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 // expected-no-diagnostics 3 #ifndef HEADER 4 #define HEADER 5 // Test host codegen. 6 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1 7 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 8 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1 9 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3 10 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 11 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3 12 13 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 14 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 15 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 16 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 17 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 18 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 19 #ifdef CK1 20 21 int a[100]; 22 23 int teams_argument_global(int n){ 24 int te = n / 128; 25 int th = 128; 26 // discard n_addr 27 28 #pragma omp target 29 #pragma omp teams distribute parallel for num_teams(te), thread_limit(th) 30 for(int i = 0; i < n; i++) { 31 a[i] = 0; 32 #pragma omp cancel for 33 } 34 35 #pragma omp target 36 {{{ 37 #pragma omp teams distribute parallel for 38 for(int i = 0; i < n; i++) { 39 a[i] = 0; 40 } 41 }}} 42 43 // outlined target regions 44 45 46 47 48 return a[0]; 49 } 50 51 #endif // CK1 52 53 // Test host codegen. 54 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9 55 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 56 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9 57 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11 58 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 59 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11 60 61 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 62 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 63 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 64 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 65 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 66 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 67 #ifdef CK2 68 69 int teams_local_arg(void) { 70 int n = 100; 71 int a[n]; 72 73 #pragma omp target 74 #pragma omp teams distribute parallel for 75 for(int i = 0; i < n; i++) { 76 a[i] = 0; 77 } 78 79 // outlined target region 80 81 82 return a[0]; 83 } 84 #endif // CK2 85 86 // Test host codegen. 87 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17 88 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 89 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK17 90 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19 91 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 92 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK19 93 94 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 95 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 96 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 97 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 98 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 99 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 100 #ifdef CK3 101 102 103 template <typename T, int X, long long Y> 104 struct SS{ 105 T a[X]; 106 float b; 107 int foo(void) { 108 109 #pragma omp target 110 #pragma omp teams distribute parallel for 111 for(int i = 0; i < X; i++) { 112 a[i] = (T)0; 113 } 114 115 // outlined target region 116 117 118 return a[0]; 119 } 120 }; 121 122 int teams_template_struct(void) { 123 SS<int, 123, 456> V; 124 return V.foo(); 125 126 } 127 #endif // CK3 128 129 // Test host codegen. 130 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK25 131 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 132 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK25 133 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK27 134 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 135 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK27 136 137 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 138 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 139 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 140 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 141 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 142 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 143 144 #ifdef CK4 145 146 template <typename T, int n> 147 int tmain(T argc) { 148 T a[n]; 149 int te = n/128; 150 int th = 128; 151 #pragma omp target 152 #pragma omp teams distribute parallel for num_teams(te) thread_limit(th) 153 for(int i = 0; i < n; i++) { 154 a[i] = (T)0; 155 } 156 return 0; 157 } 158 159 int main (int argc, char **argv) { 160 int n = 100; 161 int a[n]; 162 #pragma omp target 163 #pragma omp teams distribute parallel for 164 for(int i = 0; i < n; i++) { 165 a[i] = 0; 166 } 167 return tmain<int, 10>(argc); 168 } 169 170 171 172 173 174 175 176 #endif // CK4 177 #endif 178 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_argument_globali 179 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] { 180 // CHECK1-NEXT: entry: 181 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 182 // CHECK1-NEXT: [[TE:%.*]] = alloca i32, align 4 183 // CHECK1-NEXT: [[TH:%.*]] = alloca i32, align 4 184 // CHECK1-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8 185 // CHECK1-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8 186 // CHECK1-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 187 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8 188 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8 189 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8 190 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 191 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 192 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4 193 // CHECK1-NEXT: [[N_CASTED6:%.*]] = alloca i64, align 8 194 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8 195 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8 196 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8 197 // CHECK1-NEXT: [[_TMP11:%.*]] = alloca i32, align 4 198 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4 199 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4 200 // CHECK1-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 201 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4 202 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128 203 // CHECK1-NEXT: store i32 [[DIV]], i32* [[TE]], align 4 204 // CHECK1-NEXT: store i32 128, i32* [[TH]], align 4 205 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4 206 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32* 207 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4 208 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[TE_CASTED]], align 8 209 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4 210 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32* 211 // CHECK1-NEXT: store i32 [[TMP3]], i32* [[CONV1]], align 4 212 // CHECK1-NEXT: [[TMP4:%.*]] = load i64, i64* [[TH_CASTED]], align 8 213 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4 214 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[N_CASTED]] to i32* 215 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[CONV2]], align 4 216 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[N_CASTED]], align 8 217 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 218 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64* 219 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP8]], align 8 220 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 221 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64* 222 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP10]], align 8 223 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 224 // CHECK1-NEXT: store i8* null, i8** [[TMP11]], align 8 225 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 226 // CHECK1-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64* 227 // CHECK1-NEXT: store i64 [[TMP4]], i64* [[TMP13]], align 8 228 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 229 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64* 230 // CHECK1-NEXT: store i64 [[TMP4]], i64* [[TMP15]], align 8 231 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 232 // CHECK1-NEXT: store i8* null, i8** [[TMP16]], align 8 233 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 234 // CHECK1-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64* 235 // CHECK1-NEXT: store i64 [[TMP6]], i64* [[TMP18]], align 8 236 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 237 // CHECK1-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64* 238 // CHECK1-NEXT: store i64 [[TMP6]], i64* [[TMP20]], align 8 239 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2 240 // CHECK1-NEXT: store i8* null, i8** [[TMP21]], align 8 241 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3 242 // CHECK1-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [100 x i32]** 243 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP23]], align 8 244 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3 245 // CHECK1-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [100 x i32]** 246 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP25]], align 8 247 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3 248 // CHECK1-NEXT: store i8* null, i8** [[TMP26]], align 8 249 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 250 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 251 // CHECK1-NEXT: [[TMP29:%.*]] = load i32, i32* [[TE]], align 4 252 // CHECK1-NEXT: [[TMP30:%.*]] = load i32, i32* [[N_ADDR]], align 4 253 // CHECK1-NEXT: store i32 [[TMP30]], i32* [[DOTCAPTURE_EXPR_]], align 4 254 // CHECK1-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 255 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP31]], 0 256 // CHECK1-NEXT: [[DIV4:%.*]] = sdiv i32 [[SUB]], 1 257 // CHECK1-NEXT: [[SUB5:%.*]] = sub nsw i32 [[DIV4]], 1 258 // CHECK1-NEXT: store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4 259 // CHECK1-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4 260 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP32]], 1 261 // CHECK1-NEXT: [[TMP33:%.*]] = zext i32 [[ADD]] to i64 262 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 263 // CHECK1-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 264 // CHECK1-NEXT: store i32 1, i32* [[TMP34]], align 4 265 // CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 266 // CHECK1-NEXT: store i32 4, i32* [[TMP35]], align 4 267 // CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 268 // CHECK1-NEXT: store i8** [[TMP27]], i8*** [[TMP36]], align 8 269 // CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 270 // CHECK1-NEXT: store i8** [[TMP28]], i8*** [[TMP37]], align 8 271 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 272 // CHECK1-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP38]], align 8 273 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 274 // CHECK1-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP39]], align 8 275 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 276 // CHECK1-NEXT: store i8** null, i8*** [[TMP40]], align 8 277 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 278 // CHECK1-NEXT: store i8** null, i8*** [[TMP41]], align 8 279 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 280 // CHECK1-NEXT: store i64 [[TMP33]], i64* [[TMP42]], align 8 281 // CHECK1-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP29]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 282 // CHECK1-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0 283 // CHECK1-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 284 // CHECK1: omp_offload.failed: 285 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28(i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [100 x i32]* @a) #[[ATTR2:[0-9]+]] 286 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] 287 // CHECK1: omp_offload.cont: 288 // CHECK1-NEXT: [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4 289 // CHECK1-NEXT: [[CONV7:%.*]] = bitcast i64* [[N_CASTED6]] to i32* 290 // CHECK1-NEXT: store i32 [[TMP45]], i32* [[CONV7]], align 4 291 // CHECK1-NEXT: [[TMP46:%.*]] = load i64, i64* [[N_CASTED6]], align 8 292 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0 293 // CHECK1-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64* 294 // CHECK1-NEXT: store i64 [[TMP46]], i64* [[TMP48]], align 8 295 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0 296 // CHECK1-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i64* 297 // CHECK1-NEXT: store i64 [[TMP46]], i64* [[TMP50]], align 8 298 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0 299 // CHECK1-NEXT: store i8* null, i8** [[TMP51]], align 8 300 // CHECK1-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1 301 // CHECK1-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [100 x i32]** 302 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP53]], align 8 303 // CHECK1-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1 304 // CHECK1-NEXT: [[TMP55:%.*]] = bitcast i8** [[TMP54]] to [100 x i32]** 305 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP55]], align 8 306 // CHECK1-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1 307 // CHECK1-NEXT: store i8* null, i8** [[TMP56]], align 8 308 // CHECK1-NEXT: [[TMP57:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0 309 // CHECK1-NEXT: [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0 310 // CHECK1-NEXT: [[TMP59:%.*]] = load i32, i32* [[N_ADDR]], align 4 311 // CHECK1-NEXT: store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_12]], align 4 312 // CHECK1-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4 313 // CHECK1-NEXT: [[SUB14:%.*]] = sub nsw i32 [[TMP60]], 0 314 // CHECK1-NEXT: [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1 315 // CHECK1-NEXT: [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1 316 // CHECK1-NEXT: store i32 [[SUB16]], i32* [[DOTCAPTURE_EXPR_13]], align 4 317 // CHECK1-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4 318 // CHECK1-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP61]], 1 319 // CHECK1-NEXT: [[TMP62:%.*]] = zext i32 [[ADD17]] to i64 320 // CHECK1-NEXT: [[KERNEL_ARGS18:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8 321 // CHECK1-NEXT: [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 0 322 // CHECK1-NEXT: store i32 1, i32* [[TMP63]], align 4 323 // CHECK1-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 1 324 // CHECK1-NEXT: store i32 2, i32* [[TMP64]], align 4 325 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 2 326 // CHECK1-NEXT: store i8** [[TMP57]], i8*** [[TMP65]], align 8 327 // CHECK1-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 3 328 // CHECK1-NEXT: store i8** [[TMP58]], i8*** [[TMP66]], align 8 329 // CHECK1-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 4 330 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP67]], align 8 331 // CHECK1-NEXT: [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 5 332 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP68]], align 8 333 // CHECK1-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 6 334 // CHECK1-NEXT: store i8** null, i8*** [[TMP69]], align 8 335 // CHECK1-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 7 336 // CHECK1-NEXT: store i8** null, i8*** [[TMP70]], align 8 337 // CHECK1-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 8 338 // CHECK1-NEXT: store i64 [[TMP62]], i64* [[TMP71]], align 8 339 // CHECK1-NEXT: [[TMP72:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]]) 340 // CHECK1-NEXT: [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0 341 // CHECK1-NEXT: br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]] 342 // CHECK1: omp_offload.failed19: 343 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i64 [[TMP46]], [100 x i32]* @a) #[[ATTR2]] 344 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT20]] 345 // CHECK1: omp_offload.cont20: 346 // CHECK1-NEXT: [[TMP74:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i64 0, i64 0), align 4 347 // CHECK1-NEXT: ret i32 [[TMP74]] 348 // 349 // 350 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28 351 // CHECK1-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], i64 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] { 352 // CHECK1-NEXT: entry: 353 // CHECK1-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8 354 // CHECK1-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8 355 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 356 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 357 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]]) 358 // CHECK1-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8 359 // CHECK1-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8 360 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 361 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 362 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32* 363 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32* 364 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[N_ADDR]] to i32* 365 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 366 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4 367 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4 368 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) 369 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV2]], [100 x i32]* [[TMP1]]) 370 // CHECK1-NEXT: ret void 371 // 372 // 373 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined. 374 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 375 // CHECK1-NEXT: entry: 376 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 377 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 378 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 379 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 380 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 381 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 382 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 383 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 384 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 385 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 386 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 387 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 388 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 389 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4 390 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 391 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 392 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 393 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 394 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 395 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 396 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 397 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 398 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 399 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 400 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 401 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 402 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 403 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4 404 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 405 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 406 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 407 // CHECK1: omp.precond.then: 408 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 409 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 410 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_COMB_UB]], align 4 411 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 412 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 413 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 414 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4 415 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 416 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 417 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 418 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] 419 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 420 // CHECK1: cond.true: 421 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 422 // CHECK1-NEXT: br label [[COND_END:%.*]] 423 // CHECK1: cond.false: 424 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 425 // CHECK1-NEXT: br label [[COND_END]] 426 // CHECK1: cond.end: 427 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] 428 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 429 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 430 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4 431 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 432 // CHECK1: omp.inner.for.cond: 433 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 434 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 435 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] 436 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 437 // CHECK1: omp.inner.for.body: 438 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 439 // CHECK1-NEXT: [[TMP16:%.*]] = zext i32 [[TMP15]] to i64 440 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 441 // CHECK1-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64 442 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP16]], i64 [[TMP18]], i32* [[TMP0]], [100 x i32]* [[TMP1]]) 443 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 444 // CHECK1: omp.inner.for.inc: 445 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 446 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 447 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] 448 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 449 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 450 // CHECK1: omp.inner.for.end: 451 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 452 // CHECK1: omp.loop.exit: 453 // CHECK1-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 454 // CHECK1-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 455 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) 456 // CHECK1-NEXT: br label [[OMP_PRECOND_END]] 457 // CHECK1: omp.precond.end: 458 // CHECK1-NEXT: ret void 459 // 460 // 461 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1 462 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 463 // CHECK1-NEXT: entry: 464 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 465 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 466 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 467 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 468 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 469 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 470 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 471 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 472 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 473 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 474 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 475 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 476 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 477 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 478 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 479 // CHECK1-NEXT: [[I4:%.*]] = alloca i32, align 4 480 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 481 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 482 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 483 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 484 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 485 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 486 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 487 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 488 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 489 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 490 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 491 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 492 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 493 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 494 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 495 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4 496 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 497 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 498 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 499 // CHECK1: omp.precond.then: 500 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 501 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 502 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4 503 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 504 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP6]] to i32 505 // CHECK1-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 506 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32 507 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 508 // CHECK1-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4 509 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 510 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 511 // CHECK1-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 512 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4 513 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 514 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 515 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 516 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]] 517 // CHECK1-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 518 // CHECK1: cond.true: 519 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 520 // CHECK1-NEXT: br label [[COND_END:%.*]] 521 // CHECK1: cond.false: 522 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 523 // CHECK1-NEXT: br label [[COND_END]] 524 // CHECK1: cond.end: 525 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ] 526 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 527 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 528 // CHECK1-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4 529 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 530 // CHECK1: omp.inner.for.cond: 531 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 532 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 533 // CHECK1-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]] 534 // CHECK1-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 535 // CHECK1: omp.inner.for.body: 536 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 537 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1 538 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 539 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I4]], align 4 540 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[I4]], align 4 541 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64 542 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]] 543 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 544 // CHECK1-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 545 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 546 // CHECK1-NEXT: [[TMP21:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP20]], i32 2) 547 // CHECK1-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 548 // CHECK1-NEXT: br i1 [[TMP22]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]] 549 // CHECK1: .cancel.exit: 550 // CHECK1-NEXT: br label [[CANCEL_EXIT:%.*]] 551 // CHECK1: .cancel.continue: 552 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 553 // CHECK1: omp.body.continue: 554 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 555 // CHECK1: omp.inner.for.inc: 556 // CHECK1-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 557 // CHECK1-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP23]], 1 558 // CHECK1-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4 559 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 560 // CHECK1: omp.inner.for.end: 561 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 562 // CHECK1: omp.loop.exit: 563 // CHECK1-NEXT: [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 564 // CHECK1-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4 565 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]]) 566 // CHECK1-NEXT: br label [[OMP_PRECOND_END]] 567 // CHECK1: cancel.exit: 568 // CHECK1-NEXT: [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 569 // CHECK1-NEXT: [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4 570 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]]) 571 // CHECK1-NEXT: br label [[CANCEL_CONT:%.*]] 572 // CHECK1: omp.precond.end: 573 // CHECK1-NEXT: br label [[CANCEL_CONT]] 574 // CHECK1: cancel.cont: 575 // CHECK1-NEXT: ret void 576 // 577 // 578 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35 579 // CHECK1-SAME: (i64 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 580 // CHECK1-NEXT: entry: 581 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 582 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 583 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 584 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 585 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32* 586 // CHECK1-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 587 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], [100 x i32]* [[TMP0]]) 588 // CHECK1-NEXT: ret void 589 // 590 // 591 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2 592 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 593 // CHECK1-NEXT: entry: 594 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 595 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 596 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 597 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 598 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 599 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 600 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 601 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 602 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 603 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 604 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 605 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 606 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 607 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4 608 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 609 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 610 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 611 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 612 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 613 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 614 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 615 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 616 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 617 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 618 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 619 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 620 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 621 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4 622 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 623 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 624 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 625 // CHECK1: omp.precond.then: 626 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 627 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 628 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_COMB_UB]], align 4 629 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 630 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 631 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 632 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4 633 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 634 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 635 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 636 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] 637 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 638 // CHECK1: cond.true: 639 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 640 // CHECK1-NEXT: br label [[COND_END:%.*]] 641 // CHECK1: cond.false: 642 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 643 // CHECK1-NEXT: br label [[COND_END]] 644 // CHECK1: cond.end: 645 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] 646 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 647 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 648 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4 649 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 650 // CHECK1: omp.inner.for.cond: 651 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 652 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 653 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] 654 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 655 // CHECK1: omp.inner.for.body: 656 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 657 // CHECK1-NEXT: [[TMP16:%.*]] = zext i32 [[TMP15]] to i64 658 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 659 // CHECK1-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64 660 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, [100 x i32]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP16]], i64 [[TMP18]], i32* [[TMP0]], [100 x i32]* [[TMP1]]) 661 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 662 // CHECK1: omp.inner.for.inc: 663 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 664 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 665 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] 666 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 667 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 668 // CHECK1: omp.inner.for.end: 669 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 670 // CHECK1: omp.loop.exit: 671 // CHECK1-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 672 // CHECK1-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 673 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) 674 // CHECK1-NEXT: br label [[OMP_PRECOND_END]] 675 // CHECK1: omp.precond.end: 676 // CHECK1-NEXT: ret void 677 // 678 // 679 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3 680 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 681 // CHECK1-NEXT: entry: 682 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 683 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 684 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 685 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 686 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 687 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8 688 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 689 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 690 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 691 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 692 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 693 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 694 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 695 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 696 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 697 // CHECK1-NEXT: [[I4:%.*]] = alloca i32, align 4 698 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 699 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 700 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 701 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 702 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 703 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8 704 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 705 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8 706 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 707 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 708 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 709 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 710 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 711 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 712 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 713 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4 714 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 715 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 716 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 717 // CHECK1: omp.precond.then: 718 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 719 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 720 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4 721 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 722 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP6]] to i32 723 // CHECK1-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 724 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32 725 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 726 // CHECK1-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4 727 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 728 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 729 // CHECK1-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 730 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4 731 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 732 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 733 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 734 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]] 735 // CHECK1-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 736 // CHECK1: cond.true: 737 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 738 // CHECK1-NEXT: br label [[COND_END:%.*]] 739 // CHECK1: cond.false: 740 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 741 // CHECK1-NEXT: br label [[COND_END]] 742 // CHECK1: cond.end: 743 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ] 744 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 745 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 746 // CHECK1-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4 747 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 748 // CHECK1: omp.inner.for.cond: 749 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 750 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 751 // CHECK1-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]] 752 // CHECK1-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 753 // CHECK1: omp.inner.for.body: 754 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 755 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1 756 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 757 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I4]], align 4 758 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[I4]], align 4 759 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64 760 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]] 761 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 762 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 763 // CHECK1: omp.body.continue: 764 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 765 // CHECK1: omp.inner.for.inc: 766 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 767 // CHECK1-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1 768 // CHECK1-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4 769 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 770 // CHECK1: omp.inner.for.end: 771 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 772 // CHECK1: omp.loop.exit: 773 // CHECK1-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 774 // CHECK1-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4 775 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]]) 776 // CHECK1-NEXT: br label [[OMP_PRECOND_END]] 777 // CHECK1: omp.precond.end: 778 // CHECK1-NEXT: ret void 779 // 780 // 781 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 782 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] { 783 // CHECK1-NEXT: entry: 784 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1) 785 // CHECK1-NEXT: ret void 786 // 787 // 788 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_argument_globali 789 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] { 790 // CHECK3-NEXT: entry: 791 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 792 // CHECK3-NEXT: [[TE:%.*]] = alloca i32, align 4 793 // CHECK3-NEXT: [[TH:%.*]] = alloca i32, align 4 794 // CHECK3-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4 795 // CHECK3-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4 796 // CHECK3-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 797 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4 798 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4 799 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4 800 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 801 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 802 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 803 // CHECK3-NEXT: [[N_CASTED4:%.*]] = alloca i32, align 4 804 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4 805 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4 806 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4 807 // CHECK3-NEXT: [[_TMP8:%.*]] = alloca i32, align 4 808 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4 809 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4 810 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 811 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4 812 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128 813 // CHECK3-NEXT: store i32 [[DIV]], i32* [[TE]], align 4 814 // CHECK3-NEXT: store i32 128, i32* [[TH]], align 4 815 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4 816 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TE_CASTED]], align 4 817 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_CASTED]], align 4 818 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4 819 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[TH_CASTED]], align 4 820 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TH_CASTED]], align 4 821 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4 822 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[N_CASTED]], align 4 823 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[N_CASTED]], align 4 824 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 825 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32* 826 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP8]], align 4 827 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 828 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32* 829 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP10]], align 4 830 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 831 // CHECK3-NEXT: store i8* null, i8** [[TMP11]], align 4 832 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 833 // CHECK3-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32* 834 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP13]], align 4 835 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 836 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32* 837 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP15]], align 4 838 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 839 // CHECK3-NEXT: store i8* null, i8** [[TMP16]], align 4 840 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 841 // CHECK3-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32* 842 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP18]], align 4 843 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 844 // CHECK3-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32* 845 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP20]], align 4 846 // CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2 847 // CHECK3-NEXT: store i8* null, i8** [[TMP21]], align 4 848 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3 849 // CHECK3-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [100 x i32]** 850 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP23]], align 4 851 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3 852 // CHECK3-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [100 x i32]** 853 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP25]], align 4 854 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3 855 // CHECK3-NEXT: store i8* null, i8** [[TMP26]], align 4 856 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 857 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 858 // CHECK3-NEXT: [[TMP29:%.*]] = load i32, i32* [[TE]], align 4 859 // CHECK3-NEXT: [[TMP30:%.*]] = load i32, i32* [[N_ADDR]], align 4 860 // CHECK3-NEXT: store i32 [[TMP30]], i32* [[DOTCAPTURE_EXPR_]], align 4 861 // CHECK3-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 862 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP31]], 0 863 // CHECK3-NEXT: [[DIV2:%.*]] = sdiv i32 [[SUB]], 1 864 // CHECK3-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1 865 // CHECK3-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_1]], align 4 866 // CHECK3-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 867 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP32]], 1 868 // CHECK3-NEXT: [[TMP33:%.*]] = zext i32 [[ADD]] to i64 869 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 870 // CHECK3-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 871 // CHECK3-NEXT: store i32 1, i32* [[TMP34]], align 4 872 // CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 873 // CHECK3-NEXT: store i32 4, i32* [[TMP35]], align 4 874 // CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 875 // CHECK3-NEXT: store i8** [[TMP27]], i8*** [[TMP36]], align 4 876 // CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 877 // CHECK3-NEXT: store i8** [[TMP28]], i8*** [[TMP37]], align 4 878 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 879 // CHECK3-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP38]], align 4 880 // CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 881 // CHECK3-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP39]], align 4 882 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 883 // CHECK3-NEXT: store i8** null, i8*** [[TMP40]], align 4 884 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 885 // CHECK3-NEXT: store i8** null, i8*** [[TMP41]], align 4 886 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 887 // CHECK3-NEXT: store i64 [[TMP33]], i64* [[TMP42]], align 8 888 // CHECK3-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 [[TMP29]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 889 // CHECK3-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0 890 // CHECK3-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 891 // CHECK3: omp_offload.failed: 892 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28(i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [100 x i32]* @a) #[[ATTR2:[0-9]+]] 893 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] 894 // CHECK3: omp_offload.cont: 895 // CHECK3-NEXT: [[TMP45:%.*]] = load i32, i32* [[N_ADDR]], align 4 896 // CHECK3-NEXT: store i32 [[TMP45]], i32* [[N_CASTED4]], align 4 897 // CHECK3-NEXT: [[TMP46:%.*]] = load i32, i32* [[N_CASTED4]], align 4 898 // CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0 899 // CHECK3-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32* 900 // CHECK3-NEXT: store i32 [[TMP46]], i32* [[TMP48]], align 4 901 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0 902 // CHECK3-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32* 903 // CHECK3-NEXT: store i32 [[TMP46]], i32* [[TMP50]], align 4 904 // CHECK3-NEXT: [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0 905 // CHECK3-NEXT: store i8* null, i8** [[TMP51]], align 4 906 // CHECK3-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1 907 // CHECK3-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [100 x i32]** 908 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP53]], align 4 909 // CHECK3-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1 910 // CHECK3-NEXT: [[TMP55:%.*]] = bitcast i8** [[TMP54]] to [100 x i32]** 911 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP55]], align 4 912 // CHECK3-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1 913 // CHECK3-NEXT: store i8* null, i8** [[TMP56]], align 4 914 // CHECK3-NEXT: [[TMP57:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0 915 // CHECK3-NEXT: [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0 916 // CHECK3-NEXT: [[TMP59:%.*]] = load i32, i32* [[N_ADDR]], align 4 917 // CHECK3-NEXT: store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_9]], align 4 918 // CHECK3-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4 919 // CHECK3-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP60]], 0 920 // CHECK3-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1 921 // CHECK3-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1 922 // CHECK3-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4 923 // CHECK3-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4 924 // CHECK3-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP61]], 1 925 // CHECK3-NEXT: [[TMP62:%.*]] = zext i32 [[ADD14]] to i64 926 // CHECK3-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8 927 // CHECK3-NEXT: [[TMP63:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0 928 // CHECK3-NEXT: store i32 1, i32* [[TMP63]], align 4 929 // CHECK3-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1 930 // CHECK3-NEXT: store i32 2, i32* [[TMP64]], align 4 931 // CHECK3-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2 932 // CHECK3-NEXT: store i8** [[TMP57]], i8*** [[TMP65]], align 4 933 // CHECK3-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3 934 // CHECK3-NEXT: store i8** [[TMP58]], i8*** [[TMP66]], align 4 935 // CHECK3-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4 936 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP67]], align 4 937 // CHECK3-NEXT: [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5 938 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP68]], align 4 939 // CHECK3-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6 940 // CHECK3-NEXT: store i8** null, i8*** [[TMP69]], align 4 941 // CHECK3-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7 942 // CHECK3-NEXT: store i8** null, i8*** [[TMP70]], align 4 943 // CHECK3-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8 944 // CHECK3-NEXT: store i64 [[TMP62]], i64* [[TMP71]], align 8 945 // CHECK3-NEXT: [[TMP72:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]]) 946 // CHECK3-NEXT: [[TMP73:%.*]] = icmp ne i32 [[TMP72]], 0 947 // CHECK3-NEXT: br i1 [[TMP73]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]] 948 // CHECK3: omp_offload.failed16: 949 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i32 [[TMP46]], [100 x i32]* @a) #[[ATTR2]] 950 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT17]] 951 // CHECK3: omp_offload.cont17: 952 // CHECK3-NEXT: [[TMP74:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i32 0, i32 0), align 4 953 // CHECK3-NEXT: ret i32 [[TMP74]] 954 // 955 // 956 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l28 957 // CHECK3-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], i32 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] { 958 // CHECK3-NEXT: entry: 959 // CHECK3-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4 960 // CHECK3-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4 961 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 962 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 963 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]]) 964 // CHECK3-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4 965 // CHECK3-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4 966 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 967 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 968 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 969 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4 970 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4 971 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) 972 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP1]]) 973 // CHECK3-NEXT: ret void 974 // 975 // 976 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined. 977 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 978 // CHECK3-NEXT: entry: 979 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 980 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 981 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 982 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 983 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 984 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 985 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 986 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 987 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 988 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 989 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 990 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 991 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 992 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 993 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 994 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 995 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 996 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 997 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 998 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 999 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 1000 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 1001 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1002 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 1003 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1004 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1005 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1006 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4 1007 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1008 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 1009 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1010 // CHECK3: omp.precond.then: 1011 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1012 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1013 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_COMB_UB]], align 4 1014 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1015 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1016 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1017 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4 1018 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1019 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1020 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1021 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] 1022 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1023 // CHECK3: cond.true: 1024 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1025 // CHECK3-NEXT: br label [[COND_END:%.*]] 1026 // CHECK3: cond.false: 1027 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1028 // CHECK3-NEXT: br label [[COND_END]] 1029 // CHECK3: cond.end: 1030 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] 1031 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1032 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1033 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4 1034 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1035 // CHECK3: omp.inner.for.cond: 1036 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1037 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1038 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] 1039 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1040 // CHECK3: omp.inner.for.body: 1041 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1042 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1043 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP15]], i32 [[TMP16]], i32* [[TMP0]], [100 x i32]* [[TMP1]]) 1044 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1045 // CHECK3: omp.inner.for.inc: 1046 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1047 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1048 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]] 1049 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1050 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1051 // CHECK3: omp.inner.for.end: 1052 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1053 // CHECK3: omp.loop.exit: 1054 // CHECK3-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1055 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 1056 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) 1057 // CHECK3-NEXT: br label [[OMP_PRECOND_END]] 1058 // CHECK3: omp.precond.end: 1059 // CHECK3-NEXT: ret void 1060 // 1061 // 1062 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1 1063 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 1064 // CHECK3-NEXT: entry: 1065 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1066 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1067 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1068 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1069 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 1070 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 1071 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1072 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1073 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1074 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1075 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1076 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1077 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1078 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1079 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1080 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 1081 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1082 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1083 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1084 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1085 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 1086 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 1087 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 1088 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 1089 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 1090 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 1091 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1092 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 1093 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1094 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1095 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1096 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4 1097 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1098 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 1099 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1100 // CHECK3: omp.precond.then: 1101 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1102 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1103 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4 1104 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1105 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1106 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_LB]], align 4 1107 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_UB]], align 4 1108 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1109 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1110 // CHECK3-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1111 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4 1112 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1113 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1114 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1115 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]] 1116 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1117 // CHECK3: cond.true: 1118 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1119 // CHECK3-NEXT: br label [[COND_END:%.*]] 1120 // CHECK3: cond.false: 1121 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1122 // CHECK3-NEXT: br label [[COND_END]] 1123 // CHECK3: cond.end: 1124 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ] 1125 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1126 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1127 // CHECK3-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4 1128 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1129 // CHECK3: omp.inner.for.cond: 1130 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1131 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1132 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]] 1133 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1134 // CHECK3: omp.inner.for.body: 1135 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1136 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1 1137 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1138 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4 1139 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[I3]], align 4 1140 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP18]] 1141 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 1142 // CHECK3-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1143 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 1144 // CHECK3-NEXT: [[TMP21:%.*]] = call i32 @__kmpc_cancel(%struct.ident_t* @[[GLOB3]], i32 [[TMP20]], i32 2) 1145 // CHECK3-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0 1146 // CHECK3-NEXT: br i1 [[TMP22]], label [[DOTCANCEL_EXIT:%.*]], label [[DOTCANCEL_CONTINUE:%.*]] 1147 // CHECK3: .cancel.exit: 1148 // CHECK3-NEXT: br label [[CANCEL_EXIT:%.*]] 1149 // CHECK3: .cancel.continue: 1150 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1151 // CHECK3: omp.body.continue: 1152 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1153 // CHECK3: omp.inner.for.inc: 1154 // CHECK3-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1155 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP23]], 1 1156 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4 1157 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1158 // CHECK3: omp.inner.for.end: 1159 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1160 // CHECK3: omp.loop.exit: 1161 // CHECK3-NEXT: [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1162 // CHECK3-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4 1163 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]]) 1164 // CHECK3-NEXT: br label [[OMP_PRECOND_END]] 1165 // CHECK3: cancel.exit: 1166 // CHECK3-NEXT: [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1167 // CHECK3-NEXT: [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4 1168 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]]) 1169 // CHECK3-NEXT: br label [[CANCEL_CONT:%.*]] 1170 // CHECK3: omp.precond.end: 1171 // CHECK3-NEXT: br label [[CANCEL_CONT]] 1172 // CHECK3: cancel.cont: 1173 // CHECK3-NEXT: ret void 1174 // 1175 // 1176 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35 1177 // CHECK3-SAME: (i32 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 1178 // CHECK3-NEXT: entry: 1179 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 1180 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 1181 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 1182 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 1183 // CHECK3-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 1184 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP0]]) 1185 // CHECK3-NEXT: ret void 1186 // 1187 // 1188 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2 1189 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 1190 // CHECK3-NEXT: entry: 1191 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1192 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1193 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 1194 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 1195 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1196 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1197 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1198 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1199 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1200 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1201 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1202 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1203 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1204 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 1205 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1206 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1207 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 1208 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 1209 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 1210 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 1211 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 1212 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 1213 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1214 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 1215 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1216 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1217 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1218 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4 1219 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1220 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 1221 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1222 // CHECK3: omp.precond.then: 1223 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1224 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1225 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_COMB_UB]], align 4 1226 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1227 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1228 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1229 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4 1230 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1231 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1232 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1233 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]] 1234 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1235 // CHECK3: cond.true: 1236 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1237 // CHECK3-NEXT: br label [[COND_END:%.*]] 1238 // CHECK3: cond.false: 1239 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1240 // CHECK3-NEXT: br label [[COND_END]] 1241 // CHECK3: cond.end: 1242 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ] 1243 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1244 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1245 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4 1246 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1247 // CHECK3: omp.inner.for.cond: 1248 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1249 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1250 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]] 1251 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1252 // CHECK3: omp.inner.for.body: 1253 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1254 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1255 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, [100 x i32]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP15]], i32 [[TMP16]], i32* [[TMP0]], [100 x i32]* [[TMP1]]) 1256 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1257 // CHECK3: omp.inner.for.inc: 1258 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1259 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1260 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]] 1261 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1262 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1263 // CHECK3: omp.inner.for.end: 1264 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1265 // CHECK3: omp.loop.exit: 1266 // CHECK3-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1267 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4 1268 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]]) 1269 // CHECK3-NEXT: br label [[OMP_PRECOND_END]] 1270 // CHECK3: omp.precond.end: 1271 // CHECK3-NEXT: ret void 1272 // 1273 // 1274 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3 1275 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] { 1276 // CHECK3-NEXT: entry: 1277 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1278 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1279 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1280 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1281 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 1282 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4 1283 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1284 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1285 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1286 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1287 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1288 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1289 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1290 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1291 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1292 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4 1293 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1294 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1295 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1296 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1297 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 1298 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4 1299 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 1300 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4 1301 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4 1302 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4 1303 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1304 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0 1305 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1306 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1307 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1308 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4 1309 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1310 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]] 1311 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1312 // CHECK3: omp.precond.then: 1313 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1314 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1315 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4 1316 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1317 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1318 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_LB]], align 4 1319 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_UB]], align 4 1320 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1321 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1322 // CHECK3-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1323 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4 1324 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP9]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1325 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1326 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1327 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]] 1328 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1329 // CHECK3: cond.true: 1330 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1331 // CHECK3-NEXT: br label [[COND_END:%.*]] 1332 // CHECK3: cond.false: 1333 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1334 // CHECK3-NEXT: br label [[COND_END]] 1335 // CHECK3: cond.end: 1336 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ] 1337 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1338 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1339 // CHECK3-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4 1340 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1341 // CHECK3: omp.inner.for.cond: 1342 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1343 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1344 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]] 1345 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1346 // CHECK3: omp.inner.for.body: 1347 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1348 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1 1349 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1350 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4 1351 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[I3]], align 4 1352 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP18]] 1353 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 1354 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1355 // CHECK3: omp.body.continue: 1356 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1357 // CHECK3: omp.inner.for.inc: 1358 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1359 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP19]], 1 1360 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4 1361 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1362 // CHECK3: omp.inner.for.end: 1363 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1364 // CHECK3: omp.loop.exit: 1365 // CHECK3-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1366 // CHECK3-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4 1367 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]]) 1368 // CHECK3-NEXT: br label [[OMP_PRECOND_END]] 1369 // CHECK3: omp.precond.end: 1370 // CHECK3-NEXT: ret void 1371 // 1372 // 1373 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1374 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] { 1375 // CHECK3-NEXT: entry: 1376 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1) 1377 // CHECK3-NEXT: ret void 1378 // 1379 // 1380 // CHECK9-LABEL: define {{[^@]+}}@_Z15teams_local_argv 1381 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] { 1382 // CHECK9-NEXT: entry: 1383 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4 1384 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8 1385 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8 1386 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 1387 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8 1388 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8 1389 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8 1390 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8 1391 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1392 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1393 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1394 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4 1395 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4 1396 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64 1397 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave() 1398 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8 1399 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4 1400 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8 1401 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4 1402 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32* 1403 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4 1404 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8 1405 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4 1406 // CHECK9-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8* 1407 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false) 1408 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1409 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64* 1410 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8 1411 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1412 // CHECK9-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64* 1413 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8 1414 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 1415 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8 1416 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 1417 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64* 1418 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8 1419 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 1420 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64* 1421 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8 1422 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 1423 // CHECK9-NEXT: store i8* null, i8** [[TMP16]], align 8 1424 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 1425 // CHECK9-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32** 1426 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8 1427 // CHECK9-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 1428 // CHECK9-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32** 1429 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8 1430 // CHECK9-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2 1431 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8 1432 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2 1433 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8 1434 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1435 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1436 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0 1437 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4 1438 // CHECK9-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4 1439 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1440 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0 1441 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1442 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1443 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1444 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1445 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1 1446 // CHECK9-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64 1447 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 1448 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 1449 // CHECK9-NEXT: store i32 1, i32* [[TMP30]], align 4 1450 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 1451 // CHECK9-NEXT: store i32 3, i32* [[TMP31]], align 4 1452 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 1453 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8 1454 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 1455 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8 1456 // CHECK9-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 1457 // CHECK9-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8 1458 // CHECK9-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 1459 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8 1460 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 1461 // CHECK9-NEXT: store i8** null, i8*** [[TMP36]], align 8 1462 // CHECK9-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 1463 // CHECK9-NEXT: store i8** null, i8*** [[TMP37]], align 8 1464 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 1465 // CHECK9-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8 1466 // CHECK9-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 1467 // CHECK9-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0 1468 // CHECK9-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1469 // CHECK9: omp_offload.failed: 1470 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]] 1471 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]] 1472 // CHECK9: omp_offload.cont: 1473 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 0 1474 // CHECK9-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 1475 // CHECK9-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8 1476 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP42]]) 1477 // CHECK9-NEXT: ret i32 [[TMP41]] 1478 // 1479 // 1480 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73 1481 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] { 1482 // CHECK9-NEXT: entry: 1483 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 1484 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 1485 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 1486 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 1487 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 1488 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 1489 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32* 1490 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 1491 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8 1492 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]]) 1493 // CHECK9-NEXT: ret void 1494 // 1495 // 1496 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined. 1497 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 1498 // CHECK9-NEXT: entry: 1499 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1500 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1501 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 1502 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 1503 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 1504 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1505 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1506 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1507 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1508 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 1509 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1510 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1511 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1512 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1513 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4 1514 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1515 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1516 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 1517 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 1518 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 1519 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 1520 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 1521 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8 1522 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 1523 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 1524 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1525 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 1526 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1527 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1528 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1529 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4 1530 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1531 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 1532 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1533 // CHECK9: omp.precond.then: 1534 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1535 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1536 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4 1537 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1538 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1539 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1540 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 1541 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1542 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1543 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1544 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] 1545 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1546 // CHECK9: cond.true: 1547 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1548 // CHECK9-NEXT: br label [[COND_END:%.*]] 1549 // CHECK9: cond.false: 1550 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1551 // CHECK9-NEXT: br label [[COND_END]] 1552 // CHECK9: cond.end: 1553 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] 1554 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1555 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1556 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 1557 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1558 // CHECK9: omp.inner.for.cond: 1559 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1560 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1561 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] 1562 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1563 // CHECK9: omp.inner.for.body: 1564 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1565 // CHECK9-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64 1566 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1567 // CHECK9-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64 1568 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]) 1569 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1570 // CHECK9: omp.inner.for.inc: 1571 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1572 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1573 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]] 1574 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1575 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]] 1576 // CHECK9: omp.inner.for.end: 1577 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1578 // CHECK9: omp.loop.exit: 1579 // CHECK9-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1580 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4 1581 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]]) 1582 // CHECK9-NEXT: br label [[OMP_PRECOND_END]] 1583 // CHECK9: omp.precond.end: 1584 // CHECK9-NEXT: ret void 1585 // 1586 // 1587 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1 1588 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 1589 // CHECK9-NEXT: entry: 1590 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1591 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1592 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 1593 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 1594 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 1595 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 1596 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 1597 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1598 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1599 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1600 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1601 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 1602 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1603 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1604 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1605 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1606 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4 1607 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1608 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1609 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 1610 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 1611 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 1612 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 1613 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 1614 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 1615 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 1616 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8 1617 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 1618 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 1619 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1620 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 1621 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1622 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1623 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1624 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4 1625 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1626 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 1627 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1628 // CHECK9: omp.precond.then: 1629 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1630 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1631 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 1632 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 1633 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32 1634 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 1635 // CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32 1636 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 1637 // CHECK9-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4 1638 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1639 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1640 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1641 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4 1642 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1643 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1644 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1645 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]] 1646 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1647 // CHECK9: cond.true: 1648 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1649 // CHECK9-NEXT: br label [[COND_END:%.*]] 1650 // CHECK9: cond.false: 1651 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1652 // CHECK9-NEXT: br label [[COND_END]] 1653 // CHECK9: cond.end: 1654 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ] 1655 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1656 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1657 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4 1658 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1659 // CHECK9: omp.inner.for.cond: 1660 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1661 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1662 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]] 1663 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1664 // CHECK9: omp.inner.for.body: 1665 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1666 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1 1667 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1668 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4 1669 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4 1670 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64 1671 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]] 1672 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 1673 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1674 // CHECK9: omp.body.continue: 1675 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1676 // CHECK9: omp.inner.for.inc: 1677 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1678 // CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1 1679 // CHECK9-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4 1680 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]] 1681 // CHECK9: omp.inner.for.end: 1682 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1683 // CHECK9: omp.loop.exit: 1684 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1685 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 1686 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) 1687 // CHECK9-NEXT: br label [[OMP_PRECOND_END]] 1688 // CHECK9: omp.precond.end: 1689 // CHECK9-NEXT: ret void 1690 // 1691 // 1692 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1693 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] { 1694 // CHECK9-NEXT: entry: 1695 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1) 1696 // CHECK9-NEXT: ret void 1697 // 1698 // 1699 // CHECK11-LABEL: define {{[^@]+}}@_Z15teams_local_argv 1700 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] { 1701 // CHECK11-NEXT: entry: 1702 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4 1703 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4 1704 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4 1705 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 1706 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4 1707 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4 1708 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4 1709 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4 1710 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4 1711 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1712 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1713 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4 1714 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4 1715 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave() 1716 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4 1717 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4 1718 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4 1719 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4 1720 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4 1721 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4 1722 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4 1723 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64 1724 // CHECK11-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8* 1725 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false) 1726 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1727 // CHECK11-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32* 1728 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4 1729 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1730 // CHECK11-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32* 1731 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4 1732 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 1733 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4 1734 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 1735 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32* 1736 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4 1737 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 1738 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32* 1739 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4 1740 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 1741 // CHECK11-NEXT: store i8* null, i8** [[TMP16]], align 4 1742 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 1743 // CHECK11-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32** 1744 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4 1745 // CHECK11-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 1746 // CHECK11-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32** 1747 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4 1748 // CHECK11-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2 1749 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4 1750 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2 1751 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4 1752 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1753 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1754 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0 1755 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4 1756 // CHECK11-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4 1757 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1758 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0 1759 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1760 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1761 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1762 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1763 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1 1764 // CHECK11-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64 1765 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 1766 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 1767 // CHECK11-NEXT: store i32 1, i32* [[TMP30]], align 4 1768 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 1769 // CHECK11-NEXT: store i32 3, i32* [[TMP31]], align 4 1770 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 1771 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4 1772 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 1773 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4 1774 // CHECK11-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 1775 // CHECK11-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4 1776 // CHECK11-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 1777 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4 1778 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 1779 // CHECK11-NEXT: store i8** null, i8*** [[TMP36]], align 4 1780 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 1781 // CHECK11-NEXT: store i8** null, i8*** [[TMP37]], align 4 1782 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 1783 // CHECK11-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8 1784 // CHECK11-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 1785 // CHECK11-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0 1786 // CHECK11-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1787 // CHECK11: omp_offload.failed: 1788 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]] 1789 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]] 1790 // CHECK11: omp_offload.cont: 1791 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 0 1792 // CHECK11-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 1793 // CHECK11-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4 1794 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP42]]) 1795 // CHECK11-NEXT: ret i32 [[TMP41]] 1796 // 1797 // 1798 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73 1799 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] { 1800 // CHECK11-NEXT: entry: 1801 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 1802 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 1803 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 1804 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 1805 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 1806 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 1807 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 1808 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4 1809 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]]) 1810 // CHECK11-NEXT: ret void 1811 // 1812 // 1813 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined. 1814 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 1815 // CHECK11-NEXT: entry: 1816 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1817 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1818 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 1819 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 1820 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 1821 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1822 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4 1823 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1824 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1825 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4 1826 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1827 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1828 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1829 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1830 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4 1831 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1832 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1833 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 1834 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 1835 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 1836 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 1837 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 1838 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4 1839 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 1840 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 1841 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1842 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 1843 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1844 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1845 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1846 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4 1847 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1848 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 1849 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1850 // CHECK11: omp.precond.then: 1851 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1852 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1853 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4 1854 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1855 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1856 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1857 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 1858 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1859 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1860 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1861 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] 1862 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1863 // CHECK11: cond.true: 1864 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1865 // CHECK11-NEXT: br label [[COND_END:%.*]] 1866 // CHECK11: cond.false: 1867 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1868 // CHECK11-NEXT: br label [[COND_END]] 1869 // CHECK11: cond.end: 1870 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] 1871 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1872 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1873 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 1874 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1875 // CHECK11: omp.inner.for.cond: 1876 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1877 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1878 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] 1879 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1880 // CHECK11: omp.inner.for.body: 1881 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1882 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1883 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]) 1884 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1885 // CHECK11: omp.inner.for.inc: 1886 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1887 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1888 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 1889 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1890 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]] 1891 // CHECK11: omp.inner.for.end: 1892 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1893 // CHECK11: omp.loop.exit: 1894 // CHECK11-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1895 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4 1896 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]]) 1897 // CHECK11-NEXT: br label [[OMP_PRECOND_END]] 1898 // CHECK11: omp.precond.end: 1899 // CHECK11-NEXT: ret void 1900 // 1901 // 1902 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1 1903 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 1904 // CHECK11-NEXT: entry: 1905 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1906 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1907 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1908 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1909 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 1910 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 1911 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 1912 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1913 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4 1914 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 1915 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 1916 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4 1917 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1918 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1919 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1920 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1921 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4 1922 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1923 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1924 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1925 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1926 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 1927 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 1928 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 1929 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 1930 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 1931 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4 1932 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 1933 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 1934 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1935 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 1936 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 1937 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 1938 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 1939 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4 1940 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 1941 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 1942 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 1943 // CHECK11: omp.precond.then: 1944 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1945 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1946 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 1947 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1948 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1949 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4 1950 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4 1951 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1952 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1953 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1954 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4 1955 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1956 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1957 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1958 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]] 1959 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1960 // CHECK11: cond.true: 1961 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 1962 // CHECK11-NEXT: br label [[COND_END:%.*]] 1963 // CHECK11: cond.false: 1964 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1965 // CHECK11-NEXT: br label [[COND_END]] 1966 // CHECK11: cond.end: 1967 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ] 1968 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1969 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1970 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4 1971 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1972 // CHECK11: omp.inner.for.cond: 1973 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1974 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1975 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]] 1976 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1977 // CHECK11: omp.inner.for.body: 1978 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1979 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1 1980 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1981 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4 1982 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4 1983 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]] 1984 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 1985 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1986 // CHECK11: omp.body.continue: 1987 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1988 // CHECK11: omp.inner.for.inc: 1989 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1990 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1 1991 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4 1992 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]] 1993 // CHECK11: omp.inner.for.end: 1994 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1995 // CHECK11: omp.loop.exit: 1996 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1997 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 1998 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) 1999 // CHECK11-NEXT: br label [[OMP_PRECOND_END]] 2000 // CHECK11: omp.precond.end: 2001 // CHECK11-NEXT: ret void 2002 // 2003 // 2004 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 2005 // CHECK11-SAME: () #[[ATTR5:[0-9]+]] { 2006 // CHECK11-NEXT: entry: 2007 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1) 2008 // CHECK11-NEXT: ret void 2009 // 2010 // 2011 // CHECK17-LABEL: define {{[^@]+}}@_Z21teams_template_structv 2012 // CHECK17-SAME: () #[[ATTR0:[0-9]+]] { 2013 // CHECK17-NEXT: entry: 2014 // CHECK17-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4 2015 // CHECK17-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]]) 2016 // CHECK17-NEXT: ret i32 [[CALL]] 2017 // 2018 // 2019 // CHECK17-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv 2020 // CHECK17-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 { 2021 // CHECK17-NEXT: entry: 2022 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 2023 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8 2024 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8 2025 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8 2026 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4 2027 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 2028 // CHECK17-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 2029 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0 2030 // CHECK17-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2031 // CHECK17-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS** 2032 // CHECK17-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8 2033 // CHECK17-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2034 // CHECK17-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]** 2035 // CHECK17-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8 2036 // CHECK17-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 2037 // CHECK17-NEXT: store i8* null, i8** [[TMP4]], align 8 2038 // CHECK17-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2039 // CHECK17-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2040 // CHECK17-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 2041 // CHECK17-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 2042 // CHECK17-NEXT: store i32 1, i32* [[TMP7]], align 4 2043 // CHECK17-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 2044 // CHECK17-NEXT: store i32 1, i32* [[TMP8]], align 4 2045 // CHECK17-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 2046 // CHECK17-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8 2047 // CHECK17-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 2048 // CHECK17-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8 2049 // CHECK17-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 2050 // CHECK17-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8 2051 // CHECK17-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 2052 // CHECK17-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8 2053 // CHECK17-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 2054 // CHECK17-NEXT: store i8** null, i8*** [[TMP13]], align 8 2055 // CHECK17-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 2056 // CHECK17-NEXT: store i8** null, i8*** [[TMP14]], align 8 2057 // CHECK17-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 2058 // CHECK17-NEXT: store i64 123, i64* [[TMP15]], align 8 2059 // CHECK17-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 2060 // CHECK17-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0 2061 // CHECK17-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 2062 // CHECK17: omp_offload.failed: 2063 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]] 2064 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT]] 2065 // CHECK17: omp_offload.cont: 2066 // CHECK17-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0 2067 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i64 0, i64 0 2068 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 2069 // CHECK17-NEXT: ret i32 [[TMP18]] 2070 // 2071 // 2072 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109 2073 // CHECK17-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] { 2074 // CHECK17-NEXT: entry: 2075 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 2076 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 2077 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 2078 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]]) 2079 // CHECK17-NEXT: ret void 2080 // 2081 // 2082 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined. 2083 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] { 2084 // CHECK17-NEXT: entry: 2085 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2086 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2087 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 2088 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2089 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4 2090 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 2091 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 2092 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2093 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2094 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4 2095 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2096 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2097 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 2098 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 2099 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 2100 // CHECK17-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4 2101 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2102 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2103 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2104 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 2105 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2106 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2107 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122 2108 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2109 // CHECK17: cond.true: 2110 // CHECK17-NEXT: br label [[COND_END:%.*]] 2111 // CHECK17: cond.false: 2112 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2113 // CHECK17-NEXT: br label [[COND_END]] 2114 // CHECK17: cond.end: 2115 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 2116 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 2117 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2118 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 2119 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2120 // CHECK17: omp.inner.for.cond: 2121 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2122 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2123 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 2124 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2125 // CHECK17: omp.inner.for.body: 2126 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2127 // CHECK17-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64 2128 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2129 // CHECK17-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64 2130 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]) 2131 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2132 // CHECK17: omp.inner.for.inc: 2133 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2134 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 2135 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 2136 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 2137 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]] 2138 // CHECK17: omp.inner.for.end: 2139 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2140 // CHECK17: omp.loop.exit: 2141 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 2142 // CHECK17-NEXT: ret void 2143 // 2144 // 2145 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1 2146 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] { 2147 // CHECK17-NEXT: entry: 2148 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2149 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2150 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 2151 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 2152 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8 2153 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2154 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4 2155 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 2156 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 2157 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2158 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2159 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4 2160 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2161 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2162 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2163 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2164 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8 2165 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8 2166 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 2167 // CHECK17-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4 2168 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2169 // CHECK17-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32 2170 // CHECK17-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2171 // CHECK17-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32 2172 // CHECK17-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 2173 // CHECK17-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 2174 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2175 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2176 // CHECK17-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2177 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 2178 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2179 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2180 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122 2181 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2182 // CHECK17: cond.true: 2183 // CHECK17-NEXT: br label [[COND_END:%.*]] 2184 // CHECK17: cond.false: 2185 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2186 // CHECK17-NEXT: br label [[COND_END]] 2187 // CHECK17: cond.end: 2188 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 2189 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 2190 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 2191 // CHECK17-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 2192 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2193 // CHECK17: omp.inner.for.cond: 2194 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2195 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2196 // CHECK17-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 2197 // CHECK17-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2198 // CHECK17: omp.inner.for.body: 2199 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2200 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 2201 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 2202 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4 2203 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0 2204 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 2205 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64 2206 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]] 2207 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 2208 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 2209 // CHECK17: omp.body.continue: 2210 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2211 // CHECK17: omp.inner.for.inc: 2212 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2213 // CHECK17-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1 2214 // CHECK17-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 2215 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]] 2216 // CHECK17: omp.inner.for.end: 2217 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2218 // CHECK17: omp.loop.exit: 2219 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 2220 // CHECK17-NEXT: ret void 2221 // 2222 // 2223 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 2224 // CHECK17-SAME: () #[[ATTR3:[0-9]+]] { 2225 // CHECK17-NEXT: entry: 2226 // CHECK17-NEXT: call void @__tgt_register_requires(i64 1) 2227 // CHECK17-NEXT: ret void 2228 // 2229 // 2230 // CHECK19-LABEL: define {{[^@]+}}@_Z21teams_template_structv 2231 // CHECK19-SAME: () #[[ATTR0:[0-9]+]] { 2232 // CHECK19-NEXT: entry: 2233 // CHECK19-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4 2234 // CHECK19-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]]) 2235 // CHECK19-NEXT: ret i32 [[CALL]] 2236 // 2237 // 2238 // CHECK19-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv 2239 // CHECK19-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 { 2240 // CHECK19-NEXT: entry: 2241 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 2242 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4 2243 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4 2244 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4 2245 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4 2246 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 2247 // CHECK19-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 2248 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0 2249 // CHECK19-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2250 // CHECK19-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS** 2251 // CHECK19-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4 2252 // CHECK19-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2253 // CHECK19-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]** 2254 // CHECK19-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4 2255 // CHECK19-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 2256 // CHECK19-NEXT: store i8* null, i8** [[TMP4]], align 4 2257 // CHECK19-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2258 // CHECK19-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2259 // CHECK19-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 2260 // CHECK19-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 2261 // CHECK19-NEXT: store i32 1, i32* [[TMP7]], align 4 2262 // CHECK19-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 2263 // CHECK19-NEXT: store i32 1, i32* [[TMP8]], align 4 2264 // CHECK19-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 2265 // CHECK19-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4 2266 // CHECK19-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 2267 // CHECK19-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4 2268 // CHECK19-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 2269 // CHECK19-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4 2270 // CHECK19-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 2271 // CHECK19-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4 2272 // CHECK19-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 2273 // CHECK19-NEXT: store i8** null, i8*** [[TMP13]], align 4 2274 // CHECK19-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 2275 // CHECK19-NEXT: store i8** null, i8*** [[TMP14]], align 4 2276 // CHECK19-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 2277 // CHECK19-NEXT: store i64 123, i64* [[TMP15]], align 8 2278 // CHECK19-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 2279 // CHECK19-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0 2280 // CHECK19-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 2281 // CHECK19: omp_offload.failed: 2282 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]] 2283 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT]] 2284 // CHECK19: omp_offload.cont: 2285 // CHECK19-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0 2286 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i32 0, i32 0 2287 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 2288 // CHECK19-NEXT: ret i32 [[TMP18]] 2289 // 2290 // 2291 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109 2292 // CHECK19-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] { 2293 // CHECK19-NEXT: entry: 2294 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 2295 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 2296 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 2297 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]]) 2298 // CHECK19-NEXT: ret void 2299 // 2300 // 2301 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined. 2302 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] { 2303 // CHECK19-NEXT: entry: 2304 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 2305 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 2306 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 2307 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2308 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4 2309 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 2310 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 2311 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2312 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2313 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4 2314 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 2315 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 2316 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 2317 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 2318 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 2319 // CHECK19-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4 2320 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2321 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2322 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 2323 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 2324 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2325 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2326 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122 2327 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2328 // CHECK19: cond.true: 2329 // CHECK19-NEXT: br label [[COND_END:%.*]] 2330 // CHECK19: cond.false: 2331 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2332 // CHECK19-NEXT: br label [[COND_END]] 2333 // CHECK19: cond.end: 2334 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 2335 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 2336 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2337 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 2338 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2339 // CHECK19: omp.inner.for.cond: 2340 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2341 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2342 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 2343 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2344 // CHECK19: omp.inner.for.body: 2345 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2346 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2347 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]) 2348 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2349 // CHECK19: omp.inner.for.inc: 2350 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2351 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 2352 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]] 2353 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 2354 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]] 2355 // CHECK19: omp.inner.for.end: 2356 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2357 // CHECK19: omp.loop.exit: 2358 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 2359 // CHECK19-NEXT: ret void 2360 // 2361 // 2362 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1 2363 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] { 2364 // CHECK19-NEXT: entry: 2365 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 2366 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 2367 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 2368 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 2369 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4 2370 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2371 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4 2372 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 2373 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 2374 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2375 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2376 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4 2377 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 2378 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 2379 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 2380 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 2381 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4 2382 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4 2383 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 2384 // CHECK19-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4 2385 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 2386 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 2387 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4 2388 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4 2389 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2390 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2391 // CHECK19-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 2392 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 2393 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2394 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2395 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122 2396 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2397 // CHECK19: cond.true: 2398 // CHECK19-NEXT: br label [[COND_END:%.*]] 2399 // CHECK19: cond.false: 2400 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2401 // CHECK19-NEXT: br label [[COND_END]] 2402 // CHECK19: cond.end: 2403 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 2404 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 2405 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 2406 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 2407 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2408 // CHECK19: omp.inner.for.cond: 2409 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2410 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2411 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 2412 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2413 // CHECK19: omp.inner.for.body: 2414 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2415 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 2416 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 2417 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4 2418 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0 2419 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 2420 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]] 2421 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 2422 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 2423 // CHECK19: omp.body.continue: 2424 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2425 // CHECK19: omp.inner.for.inc: 2426 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2427 // CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1 2428 // CHECK19-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4 2429 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]] 2430 // CHECK19: omp.inner.for.end: 2431 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2432 // CHECK19: omp.loop.exit: 2433 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 2434 // CHECK19-NEXT: ret void 2435 // 2436 // 2437 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 2438 // CHECK19-SAME: () #[[ATTR3:[0-9]+]] { 2439 // CHECK19-NEXT: entry: 2440 // CHECK19-NEXT: call void @__tgt_register_requires(i64 1) 2441 // CHECK19-NEXT: ret void 2442 // 2443 // 2444 // CHECK25-LABEL: define {{[^@]+}}@main 2445 // CHECK25-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] { 2446 // CHECK25-NEXT: entry: 2447 // CHECK25-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 2448 // CHECK25-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4 2449 // CHECK25-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8 2450 // CHECK25-NEXT: [[N:%.*]] = alloca i32, align 4 2451 // CHECK25-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8 2452 // CHECK25-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8 2453 // CHECK25-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8 2454 // CHECK25-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8 2455 // CHECK25-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8 2456 // CHECK25-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8 2457 // CHECK25-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8 2458 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4 2459 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 2460 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 2461 // CHECK25-NEXT: store i32 0, i32* [[RETVAL]], align 4 2462 // CHECK25-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4 2463 // CHECK25-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8 2464 // CHECK25-NEXT: store i32 100, i32* [[N]], align 4 2465 // CHECK25-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4 2466 // CHECK25-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64 2467 // CHECK25-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave() 2468 // CHECK25-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8 2469 // CHECK25-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4 2470 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8 2471 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4 2472 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32* 2473 // CHECK25-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4 2474 // CHECK25-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8 2475 // CHECK25-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4 2476 // CHECK25-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8* 2477 // CHECK25-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false) 2478 // CHECK25-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2479 // CHECK25-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64* 2480 // CHECK25-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8 2481 // CHECK25-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2482 // CHECK25-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64* 2483 // CHECK25-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8 2484 // CHECK25-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 2485 // CHECK25-NEXT: store i8* null, i8** [[TMP11]], align 8 2486 // CHECK25-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 2487 // CHECK25-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64* 2488 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8 2489 // CHECK25-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 2490 // CHECK25-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64* 2491 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8 2492 // CHECK25-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 2493 // CHECK25-NEXT: store i8* null, i8** [[TMP16]], align 8 2494 // CHECK25-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 2495 // CHECK25-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32** 2496 // CHECK25-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8 2497 // CHECK25-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 2498 // CHECK25-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32** 2499 // CHECK25-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8 2500 // CHECK25-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2 2501 // CHECK25-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8 2502 // CHECK25-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2 2503 // CHECK25-NEXT: store i8* null, i8** [[TMP22]], align 8 2504 // CHECK25-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2505 // CHECK25-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2506 // CHECK25-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0 2507 // CHECK25-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4 2508 // CHECK25-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4 2509 // CHECK25-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 2510 // CHECK25-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0 2511 // CHECK25-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 2512 // CHECK25-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 2513 // CHECK25-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 2514 // CHECK25-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 2515 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1 2516 // CHECK25-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64 2517 // CHECK25-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 2518 // CHECK25-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 2519 // CHECK25-NEXT: store i32 1, i32* [[TMP30]], align 4 2520 // CHECK25-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 2521 // CHECK25-NEXT: store i32 3, i32* [[TMP31]], align 4 2522 // CHECK25-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 2523 // CHECK25-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8 2524 // CHECK25-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 2525 // CHECK25-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8 2526 // CHECK25-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 2527 // CHECK25-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8 2528 // CHECK25-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 2529 // CHECK25-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8 2530 // CHECK25-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 2531 // CHECK25-NEXT: store i8** null, i8*** [[TMP36]], align 8 2532 // CHECK25-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 2533 // CHECK25-NEXT: store i8** null, i8*** [[TMP37]], align 8 2534 // CHECK25-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 2535 // CHECK25-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8 2536 // CHECK25-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 2537 // CHECK25-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0 2538 // CHECK25-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 2539 // CHECK25: omp_offload.failed: 2540 // CHECK25-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]] 2541 // CHECK25-NEXT: br label [[OMP_OFFLOAD_CONT]] 2542 // CHECK25: omp_offload.cont: 2543 // CHECK25-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4 2544 // CHECK25-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP41]]) 2545 // CHECK25-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4 2546 // CHECK25-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8 2547 // CHECK25-NEXT: call void @llvm.stackrestore(i8* [[TMP42]]) 2548 // CHECK25-NEXT: [[TMP43:%.*]] = load i32, i32* [[RETVAL]], align 4 2549 // CHECK25-NEXT: ret i32 [[TMP43]] 2550 // 2551 // 2552 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162 2553 // CHECK25-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] { 2554 // CHECK25-NEXT: entry: 2555 // CHECK25-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8 2556 // CHECK25-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 2557 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 2558 // CHECK25-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8 2559 // CHECK25-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 2560 // CHECK25-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 2561 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32* 2562 // CHECK25-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 2563 // CHECK25-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8 2564 // CHECK25-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]]) 2565 // CHECK25-NEXT: ret void 2566 // 2567 // 2568 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined. 2569 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 2570 // CHECK25-NEXT: entry: 2571 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2572 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2573 // CHECK25-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 2574 // CHECK25-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 2575 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 2576 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2577 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4 2578 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 2579 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 2580 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4 2581 // CHECK25-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 2582 // CHECK25-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 2583 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2584 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2585 // CHECK25-NEXT: [[I3:%.*]] = alloca i32, align 4 2586 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2587 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2588 // CHECK25-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 2589 // CHECK25-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 2590 // CHECK25-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 2591 // CHECK25-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 2592 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 2593 // CHECK25-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8 2594 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 2595 // CHECK25-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 2596 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 2597 // CHECK25-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 2598 // CHECK25-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 2599 // CHECK25-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 2600 // CHECK25-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 2601 // CHECK25-NEXT: store i32 0, i32* [[I]], align 4 2602 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 2603 // CHECK25-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 2604 // CHECK25-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 2605 // CHECK25: omp.precond.then: 2606 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 2607 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 2608 // CHECK25-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4 2609 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2610 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2611 // CHECK25-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2612 // CHECK25-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 2613 // CHECK25-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2614 // CHECK25-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2615 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 2616 // CHECK25-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] 2617 // CHECK25-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2618 // CHECK25: cond.true: 2619 // CHECK25-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 2620 // CHECK25-NEXT: br label [[COND_END:%.*]] 2621 // CHECK25: cond.false: 2622 // CHECK25-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2623 // CHECK25-NEXT: br label [[COND_END]] 2624 // CHECK25: cond.end: 2625 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] 2626 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 2627 // CHECK25-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2628 // CHECK25-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 2629 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2630 // CHECK25: omp.inner.for.cond: 2631 // CHECK25-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2632 // CHECK25-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2633 // CHECK25-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] 2634 // CHECK25-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2635 // CHECK25: omp.inner.for.body: 2636 // CHECK25-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2637 // CHECK25-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64 2638 // CHECK25-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2639 // CHECK25-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64 2640 // CHECK25-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]) 2641 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2642 // CHECK25: omp.inner.for.inc: 2643 // CHECK25-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2644 // CHECK25-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 2645 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]] 2646 // CHECK25-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 2647 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]] 2648 // CHECK25: omp.inner.for.end: 2649 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2650 // CHECK25: omp.loop.exit: 2651 // CHECK25-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2652 // CHECK25-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4 2653 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]]) 2654 // CHECK25-NEXT: br label [[OMP_PRECOND_END]] 2655 // CHECK25: omp.precond.end: 2656 // CHECK25-NEXT: ret void 2657 // 2658 // 2659 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1 2660 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 2661 // CHECK25-NEXT: entry: 2662 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2663 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2664 // CHECK25-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 2665 // CHECK25-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 2666 // CHECK25-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8 2667 // CHECK25-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8 2668 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8 2669 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2670 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4 2671 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 2672 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 2673 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4 2674 // CHECK25-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 2675 // CHECK25-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 2676 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2677 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2678 // CHECK25-NEXT: [[I4:%.*]] = alloca i32, align 4 2679 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2680 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2681 // CHECK25-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2682 // CHECK25-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2683 // CHECK25-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8 2684 // CHECK25-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8 2685 // CHECK25-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8 2686 // CHECK25-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8 2687 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8 2688 // CHECK25-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8 2689 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 2690 // CHECK25-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 2691 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 2692 // CHECK25-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 2693 // CHECK25-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 2694 // CHECK25-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 2695 // CHECK25-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 2696 // CHECK25-NEXT: store i32 0, i32* [[I]], align 4 2697 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 2698 // CHECK25-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 2699 // CHECK25-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 2700 // CHECK25: omp.precond.then: 2701 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 2702 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 2703 // CHECK25-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 2704 // CHECK25-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2705 // CHECK25-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32 2706 // CHECK25-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2707 // CHECK25-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32 2708 // CHECK25-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 2709 // CHECK25-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4 2710 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2711 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2712 // CHECK25-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2713 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4 2714 // CHECK25-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2715 // CHECK25-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2716 // CHECK25-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 2717 // CHECK25-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]] 2718 // CHECK25-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2719 // CHECK25: cond.true: 2720 // CHECK25-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 2721 // CHECK25-NEXT: br label [[COND_END:%.*]] 2722 // CHECK25: cond.false: 2723 // CHECK25-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2724 // CHECK25-NEXT: br label [[COND_END]] 2725 // CHECK25: cond.end: 2726 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ] 2727 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 2728 // CHECK25-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 2729 // CHECK25-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4 2730 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2731 // CHECK25: omp.inner.for.cond: 2732 // CHECK25-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2733 // CHECK25-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2734 // CHECK25-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]] 2735 // CHECK25-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2736 // CHECK25: omp.inner.for.body: 2737 // CHECK25-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2738 // CHECK25-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1 2739 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 2740 // CHECK25-NEXT: store i32 [[ADD]], i32* [[I4]], align 4 2741 // CHECK25-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4 2742 // CHECK25-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64 2743 // CHECK25-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]] 2744 // CHECK25-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 2745 // CHECK25-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 2746 // CHECK25: omp.body.continue: 2747 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2748 // CHECK25: omp.inner.for.inc: 2749 // CHECK25-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2750 // CHECK25-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1 2751 // CHECK25-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4 2752 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]] 2753 // CHECK25: omp.inner.for.end: 2754 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2755 // CHECK25: omp.loop.exit: 2756 // CHECK25-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2757 // CHECK25-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 2758 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) 2759 // CHECK25-NEXT: br label [[OMP_PRECOND_END]] 2760 // CHECK25: omp.precond.end: 2761 // CHECK25-NEXT: ret void 2762 // 2763 // 2764 // CHECK25-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_ 2765 // CHECK25-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat { 2766 // CHECK25-NEXT: entry: 2767 // CHECK25-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4 2768 // CHECK25-NEXT: [[A:%.*]] = alloca [10 x i32], align 4 2769 // CHECK25-NEXT: [[TE:%.*]] = alloca i32, align 4 2770 // CHECK25-NEXT: [[TH:%.*]] = alloca i32, align 4 2771 // CHECK25-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8 2772 // CHECK25-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8 2773 // CHECK25-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8 2774 // CHECK25-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8 2775 // CHECK25-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8 2776 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4 2777 // CHECK25-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4 2778 // CHECK25-NEXT: store i32 0, i32* [[TE]], align 4 2779 // CHECK25-NEXT: store i32 128, i32* [[TH]], align 4 2780 // CHECK25-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4 2781 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32* 2782 // CHECK25-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4 2783 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[TE_CASTED]], align 8 2784 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4 2785 // CHECK25-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32* 2786 // CHECK25-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4 2787 // CHECK25-NEXT: [[TMP3:%.*]] = load i64, i64* [[TH_CASTED]], align 8 2788 // CHECK25-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2789 // CHECK25-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64* 2790 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8 2791 // CHECK25-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2792 // CHECK25-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64* 2793 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP7]], align 8 2794 // CHECK25-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 2795 // CHECK25-NEXT: store i8* null, i8** [[TMP8]], align 8 2796 // CHECK25-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 2797 // CHECK25-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64* 2798 // CHECK25-NEXT: store i64 [[TMP3]], i64* [[TMP10]], align 8 2799 // CHECK25-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 2800 // CHECK25-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64* 2801 // CHECK25-NEXT: store i64 [[TMP3]], i64* [[TMP12]], align 8 2802 // CHECK25-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 2803 // CHECK25-NEXT: store i8* null, i8** [[TMP13]], align 8 2804 // CHECK25-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 2805 // CHECK25-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]** 2806 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 8 2807 // CHECK25-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 2808 // CHECK25-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]** 2809 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 8 2810 // CHECK25-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2 2811 // CHECK25-NEXT: store i8* null, i8** [[TMP18]], align 8 2812 // CHECK25-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 2813 // CHECK25-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 2814 // CHECK25-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4 2815 // CHECK25-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 2816 // CHECK25-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 2817 // CHECK25-NEXT: store i32 1, i32* [[TMP22]], align 4 2818 // CHECK25-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 2819 // CHECK25-NEXT: store i32 3, i32* [[TMP23]], align 4 2820 // CHECK25-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 2821 // CHECK25-NEXT: store i8** [[TMP19]], i8*** [[TMP24]], align 8 2822 // CHECK25-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 2823 // CHECK25-NEXT: store i8** [[TMP20]], i8*** [[TMP25]], align 8 2824 // CHECK25-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 2825 // CHECK25-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP26]], align 8 2826 // CHECK25-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 2827 // CHECK25-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP27]], align 8 2828 // CHECK25-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 2829 // CHECK25-NEXT: store i8** null, i8*** [[TMP28]], align 8 2830 // CHECK25-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 2831 // CHECK25-NEXT: store i8** null, i8*** [[TMP29]], align 8 2832 // CHECK25-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 2833 // CHECK25-NEXT: store i64 10, i64* [[TMP30]], align 8 2834 // CHECK25-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 [[TMP21]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 2835 // CHECK25-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0 2836 // CHECK25-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 2837 // CHECK25: omp_offload.failed: 2838 // CHECK25-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]] 2839 // CHECK25-NEXT: br label [[OMP_OFFLOAD_CONT]] 2840 // CHECK25: omp_offload.cont: 2841 // CHECK25-NEXT: ret i32 0 2842 // 2843 // 2844 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151 2845 // CHECK25-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { 2846 // CHECK25-NEXT: entry: 2847 // CHECK25-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8 2848 // CHECK25-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8 2849 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8 2850 // CHECK25-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]]) 2851 // CHECK25-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8 2852 // CHECK25-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8 2853 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8 2854 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32* 2855 // CHECK25-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32* 2856 // CHECK25-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8 2857 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4 2858 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4 2859 // CHECK25-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) 2860 // CHECK25-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]]) 2861 // CHECK25-NEXT: ret void 2862 // 2863 // 2864 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..2 2865 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { 2866 // CHECK25-NEXT: entry: 2867 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2868 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2869 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8 2870 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2871 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4 2872 // CHECK25-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 2873 // CHECK25-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 2874 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2875 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2876 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4 2877 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2878 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2879 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8 2880 // CHECK25-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8 2881 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 2882 // CHECK25-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4 2883 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2884 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2885 // CHECK25-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2886 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 2887 // CHECK25-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2888 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2889 // CHECK25-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9 2890 // CHECK25-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2891 // CHECK25: cond.true: 2892 // CHECK25-NEXT: br label [[COND_END:%.*]] 2893 // CHECK25: cond.false: 2894 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2895 // CHECK25-NEXT: br label [[COND_END]] 2896 // CHECK25: cond.end: 2897 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 2898 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 2899 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2900 // CHECK25-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 2901 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2902 // CHECK25: omp.inner.for.cond: 2903 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2904 // CHECK25-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2905 // CHECK25-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 2906 // CHECK25-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2907 // CHECK25: omp.inner.for.body: 2908 // CHECK25-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2909 // CHECK25-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64 2910 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2911 // CHECK25-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64 2912 // CHECK25-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]) 2913 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2914 // CHECK25: omp.inner.for.inc: 2915 // CHECK25-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2916 // CHECK25-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 2917 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 2918 // CHECK25-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 2919 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]] 2920 // CHECK25: omp.inner.for.end: 2921 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2922 // CHECK25: omp.loop.exit: 2923 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 2924 // CHECK25-NEXT: ret void 2925 // 2926 // 2927 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..3 2928 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { 2929 // CHECK25-NEXT: entry: 2930 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2931 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2932 // CHECK25-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 2933 // CHECK25-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 2934 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8 2935 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2936 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4 2937 // CHECK25-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 2938 // CHECK25-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 2939 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2940 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2941 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4 2942 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2943 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2944 // CHECK25-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2945 // CHECK25-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2946 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8 2947 // CHECK25-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8 2948 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 2949 // CHECK25-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4 2950 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2951 // CHECK25-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32 2952 // CHECK25-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2953 // CHECK25-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32 2954 // CHECK25-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 2955 // CHECK25-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 2956 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2957 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2958 // CHECK25-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2959 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 2960 // CHECK25-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2961 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2962 // CHECK25-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9 2963 // CHECK25-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2964 // CHECK25: cond.true: 2965 // CHECK25-NEXT: br label [[COND_END:%.*]] 2966 // CHECK25: cond.false: 2967 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2968 // CHECK25-NEXT: br label [[COND_END]] 2969 // CHECK25: cond.end: 2970 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 2971 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 2972 // CHECK25-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 2973 // CHECK25-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 2974 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2975 // CHECK25: omp.inner.for.cond: 2976 // CHECK25-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2977 // CHECK25-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2978 // CHECK25-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 2979 // CHECK25-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2980 // CHECK25: omp.inner.for.body: 2981 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2982 // CHECK25-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 2983 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 2984 // CHECK25-NEXT: store i32 [[ADD]], i32* [[I]], align 4 2985 // CHECK25-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 2986 // CHECK25-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64 2987 // CHECK25-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]] 2988 // CHECK25-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 2989 // CHECK25-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 2990 // CHECK25: omp.body.continue: 2991 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2992 // CHECK25: omp.inner.for.inc: 2993 // CHECK25-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2994 // CHECK25-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1 2995 // CHECK25-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 2996 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]] 2997 // CHECK25: omp.inner.for.end: 2998 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2999 // CHECK25: omp.loop.exit: 3000 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 3001 // CHECK25-NEXT: ret void 3002 // 3003 // 3004 // CHECK25-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 3005 // CHECK25-SAME: () #[[ATTR6:[0-9]+]] { 3006 // CHECK25-NEXT: entry: 3007 // CHECK25-NEXT: call void @__tgt_register_requires(i64 1) 3008 // CHECK25-NEXT: ret void 3009 // 3010 // 3011 // CHECK27-LABEL: define {{[^@]+}}@main 3012 // CHECK27-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] { 3013 // CHECK27-NEXT: entry: 3014 // CHECK27-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 3015 // CHECK27-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4 3016 // CHECK27-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4 3017 // CHECK27-NEXT: [[N:%.*]] = alloca i32, align 4 3018 // CHECK27-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4 3019 // CHECK27-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4 3020 // CHECK27-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4 3021 // CHECK27-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4 3022 // CHECK27-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4 3023 // CHECK27-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4 3024 // CHECK27-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4 3025 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4 3026 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 3027 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 3028 // CHECK27-NEXT: store i32 0, i32* [[RETVAL]], align 4 3029 // CHECK27-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4 3030 // CHECK27-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4 3031 // CHECK27-NEXT: store i32 100, i32* [[N]], align 4 3032 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4 3033 // CHECK27-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave() 3034 // CHECK27-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4 3035 // CHECK27-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4 3036 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4 3037 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4 3038 // CHECK27-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4 3039 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4 3040 // CHECK27-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4 3041 // CHECK27-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64 3042 // CHECK27-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8* 3043 // CHECK27-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false) 3044 // CHECK27-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 3045 // CHECK27-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32* 3046 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4 3047 // CHECK27-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 3048 // CHECK27-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32* 3049 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4 3050 // CHECK27-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 3051 // CHECK27-NEXT: store i8* null, i8** [[TMP11]], align 4 3052 // CHECK27-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 3053 // CHECK27-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32* 3054 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4 3055 // CHECK27-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 3056 // CHECK27-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32* 3057 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4 3058 // CHECK27-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 3059 // CHECK27-NEXT: store i8* null, i8** [[TMP16]], align 4 3060 // CHECK27-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 3061 // CHECK27-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32** 3062 // CHECK27-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4 3063 // CHECK27-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 3064 // CHECK27-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32** 3065 // CHECK27-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4 3066 // CHECK27-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2 3067 // CHECK27-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4 3068 // CHECK27-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2 3069 // CHECK27-NEXT: store i8* null, i8** [[TMP22]], align 4 3070 // CHECK27-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 3071 // CHECK27-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 3072 // CHECK27-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0 3073 // CHECK27-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4 3074 // CHECK27-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4 3075 // CHECK27-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 3076 // CHECK27-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0 3077 // CHECK27-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 3078 // CHECK27-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 3079 // CHECK27-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 3080 // CHECK27-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 3081 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1 3082 // CHECK27-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64 3083 // CHECK27-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 3084 // CHECK27-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 3085 // CHECK27-NEXT: store i32 1, i32* [[TMP30]], align 4 3086 // CHECK27-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 3087 // CHECK27-NEXT: store i32 3, i32* [[TMP31]], align 4 3088 // CHECK27-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 3089 // CHECK27-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4 3090 // CHECK27-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 3091 // CHECK27-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4 3092 // CHECK27-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 3093 // CHECK27-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4 3094 // CHECK27-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 3095 // CHECK27-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4 3096 // CHECK27-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 3097 // CHECK27-NEXT: store i8** null, i8*** [[TMP36]], align 4 3098 // CHECK27-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 3099 // CHECK27-NEXT: store i8** null, i8*** [[TMP37]], align 4 3100 // CHECK27-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 3101 // CHECK27-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8 3102 // CHECK27-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 3103 // CHECK27-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0 3104 // CHECK27-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 3105 // CHECK27: omp_offload.failed: 3106 // CHECK27-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]] 3107 // CHECK27-NEXT: br label [[OMP_OFFLOAD_CONT]] 3108 // CHECK27: omp_offload.cont: 3109 // CHECK27-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4 3110 // CHECK27-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP41]]) 3111 // CHECK27-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4 3112 // CHECK27-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4 3113 // CHECK27-NEXT: call void @llvm.stackrestore(i8* [[TMP42]]) 3114 // CHECK27-NEXT: [[TMP43:%.*]] = load i32, i32* [[RETVAL]], align 4 3115 // CHECK27-NEXT: ret i32 [[TMP43]] 3116 // 3117 // 3118 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162 3119 // CHECK27-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] { 3120 // CHECK27-NEXT: entry: 3121 // CHECK27-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4 3122 // CHECK27-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 3123 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 3124 // CHECK27-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4 3125 // CHECK27-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 3126 // CHECK27-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 3127 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 3128 // CHECK27-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4 3129 // CHECK27-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]]) 3130 // CHECK27-NEXT: ret void 3131 // 3132 // 3133 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined. 3134 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 3135 // CHECK27-NEXT: entry: 3136 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 3137 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 3138 // CHECK27-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 3139 // CHECK27-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 3140 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 3141 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 3142 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4 3143 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 3144 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 3145 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4 3146 // CHECK27-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 3147 // CHECK27-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 3148 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 3149 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 3150 // CHECK27-NEXT: [[I3:%.*]] = alloca i32, align 4 3151 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 3152 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 3153 // CHECK27-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 3154 // CHECK27-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 3155 // CHECK27-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 3156 // CHECK27-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 3157 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 3158 // CHECK27-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4 3159 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 3160 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 3161 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 3162 // CHECK27-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 3163 // CHECK27-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 3164 // CHECK27-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 3165 // CHECK27-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 3166 // CHECK27-NEXT: store i32 0, i32* [[I]], align 4 3167 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 3168 // CHECK27-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 3169 // CHECK27-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 3170 // CHECK27: omp.precond.then: 3171 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 3172 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 3173 // CHECK27-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4 3174 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 3175 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 3176 // CHECK27-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 3177 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 3178 // CHECK27-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 3179 // CHECK27-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3180 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 3181 // CHECK27-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]] 3182 // CHECK27-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 3183 // CHECK27: cond.true: 3184 // CHECK27-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 3185 // CHECK27-NEXT: br label [[COND_END:%.*]] 3186 // CHECK27: cond.false: 3187 // CHECK27-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3188 // CHECK27-NEXT: br label [[COND_END]] 3189 // CHECK27: cond.end: 3190 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ] 3191 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 3192 // CHECK27-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 3193 // CHECK27-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4 3194 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 3195 // CHECK27: omp.inner.for.cond: 3196 // CHECK27-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3197 // CHECK27-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3198 // CHECK27-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]] 3199 // CHECK27-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 3200 // CHECK27: omp.inner.for.body: 3201 // CHECK27-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 3202 // CHECK27-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3203 // CHECK27-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]) 3204 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 3205 // CHECK27: omp.inner.for.inc: 3206 // CHECK27-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3207 // CHECK27-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 3208 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 3209 // CHECK27-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 3210 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]] 3211 // CHECK27: omp.inner.for.end: 3212 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 3213 // CHECK27: omp.loop.exit: 3214 // CHECK27-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 3215 // CHECK27-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4 3216 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]]) 3217 // CHECK27-NEXT: br label [[OMP_PRECOND_END]] 3218 // CHECK27: omp.precond.end: 3219 // CHECK27-NEXT: ret void 3220 // 3221 // 3222 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1 3223 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] { 3224 // CHECK27-NEXT: entry: 3225 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 3226 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 3227 // CHECK27-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 3228 // CHECK27-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 3229 // CHECK27-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4 3230 // CHECK27-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4 3231 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4 3232 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 3233 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4 3234 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4 3235 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4 3236 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4 3237 // CHECK27-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 3238 // CHECK27-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 3239 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 3240 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 3241 // CHECK27-NEXT: [[I3:%.*]] = alloca i32, align 4 3242 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 3243 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 3244 // CHECK27-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 3245 // CHECK27-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 3246 // CHECK27-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4 3247 // CHECK27-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4 3248 // CHECK27-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4 3249 // CHECK27-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4 3250 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4 3251 // CHECK27-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4 3252 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4 3253 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4 3254 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 3255 // CHECK27-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0 3256 // CHECK27-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1 3257 // CHECK27-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1 3258 // CHECK27-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4 3259 // CHECK27-NEXT: store i32 0, i32* [[I]], align 4 3260 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4 3261 // CHECK27-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]] 3262 // CHECK27-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]] 3263 // CHECK27: omp.precond.then: 3264 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 3265 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 3266 // CHECK27-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 3267 // CHECK27-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 3268 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 3269 // CHECK27-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4 3270 // CHECK27-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4 3271 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 3272 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 3273 // CHECK27-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 3274 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4 3275 // CHECK27-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 3276 // CHECK27-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 3277 // CHECK27-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 3278 // CHECK27-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]] 3279 // CHECK27-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 3280 // CHECK27: cond.true: 3281 // CHECK27-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4 3282 // CHECK27-NEXT: br label [[COND_END:%.*]] 3283 // CHECK27: cond.false: 3284 // CHECK27-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 3285 // CHECK27-NEXT: br label [[COND_END]] 3286 // CHECK27: cond.end: 3287 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ] 3288 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 3289 // CHECK27-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 3290 // CHECK27-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4 3291 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 3292 // CHECK27: omp.inner.for.cond: 3293 // CHECK27-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3294 // CHECK27-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 3295 // CHECK27-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]] 3296 // CHECK27-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 3297 // CHECK27: omp.inner.for.body: 3298 // CHECK27-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3299 // CHECK27-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1 3300 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 3301 // CHECK27-NEXT: store i32 [[ADD]], i32* [[I3]], align 4 3302 // CHECK27-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4 3303 // CHECK27-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]] 3304 // CHECK27-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 3305 // CHECK27-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 3306 // CHECK27: omp.body.continue: 3307 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 3308 // CHECK27: omp.inner.for.inc: 3309 // CHECK27-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3310 // CHECK27-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1 3311 // CHECK27-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4 3312 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]] 3313 // CHECK27: omp.inner.for.end: 3314 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 3315 // CHECK27: omp.loop.exit: 3316 // CHECK27-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 3317 // CHECK27-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4 3318 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]]) 3319 // CHECK27-NEXT: br label [[OMP_PRECOND_END]] 3320 // CHECK27: omp.precond.end: 3321 // CHECK27-NEXT: ret void 3322 // 3323 // 3324 // CHECK27-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_ 3325 // CHECK27-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat { 3326 // CHECK27-NEXT: entry: 3327 // CHECK27-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4 3328 // CHECK27-NEXT: [[A:%.*]] = alloca [10 x i32], align 4 3329 // CHECK27-NEXT: [[TE:%.*]] = alloca i32, align 4 3330 // CHECK27-NEXT: [[TH:%.*]] = alloca i32, align 4 3331 // CHECK27-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4 3332 // CHECK27-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4 3333 // CHECK27-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4 3334 // CHECK27-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4 3335 // CHECK27-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4 3336 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4 3337 // CHECK27-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4 3338 // CHECK27-NEXT: store i32 0, i32* [[TE]], align 4 3339 // CHECK27-NEXT: store i32 128, i32* [[TH]], align 4 3340 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4 3341 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TE_CASTED]], align 4 3342 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE_CASTED]], align 4 3343 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4 3344 // CHECK27-NEXT: store i32 [[TMP2]], i32* [[TH_CASTED]], align 4 3345 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_CASTED]], align 4 3346 // CHECK27-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 3347 // CHECK27-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32* 3348 // CHECK27-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4 3349 // CHECK27-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 3350 // CHECK27-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32* 3351 // CHECK27-NEXT: store i32 [[TMP1]], i32* [[TMP7]], align 4 3352 // CHECK27-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 3353 // CHECK27-NEXT: store i8* null, i8** [[TMP8]], align 4 3354 // CHECK27-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 3355 // CHECK27-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32* 3356 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4 3357 // CHECK27-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 3358 // CHECK27-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32* 3359 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4 3360 // CHECK27-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 3361 // CHECK27-NEXT: store i8* null, i8** [[TMP13]], align 4 3362 // CHECK27-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2 3363 // CHECK27-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]** 3364 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 4 3365 // CHECK27-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2 3366 // CHECK27-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]** 3367 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 4 3368 // CHECK27-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2 3369 // CHECK27-NEXT: store i8* null, i8** [[TMP18]], align 4 3370 // CHECK27-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 3371 // CHECK27-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 3372 // CHECK27-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4 3373 // CHECK27-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8 3374 // CHECK27-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0 3375 // CHECK27-NEXT: store i32 1, i32* [[TMP22]], align 4 3376 // CHECK27-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1 3377 // CHECK27-NEXT: store i32 3, i32* [[TMP23]], align 4 3378 // CHECK27-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2 3379 // CHECK27-NEXT: store i8** [[TMP19]], i8*** [[TMP24]], align 4 3380 // CHECK27-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3 3381 // CHECK27-NEXT: store i8** [[TMP20]], i8*** [[TMP25]], align 4 3382 // CHECK27-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4 3383 // CHECK27-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP26]], align 4 3384 // CHECK27-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5 3385 // CHECK27-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP27]], align 4 3386 // CHECK27-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6 3387 // CHECK27-NEXT: store i8** null, i8*** [[TMP28]], align 4 3388 // CHECK27-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7 3389 // CHECK27-NEXT: store i8** null, i8*** [[TMP29]], align 4 3390 // CHECK27-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8 3391 // CHECK27-NEXT: store i64 10, i64* [[TMP30]], align 8 3392 // CHECK27-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 [[TMP21]], i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]]) 3393 // CHECK27-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0 3394 // CHECK27-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 3395 // CHECK27: omp_offload.failed: 3396 // CHECK27-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]] 3397 // CHECK27-NEXT: br label [[OMP_OFFLOAD_CONT]] 3398 // CHECK27: omp_offload.cont: 3399 // CHECK27-NEXT: ret i32 0 3400 // 3401 // 3402 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151 3403 // CHECK27-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { 3404 // CHECK27-NEXT: entry: 3405 // CHECK27-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4 3406 // CHECK27-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4 3407 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4 3408 // CHECK27-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB3]]) 3409 // CHECK27-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4 3410 // CHECK27-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4 3411 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4 3412 // CHECK27-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4 3413 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4 3414 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4 3415 // CHECK27-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB3]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]]) 3416 // CHECK27-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]]) 3417 // CHECK27-NEXT: ret void 3418 // 3419 // 3420 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..2 3421 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { 3422 // CHECK27-NEXT: entry: 3423 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 3424 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 3425 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4 3426 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 3427 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4 3428 // CHECK27-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 3429 // CHECK27-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 3430 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 3431 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 3432 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4 3433 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 3434 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 3435 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4 3436 // CHECK27-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4 3437 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 3438 // CHECK27-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4 3439 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 3440 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 3441 // CHECK27-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 3442 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 3443 // CHECK27-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 3444 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3445 // CHECK27-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9 3446 // CHECK27-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 3447 // CHECK27: cond.true: 3448 // CHECK27-NEXT: br label [[COND_END:%.*]] 3449 // CHECK27: cond.false: 3450 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3451 // CHECK27-NEXT: br label [[COND_END]] 3452 // CHECK27: cond.end: 3453 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 3454 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 3455 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 3456 // CHECK27-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 3457 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 3458 // CHECK27: omp.inner.for.cond: 3459 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3460 // CHECK27-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3461 // CHECK27-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 3462 // CHECK27-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 3463 // CHECK27: omp.inner.for.body: 3464 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 3465 // CHECK27-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 3466 // CHECK27-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]) 3467 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 3468 // CHECK27: omp.inner.for.inc: 3469 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3470 // CHECK27-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 3471 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]] 3472 // CHECK27-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 3473 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]] 3474 // CHECK27: omp.inner.for.end: 3475 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 3476 // CHECK27: omp.loop.exit: 3477 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 3478 // CHECK27-NEXT: ret void 3479 // 3480 // 3481 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..3 3482 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] { 3483 // CHECK27-NEXT: entry: 3484 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 3485 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 3486 // CHECK27-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 3487 // CHECK27-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 3488 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4 3489 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 3490 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4 3491 // CHECK27-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 3492 // CHECK27-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 3493 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 3494 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 3495 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4 3496 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 3497 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 3498 // CHECK27-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 3499 // CHECK27-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 3500 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4 3501 // CHECK27-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4 3502 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 3503 // CHECK27-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4 3504 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 3505 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 3506 // CHECK27-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4 3507 // CHECK27-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4 3508 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 3509 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 3510 // CHECK27-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 3511 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 3512 // CHECK27-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 3513 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 3514 // CHECK27-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9 3515 // CHECK27-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 3516 // CHECK27: cond.true: 3517 // CHECK27-NEXT: br label [[COND_END:%.*]] 3518 // CHECK27: cond.false: 3519 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 3520 // CHECK27-NEXT: br label [[COND_END]] 3521 // CHECK27: cond.end: 3522 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 3523 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 3524 // CHECK27-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 3525 // CHECK27-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 3526 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 3527 // CHECK27: omp.inner.for.cond: 3528 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3529 // CHECK27-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 3530 // CHECK27-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 3531 // CHECK27-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 3532 // CHECK27: omp.inner.for.body: 3533 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3534 // CHECK27-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 3535 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 3536 // CHECK27-NEXT: store i32 [[ADD]], i32* [[I]], align 4 3537 // CHECK27-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 3538 // CHECK27-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]] 3539 // CHECK27-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4 3540 // CHECK27-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 3541 // CHECK27: omp.body.continue: 3542 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 3543 // CHECK27: omp.inner.for.inc: 3544 // CHECK27-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 3545 // CHECK27-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1 3546 // CHECK27-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4 3547 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]] 3548 // CHECK27: omp.inner.for.end: 3549 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 3550 // CHECK27: omp.loop.exit: 3551 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 3552 // CHECK27-NEXT: ret void 3553 // 3554 // 3555 // CHECK27-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 3556 // CHECK27-SAME: () #[[ATTR6:[0-9]+]] { 3557 // CHECK27-NEXT: entry: 3558 // CHECK27-NEXT: call void @__tgt_register_requires(i64 1) 3559 // CHECK27-NEXT: ret void 3560 // 3561