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 @[[GLOB3:[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_IV]], align 4
 100 // IR-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 10
 101 // IR-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 102 // IR-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 103 // IR-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 104 // IR-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 105 // IR-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 106 // IR-NEXT:    [[DIV6:%.*]] = sdiv i32 [[TMP12]], 10
 107 // IR-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[DIV6]], 10
 108 // IR-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL7]]
 109 // IR-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[SUB]], 1
 110 // IR-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
 111 // IR-NEXT:    store i32 [[ADD9]], ptr [[J3]], align 4
 112 // IR-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
 113 // IR-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
 114 // IR-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP14]] to i64
 115 // IR-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i64 0, i64 [[IDXPROM]]
 116 // IR-NEXT:    [[TMP15:%.*]] = load i32, ptr [[J3]], align 4
 117 // IR-NEXT:    [[IDXPROM10:%.*]] = sext i32 [[TMP15]] to i64
 118 // IR-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM10]]
 119 // IR-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX11]], align 4
 120 // IR-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP16]], [[TMP13]]
 121 // IR-NEXT:    store i32 [[ADD12]], ptr [[ARRAYIDX11]], align 4
 122 // IR-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 123 // IR:       omp.body.continue:
 124 // IR-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 125 // IR:       omp.inner.for.inc:
 126 // IR-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 127 // IR-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP17]], 1
 128 // IR-NEXT:    store i32 [[ADD13]], ptr [[DOTOMP_IV]], align 4
 129 // IR-NEXT:    br label [[OMP_INNER_FOR_COND]]
 130 // IR:       omp.inner.for.end:
 131 // IR-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 132 // IR:       omp.loop.exit:
 133 // IR-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 134 // IR-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
 135 // IR-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
 136 // IR-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 137 // IR-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 138 // IR-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 139 // IR:       .omp.lastprivate.then:
 140 // IR-NEXT:    store i32 10, ptr [[J3]], align 4
 141 // IR-NEXT:    [[TMP22:%.*]] = load i32, ptr [[J3]], align 4
 142 // IR-NEXT:    store i32 [[TMP22]], ptr [[TMP0]], align 4
 143 // IR-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 144 // IR:       .omp.lastprivate.done:
 145 // IR-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 146 // IR-NEXT:    store ptr [[SUM1]], ptr [[TMP23]], align 8
 147 // IR-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 148 // IR-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
 149 // IR-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2:[0-9]+]], i32 [[TMP25]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 150 // IR-NEXT:    switch i32 [[TMP26]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 151 // IR-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 152 // IR-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 153 // IR-NEXT:    ]
 154 // IR:       .omp.reduction.case1:
 155 // IR-NEXT:    [[TMP27:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 156 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP27]]
 157 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 158 // IR:       omp.arraycpy.body:
 159 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 160 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 161 // IR-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
 162 // IR-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 163 // IR-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
 164 // IR-NEXT:    store i32 [[ADD15]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
 165 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
 166 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 167 // IR-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP27]]
 168 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
 169 // IR:       omp.arraycpy.done18:
 170 // IR-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP25]], ptr @.gomp_critical_user_.reduction.var)
 171 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 172 // IR:       .omp.reduction.case2:
 173 // IR-NEXT:    [[TMP30:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 174 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq ptr [[TMP1]], [[TMP30]]
 175 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE26:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
 176 // IR:       omp.arraycpy.body20:
 177 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
 178 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
 179 // IR-NEXT:    [[TMP31:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 4
 180 // IR-NEXT:    [[TMP32:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 [[TMP31]] monotonic, align 4
 181 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT23]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
 182 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
 183 // IR-NEXT:    [[OMP_ARRAYCPY_DONE25:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT23]], [[TMP30]]
 184 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_BODY20]]
 185 // IR:       omp.arraycpy.done26:
 186 // IR-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 187 // IR:       .omp.reduction.default:
 188 // IR-NEXT:    ret void
 189 //
 190 //
 191 // IR-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp.reduction.reduction_func
 192 // IR-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 193 // IR-NEXT:  entry:
 194 // IR-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 195 // IR-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 196 // IR-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 197 // IR-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 198 // IR-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 199 // IR-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 200 // IR-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 201 // IR-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 202 // IR-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 203 // IR-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 204 // IR-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
 205 // IR-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
 206 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 207 // IR:       omp.arraycpy.body:
 208 // IR-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 209 // IR-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 210 // IR-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 211 // IR-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 212 // IR-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 213 // IR-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 214 // IR-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 215 // IR-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 216 // IR-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
 217 // IR-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
 218 // IR:       omp.arraycpy.done2:
 219 // IR-NEXT:    ret void
 220 //
 221 //
 222 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov
 223 // IR-PCH-SAME: () #[[ATTR0:[0-9]+]] {
 224 // IR-PCH-NEXT:  entry:
 225 // IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
 226 // IR-PCH-NEXT:    [[J:%.*]] = alloca i32, align 4
 227 // IR-PCH-NEXT:    [[SUM:%.*]] = alloca [10 x [10 x i32]], align 16
 228 // IR-PCH-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 2, ptr @_Z3foov.omp_outlined, ptr [[J]], ptr [[SUM]])
 229 // IR-PCH-NEXT:    ret i32 0
 230 //
 231 //
 232 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined
 233 // 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]+]] {
 234 // IR-PCH-NEXT:  entry:
 235 // IR-PCH-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 236 // IR-PCH-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 237 // IR-PCH-NEXT:    [[J_ADDR:%.*]] = alloca ptr, align 8
 238 // IR-PCH-NEXT:    [[SUM_ADDR:%.*]] = alloca ptr, align 8
 239 // IR-PCH-NEXT:    [[SUM1:%.*]] = alloca [10 x [10 x i32]], align 16
 240 // IR-PCH-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 241 // IR-PCH-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 242 // IR-PCH-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
 243 // IR-PCH-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 244 // IR-PCH-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 245 // IR-PCH-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 246 // IR-PCH-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 247 // IR-PCH-NEXT:    [[J3:%.*]] = alloca i32, align 4
 248 // IR-PCH-NEXT:    [[I:%.*]] = alloca i32, align 4
 249 // IR-PCH-NEXT:    [[J4:%.*]] = alloca i32, align 4
 250 // IR-PCH-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8
 251 // IR-PCH-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 252 // IR-PCH-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 253 // IR-PCH-NEXT:    store ptr [[J]], ptr [[J_ADDR]], align 8
 254 // IR-PCH-NEXT:    store ptr [[SUM]], ptr [[SUM_ADDR]], align 8
 255 // IR-PCH-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[J_ADDR]], align 8
 256 // IR-PCH-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[SUM_ADDR]], align 8
 257 // IR-PCH-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i32 0, i32 0, i32 0
 258 // IR-PCH-NEXT:    [[TMP2:%.*]] = getelementptr i32, ptr [[ARRAY_BEGIN]], i64 100
 259 // IR-PCH-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP2]]
 260 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
 261 // IR-PCH:       omp.arrayinit.body:
 262 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
 263 // IR-PCH-NEXT:    store i32 0, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 264 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 265 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP2]]
 266 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
 267 // IR-PCH:       omp.arrayinit.done:
 268 // IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 269 // IR-PCH-NEXT:    store i32 99, ptr [[DOTOMP_COMB_UB]], align 4
 270 // IR-PCH-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 271 // IR-PCH-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 272 // IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 273 // IR-PCH-NEXT:    [[TMP4:%.*]] = load i32, ptr [[TMP3]], align 4
 274 // 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)
 275 // IR-PCH-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 276 // IR-PCH-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
 277 // IR-PCH-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 278 // IR-PCH:       cond.true:
 279 // IR-PCH-NEXT:    br label [[COND_END:%.*]]
 280 // IR-PCH:       cond.false:
 281 // IR-PCH-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 282 // IR-PCH-NEXT:    br label [[COND_END]]
 283 // IR-PCH:       cond.end:
 284 // IR-PCH-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
 285 // IR-PCH-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 286 // IR-PCH-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 287 // IR-PCH-NEXT:    store i32 [[TMP7]], ptr [[DOTOMP_IV]], align 4
 288 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 289 // IR-PCH:       omp.inner.for.cond:
 290 // IR-PCH-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 291 // IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 292 // IR-PCH-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
 293 // IR-PCH-NEXT:    br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 294 // IR-PCH:       omp.inner.for.body:
 295 // IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 296 // IR-PCH-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP10]], 10
 297 // IR-PCH-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
 298 // IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 299 // IR-PCH-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 300 // IR-PCH-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 301 // IR-PCH-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 302 // IR-PCH-NEXT:    [[DIV6:%.*]] = sdiv i32 [[TMP12]], 10
 303 // IR-PCH-NEXT:    [[MUL7:%.*]] = mul nsw i32 [[DIV6]], 10
 304 // IR-PCH-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP11]], [[MUL7]]
 305 // IR-PCH-NEXT:    [[MUL8:%.*]] = mul nsw i32 [[SUB]], 1
 306 // IR-PCH-NEXT:    [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
 307 // IR-PCH-NEXT:    store i32 [[ADD9]], ptr [[J3]], align 4
 308 // IR-PCH-NEXT:    [[TMP13:%.*]] = load i32, ptr [[I]], align 4
 309 // IR-PCH-NEXT:    [[TMP14:%.*]] = load i32, ptr [[I]], align 4
 310 // IR-PCH-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP14]] to i64
 311 // IR-PCH-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [10 x i32]], ptr [[SUM1]], i64 0, i64 [[IDXPROM]]
 312 // IR-PCH-NEXT:    [[TMP15:%.*]] = load i32, ptr [[J3]], align 4
 313 // IR-PCH-NEXT:    [[IDXPROM10:%.*]] = sext i32 [[TMP15]] to i64
 314 // IR-PCH-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [10 x i32], ptr [[ARRAYIDX]], i64 0, i64 [[IDXPROM10]]
 315 // IR-PCH-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX11]], align 4
 316 // IR-PCH-NEXT:    [[ADD12:%.*]] = add nsw i32 [[TMP16]], [[TMP13]]
 317 // IR-PCH-NEXT:    store i32 [[ADD12]], ptr [[ARRAYIDX11]], align 4
 318 // IR-PCH-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 319 // IR-PCH:       omp.body.continue:
 320 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 321 // IR-PCH:       omp.inner.for.inc:
 322 // IR-PCH-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 323 // IR-PCH-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP17]], 1
 324 // IR-PCH-NEXT:    store i32 [[ADD13]], ptr [[DOTOMP_IV]], align 4
 325 // IR-PCH-NEXT:    br label [[OMP_INNER_FOR_COND]]
 326 // IR-PCH:       omp.inner.for.end:
 327 // IR-PCH-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 328 // IR-PCH:       omp.loop.exit:
 329 // IR-PCH-NEXT:    [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 330 // IR-PCH-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
 331 // IR-PCH-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP19]])
 332 // IR-PCH-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IS_LAST]], align 4
 333 // IR-PCH-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 334 // IR-PCH-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
 335 // IR-PCH:       .omp.lastprivate.then:
 336 // IR-PCH-NEXT:    store i32 10, ptr [[J3]], align 4
 337 // IR-PCH-NEXT:    [[TMP22:%.*]] = load i32, ptr [[J3]], align 4
 338 // IR-PCH-NEXT:    store i32 [[TMP22]], ptr [[TMP0]], align 4
 339 // IR-PCH-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
 340 // IR-PCH:       .omp.lastprivate.done:
 341 // IR-PCH-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
 342 // IR-PCH-NEXT:    store ptr [[SUM1]], ptr [[TMP23]], align 8
 343 // IR-PCH-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 344 // IR-PCH-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
 345 // IR-PCH-NEXT:    [[TMP26:%.*]] = call i32 @__kmpc_reduce_nowait(ptr @[[GLOB2:[0-9]+]], i32 [[TMP25]], i32 1, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST]], ptr @_Z3foov.omp_outlined.omp.reduction.reduction_func, ptr @.gomp_critical_user_.reduction.var)
 346 // IR-PCH-NEXT:    switch i32 [[TMP26]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
 347 // IR-PCH-NEXT:      i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
 348 // IR-PCH-NEXT:      i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
 349 // IR-PCH-NEXT:    ]
 350 // IR-PCH:       .omp.reduction.case1:
 351 // IR-PCH-NEXT:    [[TMP27:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 352 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP1]], [[TMP27]]
 353 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 354 // IR-PCH:       omp.arraycpy.body:
 355 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 356 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST14:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 357 // IR-PCH-NEXT:    [[TMP28:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
 358 // IR-PCH-NEXT:    [[TMP29:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 359 // IR-PCH-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
 360 // IR-PCH-NEXT:    store i32 [[ADD15]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], align 4
 361 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST14]], i32 1
 362 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 363 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP27]]
 364 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
 365 // IR-PCH:       omp.arraycpy.done18:
 366 // IR-PCH-NEXT:    call void @__kmpc_end_reduce_nowait(ptr @[[GLOB2]], i32 [[TMP25]], ptr @.gomp_critical_user_.reduction.var)
 367 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 368 // IR-PCH:       .omp.reduction.case2:
 369 // IR-PCH-NEXT:    [[TMP30:%.*]] = getelementptr i32, ptr [[TMP1]], i64 100
 370 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq ptr [[TMP1]], [[TMP30]]
 371 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE26:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
 372 // IR-PCH:       omp.arraycpy.body20:
 373 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi ptr [ [[SUM1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT24:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
 374 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi ptr [ [[TMP1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT23:%.*]], [[OMP_ARRAYCPY_BODY20]] ]
 375 // IR-PCH-NEXT:    [[TMP31:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 4
 376 // IR-PCH-NEXT:    [[TMP32:%.*]] = atomicrmw add ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 [[TMP31]] monotonic, align 4
 377 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT23]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
 378 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT24]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
 379 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE25:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT23]], [[TMP30]]
 380 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE25]], label [[OMP_ARRAYCPY_DONE26]], label [[OMP_ARRAYCPY_BODY20]]
 381 // IR-PCH:       omp.arraycpy.done26:
 382 // IR-PCH-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
 383 // IR-PCH:       .omp.reduction.default:
 384 // IR-PCH-NEXT:    ret void
 385 //
 386 //
 387 // IR-PCH-LABEL: define {{[^@]+}}@_Z3foov.omp_outlined.omp.reduction.reduction_func
 388 // IR-PCH-SAME: (ptr noundef [[TMP0:%.*]], ptr noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
 389 // IR-PCH-NEXT:  entry:
 390 // IR-PCH-NEXT:    [[DOTADDR:%.*]] = alloca ptr, align 8
 391 // IR-PCH-NEXT:    [[DOTADDR1:%.*]] = alloca ptr, align 8
 392 // IR-PCH-NEXT:    store ptr [[TMP0]], ptr [[DOTADDR]], align 8
 393 // IR-PCH-NEXT:    store ptr [[TMP1]], ptr [[DOTADDR1]], align 8
 394 // IR-PCH-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTADDR]], align 8
 395 // IR-PCH-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[DOTADDR1]], align 8
 396 // IR-PCH-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
 397 // IR-PCH-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP4]], align 8
 398 // IR-PCH-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP2]], i64 0, i64 0
 399 // IR-PCH-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
 400 // IR-PCH-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i64 100
 401 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[TMP7]], [[TMP8]]
 402 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE2:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
 403 // IR-PCH:       omp.arraycpy.body:
 404 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP5]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 405 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[TMP7]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
 406 // IR-PCH-NEXT:    [[TMP9:%.*]] = load i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 407 // IR-PCH-NEXT:    [[TMP10:%.*]] = load i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 4
 408 // IR-PCH-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
 409 // IR-PCH-NEXT:    store i32 [[ADD]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 4
 410 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
 411 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i32, ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
 412 // IR-PCH-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP8]]
 413 // IR-PCH-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE2]], label [[OMP_ARRAYCPY_BODY]]
 414 // IR-PCH:       omp.arraycpy.done2:
 415 // IR-PCH-NEXT:    ret void
 416 //