clang/test/OpenMP/distribute_parallel_for_proc_bind_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 // add -fopenmp-targets
   3
   4 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
   5 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
   6 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
   7
   8 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
   9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
  10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
  11 // expected-no-diagnostics
  12 #ifndef HEADER
  13 #define HEADER
  14
  15 typedef __INTPTR_TYPE__ intptr_t;
  16
  17
  18 void foo();
  19
  20 struct S {
  21   intptr_t a, b, c;
  22   S(intptr_t a) : a(a) {}
  23   operator char() { return a; }
  24   ~S() {}
  25 };
  26
  27 template <typename T>
  28 T tmain() {
  29 #pragma omp target
  30 #pragma omp teams
  31 #pragma omp distribute parallel for proc_bind(master)
  32   for(int i = 0; i < 1000; i++) {}
  33   return T();
  34 }
  35
  36 int main() {
  37 #pragma omp target
  38 #pragma omp teams
  39 #pragma omp distribute parallel for proc_bind(spread)
  40   for(int i = 0; i < 1000; i++) {}
  41 #pragma omp target
  42 #pragma omp teams
  43 #pragma omp distribute parallel for proc_bind(close)
  44   for(int i = 0; i < 1000; i++) {}
  45   return tmain<int>();
  46 }
  47
  48
  49
  50
  51
  52
  53
  54
  55 #endif
  56 // CHECK1-LABEL: define {{[^@]+}}@main
  57 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
  58 // CHECK1-NEXT:  entry:
  59 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
  60 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
  61 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
  62 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
  63 // CHECK1-NEXT:    [[KERNEL_ARGS2:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
  64 // CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
  65 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
  66 // CHECK1-NEXT:    store i32 2, ptr [[TMP0]], align 4
  67 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
  68 // CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
  69 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
  70 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
  71 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
  72 // CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
  73 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
  74 // CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
  75 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
  76 // CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
  77 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
  78 // CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
  79 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
  80 // CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
  81 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
  82 // CHECK1-NEXT:    store i64 1000, ptr [[TMP8]], align 8
  83 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
  84 // CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
  85 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
  86 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
  87 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
  88 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
  89 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
  90 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
  91 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.region_id, ptr [[KERNEL_ARGS]])
  92 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
  93 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
  94 // CHECK1:       omp_offload.failed:
  95 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37() #[[ATTR2:[0-9]+]]
  96 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
  97 // CHECK1:       omp_offload.cont:
  98 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 0
  99 // CHECK1-NEXT:    store i32 2, ptr [[TMP15]], align 4
 100 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 1
 101 // CHECK1-NEXT:    store i32 0, ptr [[TMP16]], align 4
 102 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 2
 103 // CHECK1-NEXT:    store ptr null, ptr [[TMP17]], align 8
 104 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 3
 105 // CHECK1-NEXT:    store ptr null, ptr [[TMP18]], align 8
 106 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 4
 107 // CHECK1-NEXT:    store ptr null, ptr [[TMP19]], align 8
 108 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 5
 109 // CHECK1-NEXT:    store ptr null, ptr [[TMP20]], align 8
 110 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 6
 111 // CHECK1-NEXT:    store ptr null, ptr [[TMP21]], align 8
 112 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 7
 113 // CHECK1-NEXT:    store ptr null, ptr [[TMP22]], align 8
 114 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 8
 115 // CHECK1-NEXT:    store i64 1000, ptr [[TMP23]], align 8
 116 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 9
 117 // CHECK1-NEXT:    store i64 0, ptr [[TMP24]], align 8
 118 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 10
 119 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP25]], align 4
 120 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 11
 121 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP26]], align 4
 122 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS2]], i32 0, i32 12
 123 // CHECK1-NEXT:    store i32 0, ptr [[TMP27]], align 4
 124 // CHECK1-NEXT:    [[TMP28:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.region_id, ptr [[KERNEL_ARGS2]])
 125 // CHECK1-NEXT:    [[TMP29:%.*]] = icmp ne i32 [[TMP28]], 0
 126 // CHECK1-NEXT:    br i1 [[TMP29]], label [[OMP_OFFLOAD_FAILED3:%.*]], label [[OMP_OFFLOAD_CONT4:%.*]]
 127 // CHECK1:       omp_offload.failed3:
 128 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41() #[[ATTR2]]
 129 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT4]]
 130 // CHECK1:       omp_offload.cont4:
 131 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
 132 // CHECK1-NEXT:    ret i32 [[CALL]]
 133 //
 134 //
 135 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37
 136 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
 137 // CHECK1-NEXT:  entry:
 138 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined)
 139 // CHECK1-NEXT:    ret void
 140 //
 141 //
 142 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined
 143 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 144 // CHECK1-NEXT:  entry:
 145 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 146 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 147 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 148 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 149 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 150 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 151 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 152 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 153 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 154 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 155 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 156 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 157 // CHECK1-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB]], align 4
 158 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 159 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 160 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 161 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 162 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 163 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 164 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
 165 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 166 // CHECK1:       cond.true:
 167 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 168 // CHECK1:       cond.false:
 169 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 170 // CHECK1-NEXT:    br label [[COND_END]]
 171 // CHECK1:       cond.end:
 172 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 173 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 174 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 175 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 176 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 177 // CHECK1:       omp.inner.for.cond:
 178 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 179 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 180 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 181 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 182 // CHECK1:       omp.inner.for.body:
 183 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(ptr @[[GLOB3]], i32 [[TMP1]], i32 4)
 184 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 185 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
 186 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 187 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 188 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
 189 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 190 // CHECK1:       omp.inner.for.inc:
 191 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 192 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 193 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
 194 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 195 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 196 // CHECK1:       omp.inner.for.end:
 197 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 198 // CHECK1:       omp.loop.exit:
 199 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 200 // CHECK1-NEXT:    ret void
 201 //
 202 //
 203 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37.omp_outlined.omp_outlined
 204 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
 205 // CHECK1-NEXT:  entry:
 206 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 207 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 208 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 209 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 210 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 211 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 212 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 213 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 214 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 215 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 216 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 217 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 218 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 219 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 220 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 221 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 222 // CHECK1-NEXT:    store i32 999, ptr [[DOTOMP_UB]], align 4
 223 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 224 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 225 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 226 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
 227 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 228 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 229 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 230 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 231 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 232 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 233 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 234 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 235 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
 236 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 237 // CHECK1:       cond.true:
 238 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 239 // CHECK1:       cond.false:
 240 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 241 // CHECK1-NEXT:    br label [[COND_END]]
 242 // CHECK1:       cond.end:
 243 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
 244 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 245 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 246 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
 247 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 248 // CHECK1:       omp.inner.for.cond:
 249 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 250 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 251 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
 252 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 253 // CHECK1:       omp.inner.for.body:
 254 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 255 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
 256 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 257 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 258 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 259 // CHECK1:       omp.body.continue:
 260 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 261 // CHECK1:       omp.inner.for.inc:
 262 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 263 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
 264 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
 265 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 266 // CHECK1:       omp.inner.for.end:
 267 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 268 // CHECK1:       omp.loop.exit:
 269 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
 270 // CHECK1-NEXT:    ret void
 271 //
 272 //
 273 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41
 274 // CHECK1-SAME: () #[[ATTR1]] {
 275 // CHECK1-NEXT:  entry:
 276 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined)
 277 // CHECK1-NEXT:    ret void
 278 //
 279 //
 280 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined
 281 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 282 // CHECK1-NEXT:  entry:
 283 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 284 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 285 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 286 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 287 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 288 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 289 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 290 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 291 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 292 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 293 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 294 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 295 // CHECK1-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB]], align 4
 296 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 297 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 298 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 299 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 300 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 301 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 302 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
 303 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 304 // CHECK1:       cond.true:
 305 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 306 // CHECK1:       cond.false:
 307 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 308 // CHECK1-NEXT:    br label [[COND_END]]
 309 // CHECK1:       cond.end:
 310 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 311 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 312 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 313 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 314 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 315 // CHECK1:       omp.inner.for.cond:
 316 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 317 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 318 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 319 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 320 // CHECK1:       omp.inner.for.body:
 321 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(ptr @[[GLOB3]], i32 [[TMP1]], i32 3)
 322 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 323 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
 324 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 325 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 326 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
 327 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 328 // CHECK1:       omp.inner.for.inc:
 329 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 330 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 331 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
 332 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 333 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 334 // CHECK1:       omp.inner.for.end:
 335 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 336 // CHECK1:       omp.loop.exit:
 337 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 338 // CHECK1-NEXT:    ret void
 339 //
 340 //
 341 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41.omp_outlined.omp_outlined
 342 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
 343 // CHECK1-NEXT:  entry:
 344 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 345 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 346 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 347 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 348 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 349 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 350 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 351 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 352 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 353 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 354 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 355 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 356 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 357 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 358 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 359 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 360 // CHECK1-NEXT:    store i32 999, ptr [[DOTOMP_UB]], align 4
 361 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 362 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 363 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 364 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
 365 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 366 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 367 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 368 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 369 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 370 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 371 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 372 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 373 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
 374 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 375 // CHECK1:       cond.true:
 376 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 377 // CHECK1:       cond.false:
 378 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 379 // CHECK1-NEXT:    br label [[COND_END]]
 380 // CHECK1:       cond.end:
 381 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
 382 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 383 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 384 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
 385 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 386 // CHECK1:       omp.inner.for.cond:
 387 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 388 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 389 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
 390 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 391 // CHECK1:       omp.inner.for.body:
 392 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 393 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
 394 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 395 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 396 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 397 // CHECK1:       omp.body.continue:
 398 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 399 // CHECK1:       omp.inner.for.inc:
 400 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 401 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
 402 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
 403 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 404 // CHECK1:       omp.inner.for.end:
 405 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 406 // CHECK1:       omp.loop.exit:
 407 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
 408 // CHECK1-NEXT:    ret void
 409 //
 410 //
 411 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
 412 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat {
 413 // CHECK1-NEXT:  entry:
 414 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 415 // CHECK1-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 416 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 417 // CHECK1-NEXT:    store i32 2, ptr [[TMP0]], align 4
 418 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 419 // CHECK1-NEXT:    store i32 0, ptr [[TMP1]], align 4
 420 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 421 // CHECK1-NEXT:    store ptr null, ptr [[TMP2]], align 8
 422 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 423 // CHECK1-NEXT:    store ptr null, ptr [[TMP3]], align 8
 424 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 425 // CHECK1-NEXT:    store ptr null, ptr [[TMP4]], align 8
 426 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 427 // CHECK1-NEXT:    store ptr null, ptr [[TMP5]], align 8
 428 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 429 // CHECK1-NEXT:    store ptr null, ptr [[TMP6]], align 8
 430 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 431 // CHECK1-NEXT:    store ptr null, ptr [[TMP7]], align 8
 432 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 433 // CHECK1-NEXT:    store i64 1000, ptr [[TMP8]], align 8
 434 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 435 // CHECK1-NEXT:    store i64 0, ptr [[TMP9]], align 8
 436 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 437 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP10]], align 4
 438 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 439 // CHECK1-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP11]], align 4
 440 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 441 // CHECK1-NEXT:    store i32 0, ptr [[TMP12]], align 4
 442 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.region_id, ptr [[KERNEL_ARGS]])
 443 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
 444 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 445 // CHECK1:       omp_offload.failed:
 446 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29() #[[ATTR2]]
 447 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 448 // CHECK1:       omp_offload.cont:
 449 // CHECK1-NEXT:    ret i32 0
 450 //
 451 //
 452 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29
 453 // CHECK1-SAME: () #[[ATTR1]] {
 454 // CHECK1-NEXT:  entry:
 455 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 0, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.omp_outlined)
 456 // CHECK1-NEXT:    ret void
 457 //
 458 //
 459 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.omp_outlined
 460 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
 461 // CHECK1-NEXT:  entry:
 462 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 463 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 464 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 465 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 466 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
 467 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
 468 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 469 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 470 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 471 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 472 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 473 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
 474 // CHECK1-NEXT:    store i32 999, ptr [[DOTOMP_COMB_UB]], align 4
 475 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 476 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 477 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 478 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 479 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 480 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 481 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
 482 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 483 // CHECK1:       cond.true:
 484 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 485 // CHECK1:       cond.false:
 486 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 487 // CHECK1-NEXT:    br label [[COND_END]]
 488 // CHECK1:       cond.end:
 489 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 490 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
 491 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 492 // CHECK1-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 493 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 494 // CHECK1:       omp.inner.for.cond:
 495 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 496 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 497 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 498 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 499 // CHECK1:       omp.inner.for.body:
 500 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(ptr @[[GLOB3]], i32 [[TMP1]], i32 2)
 501 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
 502 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
 503 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
 504 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
 505 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 2, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.omp_outlined.omp_outlined, i64 [[TMP8]], i64 [[TMP10]])
 506 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 507 // CHECK1:       omp.inner.for.inc:
 508 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 509 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
 510 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
 511 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
 512 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 513 // CHECK1:       omp.inner.for.end:
 514 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 515 // CHECK1:       omp.loop.exit:
 516 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 517 // CHECK1-NEXT:    ret void
 518 //
 519 //
 520 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.omp_outlined.omp_outlined
 521 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
 522 // CHECK1-NEXT:  entry:
 523 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 524 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 525 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
 526 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
 527 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 528 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 529 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 530 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 531 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 532 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 533 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
 534 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 535 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 536 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 537 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 538 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 539 // CHECK1-NEXT:    store i32 999, ptr [[DOTOMP_UB]], align 4
 540 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
 541 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
 542 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
 543 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
 544 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
 545 // CHECK1-NEXT:    store i32 [[CONV1]], ptr [[DOTOMP_UB]], align 4
 546 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 547 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 548 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 549 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
 550 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 551 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 552 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
 553 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 554 // CHECK1:       cond.true:
 555 // CHECK1-NEXT:    br label [[COND_END:%.*]]
 556 // CHECK1:       cond.false:
 557 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 558 // CHECK1-NEXT:    br label [[COND_END]]
 559 // CHECK1:       cond.end:
 560 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
 561 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 562 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 563 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
 564 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 565 // CHECK1:       omp.inner.for.cond:
 566 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 567 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 568 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
 569 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 570 // CHECK1:       omp.inner.for.body:
 571 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 572 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
 573 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 574 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 575 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 576 // CHECK1:       omp.body.continue:
 577 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 578 // CHECK1:       omp.inner.for.inc:
 579 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 580 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
 581 // CHECK1-NEXT:    store i32 [[ADD3]], ptr [[DOTOMP_IV]], align 4
 582 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
 583 // CHECK1:       omp.inner.for.end:
 584 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 585 // CHECK1:       omp.loop.exit:
 586 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP3]])
 587 // CHECK1-NEXT:    ret void
 588 //
 589 //
 590 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
 591 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
 592 // CHECK1-NEXT:  entry:
 593 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
 594 // CHECK1-NEXT:    ret void
 595 //