR:       omp.arraycpy.body:
 296 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 297 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 298 // IR-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
 299 // IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 300 // IR-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
 301 // IR-NEXT:    store i32 [[ADD16]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
 302 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
 303 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 304 // IR-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP26]]
 305 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
 306 // IR:       omp.arraycpy.done19:
 307 // IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
 308 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 309 // IR:       .omp.reduction.case2:
 310 // IR-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 311 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY20:%.*]] = icmp eq ptr [[TMP1]], [[TMP29]]
 312 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY20]], label [[OMP_ARRAYCPY_DONE27:%.*]], label [[OMP_ARRAYCPY_BODY21:%.*]]
 313 // IR:       omp.arraycpy.body21:
 314 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST22:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT25:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
 315 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST23:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
 316 // IR-NEXT:    [[TMP30:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], align 4
 317 // IR-NEXT:    [[TMP31:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 [[TMP30]] monotonic, align 4
 318 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 1
 319 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT25]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], i32 1
 320 // IR-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP29]]
 321 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY21]]
 322 // IR:       omp.arraycpy.done27:
 323 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 324 // IR:       .omp.reduction.default:
 325 // IR-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 326 // IR-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
 327 // IR-NEXT:    br i1 [[TMP33]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 328 // IR:       .omp.lastprivate.then:
 329 // IR-NEXT:    store i32 10, ptr [[J3]], align 4
 330 // IR-NEXT:    [[TMP34:%.*]] = load i32, ptr [[J3]], align 4
 331 // IR-NEXT:    store i32 [[TMP34]], ptr [[TMP0]], align 4
 332 // IR-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 333 // IR:       .omp.lastprivate.done:
 334 // IR-NEXT:    ret void
 335 //
 336 //
 337 // IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func
 338 // IR-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 339 // IR-NEXT:  entry:
 340 // IR-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 341 // IR-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 342 // IR-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 343 // IR-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 344 // IR-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 345 // IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 346 // IR-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 347 // IR-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 348 // IR-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 349 // IR-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 350 // IR-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
 351 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
 352 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 353 // IR:       omp.arraycpy.body:
 354 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 355 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 356 // IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 357 // IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 358 // IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 359 // IR-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 360 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 361 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 362 // IR-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
 363 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
 364 // IR:       omp.arraycpy.done2:
 365 // IR-NEXT:    ret void
 366 //
 367 //
 368 // IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp.reduction.reduction_func
 369 // IR-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
 370 // IR-NEXT:  entry:
 371 // IR-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 372 // IR-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 373 // IR-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 374 // IR-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 375 // IR-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 376 // IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 377 // IR-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 378 // IR-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 379 // IR-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 380 // IR-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 381 // IR-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
 382 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
 383 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 384 // IR:       omp.arraycpy.body:
 385 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 386 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 387 // IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 388 // IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 389 // IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 390 // IR-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 391 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 392 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 393 // IR-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
 394 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
 395 // IR:       omp.arraycpy.done2:
 396 // IR-NEXT:    ret void
 397 //
 398 //
 399 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov
 400 // IR-PCH-SAME: () #[[ATTR0:[0-9]+]] {
 401 // IR-PCH-NEXT:  entry:
 402 // IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
 403 // IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
 404 // IR-PCH-NEXT:    [[SUM:%.*]] = alloca [10 x [10 x i32]], align 16
 405 // IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB4:[0-9]+]], i32 2, ptr @_Z3foov.omp_outlined, ptr [[J]], ptr [[SUM]])
 406 // IR-PCH-NEXT:    ret i32 0
 407 //
 408 //
 409 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined
 410 // IR-PCH-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1:[0-9]+]] {
 411 // IR-PCH-NEXT:  entry:
 412 // IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 413 // IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 414 // IR-PCH-NEXT:    [[J_ADDR:%.*]] = alloca ptr, align 8
 415 // IR-PCH-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
 416 // IR-PCH-NEXT:    [[SUM1:%.*]] = alloca [10 x [10 x i32]], align 16
 417 // IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 418 // IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 419 // IR-PCH-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 420 // IR-PCH-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 421 // IR-PCH-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 422 // IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 423 // IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 424 // IR-PCH-NEXT:    [[J3:%.*]] = alloca i32, align 4
 425 // IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
 426 // IR-PCH-NEXT:    [[J4:%.*]] = alloca i32, align 4
 427 // IR-PCH-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 428 // IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 429 // IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 430 // IR-PCH-NEXT:    store ptr [[J]], ptr [[J_ADDR]], align 8
 431 // IR-PCH-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
 432 // IR-PCH-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[J_ADDR]], align 8
 433 // IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
 434 // IR-PCH-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i32 0, i32 0, i32 0
 435 // IR-PCH-NEXT:    [[TMP2:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
 436 // IR-PCH-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP2]]
 437 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 438 // IR-PCH:       omp.arrayinit.body:
 439 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 440 // IR-PCH-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 441 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 442 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP2]]
 443 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 444 // IR-PCH:       omp.arrayinit.done:
 445 // IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 446 // IR-PCH-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
 447 // IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 448 // IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 449 // IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 450 // IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 451 // IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 452 // IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 453 // IR-PCH-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
 454 // IR-PCH-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 455 // IR-PCH:       cond.true:
 456 // IR-PCH-NEXT:    br label [[COND_END:%.*]]
 457 // IR-PCH:       cond.false:
 458 // IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 459 // IR-PCH-NEXT:    br label [[COND_END]]
 460 // IR-PCH:       cond.end:
 461 // IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 462 // IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 463 // IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 464 // IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 465 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 466 // IR-PCH:       omp.inner.for.cond:
 467 // IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 468 // IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 469 // IR-PCH-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 470 // IR-PCH-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 471 // IR-PCH:       omp.inner.for.body:
 472 // IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 473 // IR-PCH-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
 474 // IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 475 // IR-PCH-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
 476 // IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB4]], i32 4, ptr @_Z3foov.omp_outlined.omp_outlined, i64 [[TMP11]], i64 [[TMP13]], ptr [[J3]], ptr [[SUM1]])
 477 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 478 // IR-PCH:       omp.inner.for.inc:
 479 // IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 480 // IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 481 // IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
 482 // IR-PCH-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 483 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
 484 // IR-PCH:       omp.inner.for.end:
 485 // IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 486 // IR-PCH:       omp.loop.exit:
 487 // IR-PCH-NEXT:    [[TMP16:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 488 // IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[TMP16]], align 4
 489 // IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2:[0-9]+]], i32 [[TMP17]])
 490 // IR-PCH-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 491 // IR-PCH-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
 492 // IR-PCH-NEXT:    br i1 [[TMP19]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 493 // IR-PCH:       .omp.lastprivate.then:
 494 // IR-PCH-NEXT:    store i32 10, ptr [[J3]], align 4
 495 // IR-PCH-NEXT:    [[TMP20:%.*]] = load i32, ptr [[J3]], align 4
 496 // IR-PCH-NEXT:    store i32 [[TMP20]], ptr [[TMP0]], align 4
 497 // IR-PCH-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 498 // IR-PCH:       .omp.lastprivate.done:
 499 // IR-PCH-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 500 // IR-PCH-NEXT:    store ptr [[SUM1]], ptr [[TMP21]], align 8
 501 // IR-PCH-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 502 // IR-PCH-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
 503 // IR-PCH-NEXT:    [[TMP24:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3:[0-9]+]], i32 [[TMP23]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 504 // IR-PCH-NEXT:    switch i32 [[TMP24]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 505 // IR-PCH-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 506 // IR-PCH-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 507 // IR-PCH-NEXT:    ]
 508 // IR-PCH:       .omp.reduction.case1:
 509 // IR-PCH-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 510 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP25]]
 511 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE10:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 512 // IR-PCH:       omp.arraycpy.body:
 513 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 514 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST6:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT8:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 515 // IR-PCH-NEXT:    [[TMP26:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
 516 // IR-PCH-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 517 // IR-PCH-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
 518 // IR-PCH-NEXT:    store i32 [[ADD7]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], align 4
 519 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT8]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST6]], i32 1
 520 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 521 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE9:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT8]], [[TMP25]]
 522 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE9]], label [[OMP_ARRAYCPY_DONE10]], label [[OMP_ARRAYCPY_BODY]]
 523 // IR-PCH:       omp.arraycpy.done10:
 524 // IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP23]], ptr @.gomp_critical_user_.reduction.var)
 525 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 526 // IR-PCH:       .omp.reduction.case2:
 527 // IR-PCH-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 528 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY11:%.*]] = icmp eq ptr [[TMP1]], [[TMP28]]
 529 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY11]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY12:%.*]]
 530 // IR-PCH:       omp.arraycpy.body12:
 531 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST13:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
 532 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY12]] ]
 533 // IR-PCH-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], align 4
 534 // IR-PCH-NEXT:    [[TMP30:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 [[TMP29]] monotonic, align 4
 535 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
 536 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST13]], i32 1
 537 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP28]]
 538 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY12]]
 539 // IR-PCH:       omp.arraycpy.done18:
 540 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 541 // IR-PCH:       .omp.reduction.default:
 542 // IR-PCH-NEXT:    ret void
 543 //
 544 //
 545 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined
 546 // IR-PCH-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) [[J:%.*]], ptr noundef nonnull align 4 dereferenceable(400) [[SUM:%.*]]) #[[ATTR1]] {
 547 // IR-PCH-NEXT:  entry:
 548 // IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 549 // IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 550 // IR-PCH-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 551 // IR-PCH-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 552 // IR-PCH-NEXT:    [[J_ADDR:%.*]] = alloca ptr, align 8
 553 // IR-PCH-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
 554 // IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 555 // IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 556 // IR-PCH-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
 557 // IR-PCH-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 558 // IR-PCH-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 559 // IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 560 // IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 561 // IR-PCH-NEXT:    [[J3:%.*]] = alloca i32, align 4
 562 // IR-PCH-NEXT:    [[SUM4:%.*]] = alloca [10 x [10 x i32]], align 16
 563 // IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
 564 // IR-PCH-NEXT:    [[J5:%.*]] = alloca i32, align 4
 565 // IR-PCH-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 566 // IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 567 // IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 568 // IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 569 // IR-PCH-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 570 // IR-PCH-NEXT:    store ptr [[J]], ptr [[J_ADDR]], align 8
 571 // IR-PCH-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
 572 // IR-PCH-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[J_ADDR]], align 8
 573 // IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
 574 // IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 575 // IR-PCH-NEXT:    store i32 99, ptr [[DOTOMP_UB]], align 4
 576 // IR-PCH-NEXT:    [[TMP2:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 577 // IR-PCH-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP2]] to i32
 578 // IR-PCH-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 579 // IR-PCH-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP3]] to i32
 580 // IR-PCH-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 581 // IR-PCH-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
 582 // IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 583 // IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 584 // IR-PCH-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i32 0, i32 0, i32 0
 585 // IR-PCH-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
 586 // IR-PCH-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP4]]
 587 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 588 // IR-PCH:       omp.arrayinit.body:
 589 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 590 // IR-PCH-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 591 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 592 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
 593 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 594 // IR-PCH:       omp.arrayinit.done:
 595 // IR-PCH-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 596 // IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
 597 // IR-PCH-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP6]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 598 // IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 599 // IR-PCH-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 99
 600 // IR-PCH-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 601 // IR-PCH:       cond.true:
 602 // IR-PCH-NEXT:    br label [[COND_END:%.*]]
 603 // IR-PCH:       cond.false:
 604 // IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 605 // IR-PCH-NEXT:    br label [[COND_END]]
 606 // IR-PCH:       cond.end:
 607 // IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
 608 // IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 609 // IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 610 // IR-PCH-NEXT:    store i32 [[TMP9]], ptr [[DOTOMP_IV]], align 4
 611 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 612 // IR-PCH:       omp.inner.for.cond:
 613 // IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3:![0-9]+]]
 614 // IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4, !llvm.access.group [[ACC_GRP3]]
 615 // IR-PCH-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
 616 // IR-PCH-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 617 // IR-PCH:       omp.inner.for.body:
 618 // IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 619 // IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP12]], 10
 620 // IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 621 // IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 622 // IR-PCH-NEXT:    store i32 [[ADD]], ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 623 // IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 624 // IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 625 // IR-PCH-NEXT:    [[DIV7:%.*]] = sdiv i32 [[TMP14]], 10
 626 // IR-PCH-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 10
 627 // IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP13]], [[MUL8]]
 628 // IR-PCH-NEXT:    [[MUL9:%.*]] = mul nsw i32 [[SUB]], 1
 629 // IR-PCH-NEXT:    [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
 630 // IR-PCH-NEXT:    store i32 [[ADD10]], ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
 631 // IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 632 // IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4, !llvm.access.group [[ACC_GRP3]]
 633 // IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
 634 // IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM4]], i64 0, i64 [[IDXPROM]]
 635 // IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[J3]], align 4, !llvm.access.group [[ACC_GRP3]]
 636 // IR-PCH-NEXT:    [[IDXPROM11:%.*]] = sext i32 [[TMP17]] to i64
 637 // IR-PCH-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM11]]
 638 // IR-PCH-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
 639 // IR-PCH-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP18]], [[TMP15]]
 640 // IR-PCH-NEXT:    store i32 [[ADD13]], ptr [[ARRAYIDX12]], align 4, !llvm.access.group [[ACC_GRP3]]
 641 // IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 642 // IR-PCH:       omp.body.continue:
 643 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 644 // IR-PCH:       omp.inner.for.inc:
 645 // IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 646 // IR-PCH-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP19]], 1
 647 // IR-PCH-NEXT:    store i32 [[ADD14]], ptr [[DOTOMP_IV]], align 4, !llvm.access.group [[ACC_GRP3]]
 648 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
 649 // IR-PCH:       omp.inner.for.end:
 650 // IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 651 // IR-PCH:       omp.loop.exit:
 652 // IR-PCH-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 653 // IR-PCH-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
 654 // IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
 655 // IR-PCH-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 656 // IR-PCH-NEXT:    store ptr [[SUM4]], ptr [[TMP22]], align 8
 657 // IR-PCH-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 658 // IR-PCH-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
 659 // IR-PCH-NEXT:    [[TMP25:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 660 // IR-PCH-NEXT:    switch i32 [[TMP25]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 661 // IR-PCH-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 662 // IR-PCH-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 663 // IR-PCH-NEXT:    ]
 664 // IR-PCH:       .omp.reduction.case1:
 665 // IR-PCH-NEXT:    [[TMP26:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 666 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP26]]
 667 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE19:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 668 // IR-PCH:       omp.arraycpy.body:
 669 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 670 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT17:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 671 // IR-PCH-NEXT:    [[TMP27:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
 672 // IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 673 // IR-PCH-NEXT:    [[ADD16:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
 674 // IR-PCH-NEXT:    store i32 [[ADD16]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 4
 675 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT17]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
 676 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 677 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE18:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT17]], [[TMP26]]
 678 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_DONE19]], label [[OMP_ARRAYCPY_BODY]]
 679 // IR-PCH:       omp.arraycpy.done19:
 680 // IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB3]], i32 [[TMP24]], ptr @.gomp_critical_user_.reduction.var)
 681 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 682 // IR-PCH:       .omp.reduction.case2:
 683 // IR-PCH-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 684 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY20:%.*]] = icmp eq ptr [[TMP1]], [[TMP29]]
 685 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY20]], label [[OMP_ARRAYCPY_DONE27:%.*]], label [[OMP_ARRAYCPY_BODY21:%.*]]
 686 // IR-PCH:       omp.arraycpy.body21:
 687 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST22:%.*]] = phi ptr [ [[SUM4]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT25:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
 688 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST23:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY21]] ]
 689 // IR-PCH-NEXT:    [[TMP30:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], align 4
 690 // IR-PCH-NEXT:    [[TMP31:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 [[TMP30]] monotonic, align 4
 691 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST23]], i32 1
 692 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT25]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST22]], i32 1
 693 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE26:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT24]], [[TMP29]]
 694 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_DONE27]], label [[OMP_ARRAYCPY_BODY21]]
 695 // IR-PCH:       omp.arraycpy.done27:
 696 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 697 // IR-PCH:       .omp.reduction.default:
 698 // IR-PCH-NEXT:    [[TMP32:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 699 // IR-PCH-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
 700 // IR-PCH-NEXT:    br i1 [[TMP33]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 701 // IR-PCH:       .omp.lastprivate.then:
 702 // IR-PCH-NEXT:    store i32 10, ptr [[J3]], align 4
 703 // IR-PCH-NEXT:    [[TMP34:%.*]] = load i32, ptr [[J3]], align 4
 704 // IR-PCH-NEXT:    store i32 [[TMP34]], ptr [[TMP0]], align 4
 705 // IR-PCH-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 706 // IR-PCH:       .omp.lastprivate.done:
 707 // IR-PCH-NEXT:    ret void
 708 //
 709 //
 710 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp_outlined.omp.reduction.reduction_func
 711 // IR-PCH-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 712 // IR-PCH-NEXT:  entry:
 713 // IR-PCH-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 714 // IR-PCH-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 715 // IR-PCH-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 716 // IR-PCH-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 717 // IR-PCH-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 718 // IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 719 // IR-PCH-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 720 // IR-PCH-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 721 // IR-PCH-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 722 // IR-PCH-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 723 // IR-PCH-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
 724 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
 725 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 726 // IR-PCH:       omp.arraycpy.body:
 727 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 728 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 729 // IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 730 // IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 731 // IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 732 // IR-PCH-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 733 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 734 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 735 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
 736 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
 737 // IR-PCH:       omp.arraycpy.done2:
 738 // IR-PCH-NEXT:    ret void
 739 //
 740 //
 741 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp.reduction.reduction_func
 742 // IR-PCH-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3]] {
 743 // IR-PCH-NEXT:  entry:
 744 // IR-PCH-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 745 // IR-PCH-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 746 // IR-PCH-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 747 // IR-PCH-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 748 // IR-PCH-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 749 // IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 750 // IR-PCH-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 751 // IR-PCH-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 752 // IR-PCH-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 753 // IR-PCH-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 754 // IR-PCH-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
 755 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
 756 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 757 // IR-PCH:       omp.arraycpy.body:
 758 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 759 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 760 // IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 761 // IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 762 // IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 763 // IR-PCH-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 764 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 765 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 766 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
 767 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
 768 // IR-PCH:       omp.arraycpy.done2:
 769 // IR-PCH-NEXT:    ret void
 770 //