clang/test/OpenMP/target_parallel_generic_loop_codegen-2.cpp

   1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _ --version 2
   2 // Test host codegen.
   3 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s
   4 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
   5 // 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
   6 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix CHECK-X86
   7 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
   8 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix CHECK-X86
   9
  10 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0 %s
  11 // 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
  12 // 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 --check-prefix SIMD-ONLY0 %s
  13 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck --check-prefix SIMD-ONLY0-X86 %s
  14 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
  15 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY0-X86 %s
  16
  17 // Test target parallel for codegen - host bc file has to be created first.
  18 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
  19 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s -check-prefix=TCHECK-TARGET
  20 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
  21 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s -check-prefix=TCHECK-TARGET
  22 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
  23 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s -check-prefix=TCHECK-TARGET-X86
  24 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
  25 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s -check-prefix=TCHECK-TARGET-X86
  26
  27 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
  28 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck --check-prefix SIMD-ONLY1-TARGET %s
  29 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
  30 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY1-TARGET %s
  31 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
  32 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck --check-prefix SIMD-ONLY1-TARGET-X86 %s
  33 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
  34 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -fopenmp-is-target-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix SIMD-ONLY1-TARGET-X86 %s
  35
  36 // expected-no-diagnostics
  37 #ifndef HEADER
  38 #define HEADER
  39
  40
  41 int nested(int a){
  42   #pragma omp target parallel loop
  43   for (int i = 0; i < 10; ++i)
  44     ++a;
  45
  46   auto F = [&](){
  47     #pragma omp parallel
  48     {
  49       #pragma omp target parallel loop
  50       for (int i = 0; i < 10; ++i)
  51         ++a;
  52     }
  53   };
  54
  55   F();
  56
  57   return a;
  58 }
  59
  60
  61
  62
  63
  64
  65 // Check metadata is properly generated:
  66
  67 #endif
  68 // CHECK-LABEL: define dso_local noundef signext i32 @_Z6nestedi
  69 // CHECK-SAME: (i32 noundef signext [[A:%.*]]) #[[ATTR0:[0-9]+]] {
  70 // CHECK-NEXT:  entry:
  71 // CHECK-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
  72 // CHECK-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
  73 // CHECK-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 8
  74 // CHECK-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 8
  75 // CHECK-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 8
  76 // CHECK-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
  77 // CHECK-NEXT:    [[F:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
  78 // CHECK-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
  79 // CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
  80 // CHECK-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
  81 // CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[A_CASTED]], align 8
  82 // CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
  83 // CHECK-NEXT:    store i64 [[TMP1]], ptr [[TMP2]], align 8
  84 // CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
  85 // CHECK-NEXT:    store i64 [[TMP1]], ptr [[TMP3]], align 8
  86 // CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
  87 // CHECK-NEXT:    store ptr null, ptr [[TMP4]], align 8
  88 // CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
  89 // CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
  90 // CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
  91 // CHECK-NEXT:    store i32 3, ptr [[TMP7]], align 4
  92 // CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
  93 // CHECK-NEXT:    store i32 1, ptr [[TMP8]], align 4
  94 // CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
  95 // CHECK-NEXT:    store ptr [[TMP5]], ptr [[TMP9]], align 8
  96 // CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
  97 // CHECK-NEXT:    store ptr [[TMP6]], ptr [[TMP10]], align 8
  98 // CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
  99 // CHECK-NEXT:    store ptr @.offload_sizes, ptr [[TMP11]], align 8
 100 // CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 101 // CHECK-NEXT:    store ptr @.offload_maptypes, ptr [[TMP12]], align 8
 102 // CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 103 // CHECK-NEXT:    store ptr null, ptr [[TMP13]], align 8
 104 // CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 105 // CHECK-NEXT:    store ptr null, ptr [[TMP14]], align 8
 106 // CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 107 // CHECK-NEXT:    store i64 0, ptr [[TMP15]], align 8
 108 // CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 109 // CHECK-NEXT:    store i64 0, ptr [[TMP16]], align 8
 110 // CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 111 // CHECK-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP17]], align 4
 112 // CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 113 // CHECK-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP18]], align 4
 114 // CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 115 // CHECK-NEXT:    store i32 0, ptr [[TMP19]], align 4
 116 // CHECK-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 1, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.region_id, ptr [[KERNEL_ARGS]])
 117 // CHECK-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 118 // CHECK-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 119 // CHECK:       omp_offload.failed:
 120 // CHECK-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42(i64 [[TMP1]]) #[[ATTR2:[0-9]+]]
 121 // CHECK-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 122 // CHECK:       omp_offload.cont:
 123 // CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[F]], i32 0, i32 0
 124 // CHECK-NEXT:    store ptr [[A_ADDR]], ptr [[TMP22]], align 8
 125 // CHECK-NEXT:    call void @"_ZZ6nestediENK3$_0clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[F]])
 126 // CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[A_ADDR]], align 4
 127 // CHECK-NEXT:    ret i32 [[TMP23]]
 128 //
 129 //
 130 // CHECK-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42
 131 // CHECK-SAME: (i64 noundef [[A:%.*]]) #[[ATTR1:[0-9]+]] {
 132 // CHECK-NEXT:  entry:
 133 // CHECK-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 134 // CHECK-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
 135 // CHECK-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 136 // CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 137 // CHECK-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 138 // CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[A_CASTED]], align 8
 139 // CHECK-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined, i64 [[TMP1]])
 140 // CHECK-NEXT:    ret void
 141 //
 142 //
 143 // CHECK-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined
 144 // CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR1]] {
 145 // CHECK-NEXT:  entry:
 146 // CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 147 // CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 148 // CHECK-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 149 // CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 150 // CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 151 // CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 152 // CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 153 // CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 154 // CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 155 // CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
 156 // CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 157 // CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 158 // CHECK-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 159 // CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 160 // CHECK-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 161 // CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 162 // CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 163 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 164 // CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 165 // CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 166 // CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 167 // CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 168 // CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 169 // CHECK:       cond.true:
 170 // CHECK-NEXT:    br label [[COND_END:%.*]]
 171 // CHECK:       cond.false:
 172 // CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 173 // CHECK-NEXT:    br label [[COND_END]]
 174 // CHECK:       cond.end:
 175 // CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 176 // CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 177 // CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 178 // CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 179 // CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 180 // CHECK:       omp.inner.for.cond:
 181 // CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 182 // CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 183 // CHECK-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 184 // CHECK-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 185 // CHECK:       omp.inner.for.body:
 186 // CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 187 // CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 188 // CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 189 // CHECK-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 190 // CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 191 // CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 192 // CHECK-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 193 // CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 194 // CHECK:       omp.body.continue:
 195 // CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 196 // CHECK:       omp.inner.for.inc:
 197 // CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 198 // CHECK-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 199 // CHECK-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 200 // CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
 201 // CHECK:       omp.inner.for.end:
 202 // CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 203 // CHECK:       omp.loop.exit:
 204 // CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 205 // CHECK-NEXT:    ret void
 206 //
 207 //
 208 // CHECK-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49
 209 // CHECK-SAME: (i64 noundef [[A:%.*]]) #[[ATTR1]] {
 210 // CHECK-NEXT:  entry:
 211 // CHECK-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 212 // CHECK-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
 213 // CHECK-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 214 // CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 215 // CHECK-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 216 // CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr [[A_CASTED]], align 8
 217 // CHECK-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined, i64 [[TMP1]])
 218 // CHECK-NEXT:    ret void
 219 //
 220 //
 221 // CHECK-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined
 222 // CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR1]] {
 223 // CHECK-NEXT:  entry:
 224 // CHECK-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 225 // CHECK-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 226 // CHECK-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 227 // CHECK-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 228 // CHECK-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 229 // CHECK-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 230 // CHECK-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 231 // CHECK-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 232 // CHECK-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 233 // CHECK-NEXT:    [[I:%.*]] = alloca i32, align 4
 234 // CHECK-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 235 // CHECK-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 236 // CHECK-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 237 // CHECK-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 238 // CHECK-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 239 // CHECK-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 240 // CHECK-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 241 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 242 // CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 243 // CHECK-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 244 // CHECK-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 245 // CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 246 // CHECK-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 247 // CHECK:       cond.true:
 248 // CHECK-NEXT:    br label [[COND_END:%.*]]
 249 // CHECK:       cond.false:
 250 // CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 251 // CHECK-NEXT:    br label [[COND_END]]
 252 // CHECK:       cond.end:
 253 // CHECK-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 254 // CHECK-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 255 // CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 256 // CHECK-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 257 // CHECK-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 258 // CHECK:       omp.inner.for.cond:
 259 // CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 260 // CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 261 // CHECK-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 262 // CHECK-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 263 // CHECK:       omp.inner.for.body:
 264 // CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 265 // CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 266 // CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 267 // CHECK-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 268 // CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 269 // CHECK-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 270 // CHECK-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 271 // CHECK-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 272 // CHECK:       omp.body.continue:
 273 // CHECK-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 274 // CHECK:       omp.inner.for.inc:
 275 // CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 276 // CHECK-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 277 // CHECK-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 278 // CHECK-NEXT:    br label [[OMP_INNER_FOR_COND]]
 279 // CHECK:       omp.inner.for.end:
 280 // CHECK-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 281 // CHECK:       omp.loop.exit:
 282 // CHECK-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 283 // CHECK-NEXT:    ret void
 284 //
 285 //
 286 // CHECK-X86-LABEL: define dso_local noundef i32 @_Z6nestedi
 287 // CHECK-X86-SAME: (i32 noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 288 // CHECK-X86-NEXT:  entry:
 289 // CHECK-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 290 // CHECK-X86-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
 291 // CHECK-X86-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x ptr], align 4
 292 // CHECK-X86-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x ptr], align 4
 293 // CHECK-X86-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x ptr], align 4
 294 // CHECK-X86-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
 295 // CHECK-X86-NEXT:    [[F:%.*]] = alloca [[CLASS_ANON:%.*]], align 4
 296 // CHECK-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 297 // CHECK-X86-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 298 // CHECK-X86-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 299 // CHECK-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_CASTED]], align 4
 300 // CHECK-X86-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 301 // CHECK-X86-NEXT:    store i32 [[TMP1]], ptr [[TMP2]], align 4
 302 // CHECK-X86-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 303 // CHECK-X86-NEXT:    store i32 [[TMP1]], ptr [[TMP3]], align 4
 304 // CHECK-X86-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
 305 // CHECK-X86-NEXT:    store ptr null, ptr [[TMP4]], align 4
 306 // CHECK-X86-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
 307 // CHECK-X86-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
 308 // CHECK-X86-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
 309 // CHECK-X86-NEXT:    store i32 3, ptr [[TMP7]], align 4
 310 // CHECK-X86-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
 311 // CHECK-X86-NEXT:    store i32 1, ptr [[TMP8]], align 4
 312 // CHECK-X86-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
 313 // CHECK-X86-NEXT:    store ptr [[TMP5]], ptr [[TMP9]], align 4
 314 // CHECK-X86-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
 315 // CHECK-X86-NEXT:    store ptr [[TMP6]], ptr [[TMP10]], align 4
 316 // CHECK-X86-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
 317 // CHECK-X86-NEXT:    store ptr @.offload_sizes, ptr [[TMP11]], align 4
 318 // CHECK-X86-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
 319 // CHECK-X86-NEXT:    store ptr @.offload_maptypes, ptr [[TMP12]], align 4
 320 // CHECK-X86-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
 321 // CHECK-X86-NEXT:    store ptr null, ptr [[TMP13]], align 4
 322 // CHECK-X86-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
 323 // CHECK-X86-NEXT:    store ptr null, ptr [[TMP14]], align 4
 324 // CHECK-X86-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
 325 // CHECK-X86-NEXT:    store i64 0, ptr [[TMP15]], align 8
 326 // CHECK-X86-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
 327 // CHECK-X86-NEXT:    store i64 0, ptr [[TMP16]], align 8
 328 // CHECK-X86-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
 329 // CHECK-X86-NEXT:    store [3 x i32] [i32 1, i32 0, i32 0], ptr [[TMP17]], align 4
 330 // CHECK-X86-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
 331 // CHECK-X86-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP18]], align 4
 332 // CHECK-X86-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
 333 // CHECK-X86-NEXT:    store i32 0, ptr [[TMP19]], align 4
 334 // CHECK-X86-NEXT:    [[TMP20:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB2:[0-9]+]], i64 -1, i32 1, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.region_id, ptr [[KERNEL_ARGS]])
 335 // CHECK-X86-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
 336 // CHECK-X86-NEXT:    br i1 [[TMP21]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
 337 // CHECK-X86:       omp_offload.failed:
 338 // CHECK-X86-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42(i32 [[TMP1]]) #[[ATTR2:[0-9]+]]
 339 // CHECK-X86-NEXT:    br label [[OMP_OFFLOAD_CONT]]
 340 // CHECK-X86:       omp_offload.cont:
 341 // CHECK-X86-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[F]], i32 0, i32 0
 342 // CHECK-X86-NEXT:    store ptr [[A_ADDR]], ptr [[TMP22]], align 4
 343 // CHECK-X86-NEXT:    call void @"_ZZ6nestediENK3$_0clEv"(ptr noundef nonnull align 4 dereferenceable(4) [[F]])
 344 // CHECK-X86-NEXT:    [[TMP23:%.*]] = load i32, ptr [[A_ADDR]], align 4
 345 // CHECK-X86-NEXT:    ret i32 [[TMP23]]
 346 //
 347 //
 348 // CHECK-X86-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42
 349 // CHECK-X86-SAME: (i32 noundef [[A:%.*]]) #[[ATTR1:[0-9]+]] {
 350 // CHECK-X86-NEXT:  entry:
 351 // CHECK-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 352 // CHECK-X86-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
 353 // CHECK-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 354 // CHECK-X86-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 355 // CHECK-X86-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 356 // CHECK-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_CASTED]], align 4
 357 // CHECK-X86-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined, i32 [[TMP1]])
 358 // CHECK-X86-NEXT:    ret void
 359 //
 360 //
 361 // CHECK-X86-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined
 362 // CHECK-X86-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
 363 // CHECK-X86-NEXT:  entry:
 364 // CHECK-X86-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 365 // CHECK-X86-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 366 // CHECK-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 367 // CHECK-X86-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 368 // CHECK-X86-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 369 // CHECK-X86-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 370 // CHECK-X86-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 371 // CHECK-X86-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 372 // CHECK-X86-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 373 // CHECK-X86-NEXT:    [[I:%.*]] = alloca i32, align 4
 374 // CHECK-X86-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 375 // CHECK-X86-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 376 // CHECK-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 377 // CHECK-X86-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 378 // CHECK-X86-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 379 // CHECK-X86-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 380 // CHECK-X86-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 381 // CHECK-X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 382 // CHECK-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 383 // CHECK-X86-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 384 // CHECK-X86-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 385 // CHECK-X86-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 386 // CHECK-X86-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 387 // CHECK-X86:       cond.true:
 388 // CHECK-X86-NEXT:    br label [[COND_END:%.*]]
 389 // CHECK-X86:       cond.false:
 390 // CHECK-X86-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 391 // CHECK-X86-NEXT:    br label [[COND_END]]
 392 // CHECK-X86:       cond.end:
 393 // CHECK-X86-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 394 // CHECK-X86-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 395 // CHECK-X86-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 396 // CHECK-X86-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 397 // CHECK-X86-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 398 // CHECK-X86:       omp.inner.for.cond:
 399 // CHECK-X86-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 400 // CHECK-X86-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 401 // CHECK-X86-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 402 // CHECK-X86-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 403 // CHECK-X86:       omp.inner.for.body:
 404 // CHECK-X86-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 405 // CHECK-X86-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 406 // CHECK-X86-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 407 // CHECK-X86-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 408 // CHECK-X86-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 409 // CHECK-X86-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 410 // CHECK-X86-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 411 // CHECK-X86-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 412 // CHECK-X86:       omp.body.continue:
 413 // CHECK-X86-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 414 // CHECK-X86:       omp.inner.for.inc:
 415 // CHECK-X86-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 416 // CHECK-X86-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 417 // CHECK-X86-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 418 // CHECK-X86-NEXT:    br label [[OMP_INNER_FOR_COND]]
 419 // CHECK-X86:       omp.inner.for.end:
 420 // CHECK-X86-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 421 // CHECK-X86:       omp.loop.exit:
 422 // CHECK-X86-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 423 // CHECK-X86-NEXT:    ret void
 424 //
 425 //
 426 // CHECK-X86-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49
 427 // CHECK-X86-SAME: (i32 noundef [[A:%.*]]) #[[ATTR1]] {
 428 // CHECK-X86-NEXT:  entry:
 429 // CHECK-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 430 // CHECK-X86-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
 431 // CHECK-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 432 // CHECK-X86-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 433 // CHECK-X86-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 434 // CHECK-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_CASTED]], align 4
 435 // CHECK-X86-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined, i32 [[TMP1]])
 436 // CHECK-X86-NEXT:    ret void
 437 //
 438 //
 439 // CHECK-X86-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined
 440 // CHECK-X86-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
 441 // CHECK-X86-NEXT:  entry:
 442 // CHECK-X86-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 443 // CHECK-X86-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 444 // CHECK-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 445 // CHECK-X86-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 446 // CHECK-X86-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 447 // CHECK-X86-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 448 // CHECK-X86-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 449 // CHECK-X86-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 450 // CHECK-X86-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 451 // CHECK-X86-NEXT:    [[I:%.*]] = alloca i32, align 4
 452 // CHECK-X86-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 453 // CHECK-X86-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 454 // CHECK-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 455 // CHECK-X86-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 456 // CHECK-X86-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 457 // CHECK-X86-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 458 // CHECK-X86-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 459 // CHECK-X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 460 // CHECK-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 461 // CHECK-X86-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 462 // CHECK-X86-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 463 // CHECK-X86-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 464 // CHECK-X86-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 465 // CHECK-X86:       cond.true:
 466 // CHECK-X86-NEXT:    br label [[COND_END:%.*]]
 467 // CHECK-X86:       cond.false:
 468 // CHECK-X86-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 469 // CHECK-X86-NEXT:    br label [[COND_END]]
 470 // CHECK-X86:       cond.end:
 471 // CHECK-X86-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 472 // CHECK-X86-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 473 // CHECK-X86-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 474 // CHECK-X86-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 475 // CHECK-X86-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 476 // CHECK-X86:       omp.inner.for.cond:
 477 // CHECK-X86-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 478 // CHECK-X86-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 479 // CHECK-X86-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 480 // CHECK-X86-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 481 // CHECK-X86:       omp.inner.for.body:
 482 // CHECK-X86-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 483 // CHECK-X86-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 484 // CHECK-X86-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 485 // CHECK-X86-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 486 // CHECK-X86-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 487 // CHECK-X86-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 488 // CHECK-X86-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 489 // CHECK-X86-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 490 // CHECK-X86:       omp.body.continue:
 491 // CHECK-X86-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 492 // CHECK-X86:       omp.inner.for.inc:
 493 // CHECK-X86-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 494 // CHECK-X86-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 495 // CHECK-X86-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 496 // CHECK-X86-NEXT:    br label [[OMP_INNER_FOR_COND]]
 497 // CHECK-X86:       omp.inner.for.end:
 498 // CHECK-X86-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 499 // CHECK-X86:       omp.loop.exit:
 500 // CHECK-X86-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 501 // CHECK-X86-NEXT:    ret void
 502 //
 503 //
 504 // SIMD-ONLY0-LABEL: define dso_local noundef signext i32 @_Z6nestedi
 505 // SIMD-ONLY0-SAME: (i32 noundef signext [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 506 // SIMD-ONLY0-NEXT:  entry:
 507 // SIMD-ONLY0-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 508 // SIMD-ONLY0-NEXT:    [[I:%.*]] = alloca i32, align 4
 509 // SIMD-ONLY0-NEXT:    [[F:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
 510 // SIMD-ONLY0-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 511 // SIMD-ONLY0-NEXT:    store i32 0, ptr [[I]], align 4
 512 // SIMD-ONLY0-NEXT:    br label [[FOR_COND:%.*]]
 513 // SIMD-ONLY0:       for.cond:
 514 // SIMD-ONLY0-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4
 515 // SIMD-ONLY0-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 10
 516 // SIMD-ONLY0-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
 517 // SIMD-ONLY0:       for.body:
 518 // SIMD-ONLY0-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
 519 // SIMD-ONLY0-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP1]], 1
 520 // SIMD-ONLY0-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 521 // SIMD-ONLY0-NEXT:    br label [[FOR_INC:%.*]]
 522 // SIMD-ONLY0:       for.inc:
 523 // SIMD-ONLY0-NEXT:    [[TMP2:%.*]] = load i32, ptr [[I]], align 4
 524 // SIMD-ONLY0-NEXT:    [[INC1:%.*]] = add nsw i32 [[TMP2]], 1
 525 // SIMD-ONLY0-NEXT:    store i32 [[INC1]], ptr [[I]], align 4
 526 // SIMD-ONLY0-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
 527 // SIMD-ONLY0:       for.end:
 528 // SIMD-ONLY0-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[F]], i32 0, i32 0
 529 // SIMD-ONLY0-NEXT:    store ptr [[A_ADDR]], ptr [[TMP3]], align 8
 530 // SIMD-ONLY0-NEXT:    call void @"_ZZ6nestediENK3$_0clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[F]])
 531 // SIMD-ONLY0-NEXT:    [[TMP4:%.*]] = load i32, ptr [[A_ADDR]], align 4
 532 // SIMD-ONLY0-NEXT:    ret i32 [[TMP4]]
 533 //
 534 //
 535 // SIMD-ONLY0-X86-LABEL: define dso_local noundef i32 @_Z6nestedi
 536 // SIMD-ONLY0-X86-SAME: (i32 noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 537 // SIMD-ONLY0-X86-NEXT:  entry:
 538 // SIMD-ONLY0-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 539 // SIMD-ONLY0-X86-NEXT:    [[I:%.*]] = alloca i32, align 4
 540 // SIMD-ONLY0-X86-NEXT:    [[F:%.*]] = alloca [[CLASS_ANON:%.*]], align 4
 541 // SIMD-ONLY0-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 542 // SIMD-ONLY0-X86-NEXT:    store i32 0, ptr [[I]], align 4
 543 // SIMD-ONLY0-X86-NEXT:    br label [[FOR_COND:%.*]]
 544 // SIMD-ONLY0-X86:       for.cond:
 545 // SIMD-ONLY0-X86-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4
 546 // SIMD-ONLY0-X86-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 10
 547 // SIMD-ONLY0-X86-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
 548 // SIMD-ONLY0-X86:       for.body:
 549 // SIMD-ONLY0-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
 550 // SIMD-ONLY0-X86-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP1]], 1
 551 // SIMD-ONLY0-X86-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 552 // SIMD-ONLY0-X86-NEXT:    br label [[FOR_INC:%.*]]
 553 // SIMD-ONLY0-X86:       for.inc:
 554 // SIMD-ONLY0-X86-NEXT:    [[TMP2:%.*]] = load i32, ptr [[I]], align 4
 555 // SIMD-ONLY0-X86-NEXT:    [[INC1:%.*]] = add nsw i32 [[TMP2]], 1
 556 // SIMD-ONLY0-X86-NEXT:    store i32 [[INC1]], ptr [[I]], align 4
 557 // SIMD-ONLY0-X86-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
 558 // SIMD-ONLY0-X86:       for.end:
 559 // SIMD-ONLY0-X86-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[F]], i32 0, i32 0
 560 // SIMD-ONLY0-X86-NEXT:    store ptr [[A_ADDR]], ptr [[TMP3]], align 4
 561 // SIMD-ONLY0-X86-NEXT:    call void @"_ZZ6nestediENK3$_0clEv"(ptr noundef nonnull align 4 dereferenceable(4) [[F]])
 562 // SIMD-ONLY0-X86-NEXT:    [[TMP4:%.*]] = load i32, ptr [[A_ADDR]], align 4
 563 // SIMD-ONLY0-X86-NEXT:    ret i32 [[TMP4]]
 564 //
 565 //
 566 // TCHECK-TARGET-LABEL: define weak_odr protected void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42
 567 // TCHECK-TARGET-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 568 // TCHECK-TARGET-NEXT:  entry:
 569 // TCHECK-TARGET-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 570 // TCHECK-TARGET-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 571 // TCHECK-TARGET-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
 572 // TCHECK-TARGET-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 573 // TCHECK-TARGET-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 574 // TCHECK-TARGET-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 575 // TCHECK-TARGET-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 576 // TCHECK-TARGET-NEXT:    [[TMP1:%.*]] = load i64, ptr [[A_CASTED]], align 8
 577 // TCHECK-TARGET-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2:[0-9]+]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined, i64 [[TMP1]])
 578 // TCHECK-TARGET-NEXT:    ret void
 579 //
 580 //
 581 // TCHECK-TARGET-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined
 582 // TCHECK-TARGET-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR0]] {
 583 // TCHECK-TARGET-NEXT:  entry:
 584 // TCHECK-TARGET-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 585 // TCHECK-TARGET-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 586 // TCHECK-TARGET-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 587 // TCHECK-TARGET-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 588 // TCHECK-TARGET-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 589 // TCHECK-TARGET-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 590 // TCHECK-TARGET-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 591 // TCHECK-TARGET-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 592 // TCHECK-TARGET-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 593 // TCHECK-TARGET-NEXT:    [[I:%.*]] = alloca i32, align 4
 594 // TCHECK-TARGET-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 595 // TCHECK-TARGET-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 596 // TCHECK-TARGET-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 597 // TCHECK-TARGET-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 598 // TCHECK-TARGET-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 599 // TCHECK-TARGET-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 600 // TCHECK-TARGET-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 601 // TCHECK-TARGET-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 602 // TCHECK-TARGET-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 603 // TCHECK-TARGET-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 604 // TCHECK-TARGET-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 605 // TCHECK-TARGET-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 606 // TCHECK-TARGET-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 607 // TCHECK-TARGET:       cond.true:
 608 // TCHECK-TARGET-NEXT:    br label [[COND_END:%.*]]
 609 // TCHECK-TARGET:       cond.false:
 610 // TCHECK-TARGET-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 611 // TCHECK-TARGET-NEXT:    br label [[COND_END]]
 612 // TCHECK-TARGET:       cond.end:
 613 // TCHECK-TARGET-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 614 // TCHECK-TARGET-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 615 // TCHECK-TARGET-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 616 // TCHECK-TARGET-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 617 // TCHECK-TARGET-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 618 // TCHECK-TARGET:       omp.inner.for.cond:
 619 // TCHECK-TARGET-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 620 // TCHECK-TARGET-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 621 // TCHECK-TARGET-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 622 // TCHECK-TARGET-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 623 // TCHECK-TARGET:       omp.inner.for.body:
 624 // TCHECK-TARGET-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 625 // TCHECK-TARGET-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 626 // TCHECK-TARGET-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 627 // TCHECK-TARGET-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 628 // TCHECK-TARGET-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 629 // TCHECK-TARGET-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 630 // TCHECK-TARGET-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 631 // TCHECK-TARGET-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 632 // TCHECK-TARGET:       omp.body.continue:
 633 // TCHECK-TARGET-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 634 // TCHECK-TARGET:       omp.inner.for.inc:
 635 // TCHECK-TARGET-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 636 // TCHECK-TARGET-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 637 // TCHECK-TARGET-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 638 // TCHECK-TARGET-NEXT:    br label [[OMP_INNER_FOR_COND]]
 639 // TCHECK-TARGET:       omp.inner.for.end:
 640 // TCHECK-TARGET-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 641 // TCHECK-TARGET:       omp.loop.exit:
 642 // TCHECK-TARGET-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 643 // TCHECK-TARGET-NEXT:    ret void
 644 //
 645 //
 646 // TCHECK-TARGET-LABEL: define weak_odr protected void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49
 647 // TCHECK-TARGET-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i64 noundef [[A:%.*]]) #[[ATTR0]] {
 648 // TCHECK-TARGET-NEXT:  entry:
 649 // TCHECK-TARGET-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8
 650 // TCHECK-TARGET-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 651 // TCHECK-TARGET-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
 652 // TCHECK-TARGET-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 8
 653 // TCHECK-TARGET-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 654 // TCHECK-TARGET-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 655 // TCHECK-TARGET-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 656 // TCHECK-TARGET-NEXT:    [[TMP1:%.*]] = load i64, ptr [[A_CASTED]], align 8
 657 // TCHECK-TARGET-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined, i64 [[TMP1]])
 658 // TCHECK-TARGET-NEXT:    ret void
 659 //
 660 //
 661 // TCHECK-TARGET-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined
 662 // TCHECK-TARGET-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR0]] {
 663 // TCHECK-TARGET-NEXT:  entry:
 664 // TCHECK-TARGET-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
 665 // TCHECK-TARGET-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
 666 // TCHECK-TARGET-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
 667 // TCHECK-TARGET-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 668 // TCHECK-TARGET-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 669 // TCHECK-TARGET-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 670 // TCHECK-TARGET-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 671 // TCHECK-TARGET-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 672 // TCHECK-TARGET-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 673 // TCHECK-TARGET-NEXT:    [[I:%.*]] = alloca i32, align 4
 674 // TCHECK-TARGET-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
 675 // TCHECK-TARGET-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
 676 // TCHECK-TARGET-NEXT:    store i64 [[A]], ptr [[A_ADDR]], align 8
 677 // TCHECK-TARGET-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 678 // TCHECK-TARGET-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 679 // TCHECK-TARGET-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 680 // TCHECK-TARGET-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 681 // TCHECK-TARGET-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
 682 // TCHECK-TARGET-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 683 // TCHECK-TARGET-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 684 // TCHECK-TARGET-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 685 // TCHECK-TARGET-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 686 // TCHECK-TARGET-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 687 // TCHECK-TARGET:       cond.true:
 688 // TCHECK-TARGET-NEXT:    br label [[COND_END:%.*]]
 689 // TCHECK-TARGET:       cond.false:
 690 // TCHECK-TARGET-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 691 // TCHECK-TARGET-NEXT:    br label [[COND_END]]
 692 // TCHECK-TARGET:       cond.end:
 693 // TCHECK-TARGET-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 694 // TCHECK-TARGET-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 695 // TCHECK-TARGET-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 696 // TCHECK-TARGET-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 697 // TCHECK-TARGET-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 698 // TCHECK-TARGET:       omp.inner.for.cond:
 699 // TCHECK-TARGET-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 700 // TCHECK-TARGET-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 701 // TCHECK-TARGET-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 702 // TCHECK-TARGET-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 703 // TCHECK-TARGET:       omp.inner.for.body:
 704 // TCHECK-TARGET-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 705 // TCHECK-TARGET-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 706 // TCHECK-TARGET-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 707 // TCHECK-TARGET-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 708 // TCHECK-TARGET-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 709 // TCHECK-TARGET-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 710 // TCHECK-TARGET-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 711 // TCHECK-TARGET-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 712 // TCHECK-TARGET:       omp.body.continue:
 713 // TCHECK-TARGET-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 714 // TCHECK-TARGET:       omp.inner.for.inc:
 715 // TCHECK-TARGET-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 716 // TCHECK-TARGET-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 717 // TCHECK-TARGET-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 718 // TCHECK-TARGET-NEXT:    br label [[OMP_INNER_FOR_COND]]
 719 // TCHECK-TARGET:       omp.inner.for.end:
 720 // TCHECK-TARGET-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 721 // TCHECK-TARGET:       omp.loop.exit:
 722 // TCHECK-TARGET-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 723 // TCHECK-TARGET-NEXT:    ret void
 724 //
 725 //
 726 // TCHECK-TARGET-X86-LABEL: define weak_odr protected void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42
 727 // TCHECK-TARGET-X86-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 728 // TCHECK-TARGET-X86-NEXT:  entry:
 729 // TCHECK-TARGET-X86-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 730 // TCHECK-TARGET-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 731 // TCHECK-TARGET-X86-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
 732 // TCHECK-TARGET-X86-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
 733 // TCHECK-TARGET-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 734 // TCHECK-TARGET-X86-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 735 // TCHECK-TARGET-X86-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 736 // TCHECK-TARGET-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_CASTED]], align 4
 737 // TCHECK-TARGET-X86-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2:[0-9]+]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined, i32 [[TMP1]])
 738 // TCHECK-TARGET-X86-NEXT:    ret void
 739 //
 740 //
 741 // TCHECK-TARGET-X86-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l42.omp_outlined
 742 // TCHECK-TARGET-X86-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0]] {
 743 // TCHECK-TARGET-X86-NEXT:  entry:
 744 // TCHECK-TARGET-X86-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 745 // TCHECK-TARGET-X86-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 746 // TCHECK-TARGET-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 747 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 748 // TCHECK-TARGET-X86-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 749 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 750 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 751 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 752 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 753 // TCHECK-TARGET-X86-NEXT:    [[I:%.*]] = alloca i32, align 4
 754 // TCHECK-TARGET-X86-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 755 // TCHECK-TARGET-X86-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 756 // TCHECK-TARGET-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 757 // TCHECK-TARGET-X86-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 758 // TCHECK-TARGET-X86-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 759 // TCHECK-TARGET-X86-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 760 // TCHECK-TARGET-X86-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 761 // TCHECK-TARGET-X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 762 // TCHECK-TARGET-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 763 // TCHECK-TARGET-X86-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 764 // TCHECK-TARGET-X86-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 765 // TCHECK-TARGET-X86-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 766 // TCHECK-TARGET-X86-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 767 // TCHECK-TARGET-X86:       cond.true:
 768 // TCHECK-TARGET-X86-NEXT:    br label [[COND_END:%.*]]
 769 // TCHECK-TARGET-X86:       cond.false:
 770 // TCHECK-TARGET-X86-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 771 // TCHECK-TARGET-X86-NEXT:    br label [[COND_END]]
 772 // TCHECK-TARGET-X86:       cond.end:
 773 // TCHECK-TARGET-X86-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 774 // TCHECK-TARGET-X86-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 775 // TCHECK-TARGET-X86-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 776 // TCHECK-TARGET-X86-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 777 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 778 // TCHECK-TARGET-X86:       omp.inner.for.cond:
 779 // TCHECK-TARGET-X86-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 780 // TCHECK-TARGET-X86-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 781 // TCHECK-TARGET-X86-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 782 // TCHECK-TARGET-X86-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 783 // TCHECK-TARGET-X86:       omp.inner.for.body:
 784 // TCHECK-TARGET-X86-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 785 // TCHECK-TARGET-X86-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 786 // TCHECK-TARGET-X86-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 787 // TCHECK-TARGET-X86-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 788 // TCHECK-TARGET-X86-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 789 // TCHECK-TARGET-X86-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 790 // TCHECK-TARGET-X86-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 791 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 792 // TCHECK-TARGET-X86:       omp.body.continue:
 793 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 794 // TCHECK-TARGET-X86:       omp.inner.for.inc:
 795 // TCHECK-TARGET-X86-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 796 // TCHECK-TARGET-X86-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 797 // TCHECK-TARGET-X86-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 798 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_INNER_FOR_COND]]
 799 // TCHECK-TARGET-X86:       omp.inner.for.end:
 800 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 801 // TCHECK-TARGET-X86:       omp.loop.exit:
 802 // TCHECK-TARGET-X86-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 803 // TCHECK-TARGET-X86-NEXT:    ret void
 804 //
 805 //
 806 // TCHECK-TARGET-X86-LABEL: define weak_odr protected void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49
 807 // TCHECK-TARGET-X86-SAME: (ptr noalias noundef [[DYN_PTR:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0]] {
 808 // TCHECK-TARGET-X86-NEXT:  entry:
 809 // TCHECK-TARGET-X86-NEXT:    [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 4
 810 // TCHECK-TARGET-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 811 // TCHECK-TARGET-X86-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
 812 // TCHECK-TARGET-X86-NEXT:    store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR]], align 4
 813 // TCHECK-TARGET-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 814 // TCHECK-TARGET-X86-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
 815 // TCHECK-TARGET-X86-NEXT:    store i32 [[TMP0]], ptr [[A_CASTED]], align 4
 816 // TCHECK-TARGET-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_CASTED]], align 4
 817 // TCHECK-TARGET-X86-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB2]], i32 1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined, i32 [[TMP1]])
 818 // TCHECK-TARGET-X86-NEXT:    ret void
 819 //
 820 //
 821 // TCHECK-TARGET-X86-LABEL: define internal void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6nestedi_l49.omp_outlined
 822 // TCHECK-TARGET-X86-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0]] {
 823 // TCHECK-TARGET-X86-NEXT:  entry:
 824 // TCHECK-TARGET-X86-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
 825 // TCHECK-TARGET-X86-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
 826 // TCHECK-TARGET-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 827 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
 828 // TCHECK-TARGET-X86-NEXT:    [[TMP:%.*]] = alloca i32, align 4
 829 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
 830 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
 831 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
 832 // TCHECK-TARGET-X86-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
 833 // TCHECK-TARGET-X86-NEXT:    [[I:%.*]] = alloca i32, align 4
 834 // TCHECK-TARGET-X86-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
 835 // TCHECK-TARGET-X86-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
 836 // TCHECK-TARGET-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 837 // TCHECK-TARGET-X86-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
 838 // TCHECK-TARGET-X86-NEXT:    store i32 9, ptr [[DOTOMP_UB]], align 4
 839 // TCHECK-TARGET-X86-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
 840 // TCHECK-TARGET-X86-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
 841 // TCHECK-TARGET-X86-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
 842 // TCHECK-TARGET-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[TMP0]], align 4
 843 // TCHECK-TARGET-X86-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP1]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
 844 // TCHECK-TARGET-X86-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 845 // TCHECK-TARGET-X86-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
 846 // TCHECK-TARGET-X86-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
 847 // TCHECK-TARGET-X86:       cond.true:
 848 // TCHECK-TARGET-X86-NEXT:    br label [[COND_END:%.*]]
 849 // TCHECK-TARGET-X86:       cond.false:
 850 // TCHECK-TARGET-X86-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 851 // TCHECK-TARGET-X86-NEXT:    br label [[COND_END]]
 852 // TCHECK-TARGET-X86:       cond.end:
 853 // TCHECK-TARGET-X86-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
 854 // TCHECK-TARGET-X86-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
 855 // TCHECK-TARGET-X86-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
 856 // TCHECK-TARGET-X86-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_IV]], align 4
 857 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
 858 // TCHECK-TARGET-X86:       omp.inner.for.cond:
 859 // TCHECK-TARGET-X86-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 860 // TCHECK-TARGET-X86-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
 861 // TCHECK-TARGET-X86-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
 862 // TCHECK-TARGET-X86-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
 863 // TCHECK-TARGET-X86:       omp.inner.for.body:
 864 // TCHECK-TARGET-X86-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 865 // TCHECK-TARGET-X86-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
 866 // TCHECK-TARGET-X86-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
 867 // TCHECK-TARGET-X86-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
 868 // TCHECK-TARGET-X86-NEXT:    [[TMP8:%.*]] = load i32, ptr [[A_ADDR]], align 4
 869 // TCHECK-TARGET-X86-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP8]], 1
 870 // TCHECK-TARGET-X86-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 871 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
 872 // TCHECK-TARGET-X86:       omp.body.continue:
 873 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
 874 // TCHECK-TARGET-X86:       omp.inner.for.inc:
 875 // TCHECK-TARGET-X86-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
 876 // TCHECK-TARGET-X86-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
 877 // TCHECK-TARGET-X86-NEXT:    store i32 [[ADD2]], ptr [[DOTOMP_IV]], align 4
 878 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_INNER_FOR_COND]]
 879 // TCHECK-TARGET-X86:       omp.inner.for.end:
 880 // TCHECK-TARGET-X86-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
 881 // TCHECK-TARGET-X86:       omp.loop.exit:
 882 // TCHECK-TARGET-X86-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP1]])
 883 // TCHECK-TARGET-X86-NEXT:    ret void
 884 //
 885 //
 886 // SIMD-ONLY1-TARGET-LABEL: define dso_local noundef signext i32 @_Z6nestedi
 887 // SIMD-ONLY1-TARGET-SAME: (i32 noundef signext [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 888 // SIMD-ONLY1-TARGET-NEXT:  entry:
 889 // SIMD-ONLY1-TARGET-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 890 // SIMD-ONLY1-TARGET-NEXT:    [[I:%.*]] = alloca i32, align 4
 891 // SIMD-ONLY1-TARGET-NEXT:    [[F:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
 892 // SIMD-ONLY1-TARGET-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 893 // SIMD-ONLY1-TARGET-NEXT:    store i32 0, ptr [[I]], align 4
 894 // SIMD-ONLY1-TARGET-NEXT:    br label [[FOR_COND:%.*]]
 895 // SIMD-ONLY1-TARGET:       for.cond:
 896 // SIMD-ONLY1-TARGET-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4
 897 // SIMD-ONLY1-TARGET-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 10
 898 // SIMD-ONLY1-TARGET-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
 899 // SIMD-ONLY1-TARGET:       for.body:
 900 // SIMD-ONLY1-TARGET-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
 901 // SIMD-ONLY1-TARGET-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP1]], 1
 902 // SIMD-ONLY1-TARGET-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 903 // SIMD-ONLY1-TARGET-NEXT:    br label [[FOR_INC:%.*]]
 904 // SIMD-ONLY1-TARGET:       for.inc:
 905 // SIMD-ONLY1-TARGET-NEXT:    [[TMP2:%.*]] = load i32, ptr [[I]], align 4
 906 // SIMD-ONLY1-TARGET-NEXT:    [[INC1:%.*]] = add nsw i32 [[TMP2]], 1
 907 // SIMD-ONLY1-TARGET-NEXT:    store i32 [[INC1]], ptr [[I]], align 4
 908 // SIMD-ONLY1-TARGET-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
 909 // SIMD-ONLY1-TARGET:       for.end:
 910 // SIMD-ONLY1-TARGET-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[F]], i32 0, i32 0
 911 // SIMD-ONLY1-TARGET-NEXT:    store ptr [[A_ADDR]], ptr [[TMP3]], align 8
 912 // SIMD-ONLY1-TARGET-NEXT:    call void @"_ZZ6nestediENK3$_0clEv"(ptr noundef nonnull align 8 dereferenceable(8) [[F]])
 913 // SIMD-ONLY1-TARGET-NEXT:    [[TMP4:%.*]] = load i32, ptr [[A_ADDR]], align 4
 914 // SIMD-ONLY1-TARGET-NEXT:    ret i32 [[TMP4]]
 915 //
 916 //
 917 // SIMD-ONLY1-TARGET-X86-LABEL: define dso_local noundef i32 @_Z6nestedi
 918 // SIMD-ONLY1-TARGET-X86-SAME: (i32 noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] {
 919 // SIMD-ONLY1-TARGET-X86-NEXT:  entry:
 920 // SIMD-ONLY1-TARGET-X86-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
 921 // SIMD-ONLY1-TARGET-X86-NEXT:    [[I:%.*]] = alloca i32, align 4
 922 // SIMD-ONLY1-TARGET-X86-NEXT:    [[F:%.*]] = alloca [[CLASS_ANON:%.*]], align 4
 923 // SIMD-ONLY1-TARGET-X86-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
 924 // SIMD-ONLY1-TARGET-X86-NEXT:    store i32 0, ptr [[I]], align 4
 925 // SIMD-ONLY1-TARGET-X86-NEXT:    br label [[FOR_COND:%.*]]
 926 // SIMD-ONLY1-TARGET-X86:       for.cond:
 927 // SIMD-ONLY1-TARGET-X86-NEXT:    [[TMP0:%.*]] = load i32, ptr [[I]], align 4
 928 // SIMD-ONLY1-TARGET-X86-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP0]], 10
 929 // SIMD-ONLY1-TARGET-X86-NEXT:    br i1 [[CMP]], label [[FOR_BODY:%.*]], label [[FOR_END:%.*]]
 930 // SIMD-ONLY1-TARGET-X86:       for.body:
 931 // SIMD-ONLY1-TARGET-X86-NEXT:    [[TMP1:%.*]] = load i32, ptr [[A_ADDR]], align 4
 932 // SIMD-ONLY1-TARGET-X86-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP1]], 1
 933 // SIMD-ONLY1-TARGET-X86-NEXT:    store i32 [[INC]], ptr [[A_ADDR]], align 4
 934 // SIMD-ONLY1-TARGET-X86-NEXT:    br label [[FOR_INC:%.*]]
 935 // SIMD-ONLY1-TARGET-X86:       for.inc:
 936 // SIMD-ONLY1-TARGET-X86-NEXT:    [[TMP2:%.*]] = load i32, ptr [[I]], align 4
 937 // SIMD-ONLY1-TARGET-X86-NEXT:    [[INC1:%.*]] = add nsw i32 [[TMP2]], 1
 938 // SIMD-ONLY1-TARGET-X86-NEXT:    store i32 [[INC1]], ptr [[I]], align 4
 939 // SIMD-ONLY1-TARGET-X86-NEXT:    br label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
 940 // SIMD-ONLY1-TARGET-X86:       for.end:
 941 // SIMD-ONLY1-TARGET-X86-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[F]], i32 0, i32 0
 942 // SIMD-ONLY1-TARGET-X86-NEXT:    store ptr [[A_ADDR]], ptr [[TMP3]], align 4
 943 // SIMD-ONLY1-TARGET-X86-NEXT:    call void @"_ZZ6nestediENK3$_0clEv"(ptr noundef nonnull align 4 dereferenceable(4) [[F]])
 944 // SIMD-ONLY1-TARGET-X86-NEXT:    [[TMP4:%.*]] = load i32, ptr [[A_ADDR]], align 4
 945 // SIMD-ONLY1-TARGET-X86-NEXT:    ret i32 [[TMP4]]
 946 //