1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _ 2 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1 3 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 4 // RUN: %clang_cc1 -DCHECK -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=CHECK2 5 // RUN: %clang_cc1 -DCHECK -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3 6 // RUN: %clang_cc1 -DCHECK -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 7 // RUN: %clang_cc1 -DCHECK -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=CHECK4 8 9 // RUN: %clang_cc1 -DCHECK -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}}" 10 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 11 // RUN: %clang_cc1 -DCHECK -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}}" 12 // RUN: %clang_cc1 -DCHECK -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}}" 13 // RUN: %clang_cc1 -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s 14 // RUN: %clang_cc1 -DCHECK -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}}" 15 16 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9 17 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 18 // RUN: %clang_cc1 -DLAMBDA -fopenmp -x c++ -std=c++11 -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=CHECK10 19 20 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}" 21 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s 22 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -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}}" 23 24 // expected-no-diagnostics 25 #ifndef HEADER 26 #define HEADER 27 28 int x; 29 #pragma omp threadprivate(x) 30 31 template <typename T> 32 T tmain() { 33 int a[2]; 34 #pragma omp target 35 #pragma omp teams distribute parallel for copyin(x) 36 for (int i = 0; i < 2; ++i) { 37 a[i] = x; 38 } 39 return T(); 40 } 41 42 int main() { 43 int a[2]; 44 #ifdef LAMBDA 45 [&]() { 46 #pragma omp target 47 #pragma omp teams distribute parallel for copyin(x) 48 for (int i = 0; i < 2; ++i) { 49 50 // Skip global, bound tid and loop vars 51 a[i] = x; 52 53 54 // Skip global, bound tid and loop vars 55 56 57 [&]() { 58 a[i] = x; 59 }(); 60 } 61 }(); 62 return 0; 63 #else 64 #pragma omp target 65 #pragma omp teams distribute parallel for copyin(x) 66 for (int i = 0; i < 2; ++i) { 67 a[i] = x; 68 } 69 return tmain<int>(); 70 //return 0; 71 #endif 72 } 73 74 75 76 // Skip global, bound tid and loop vars 77 78 79 // Skip global, bound tid and loop vars 80 81 82 83 84 // Skip global, bound tid and loop vars 85 86 87 // Skip global, bound tid and loop vars 88 // prev lb and ub 89 90 // iter variables 91 92 93 94 #endif 95 // CHECK1-LABEL: define {{[^@]+}}@main 96 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] { 97 // CHECK1-NEXT: entry: 98 // CHECK1-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 99 // CHECK1-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 100 // CHECK1-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8 101 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8 102 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8 103 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8 104 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 105 // CHECK1-NEXT: store i32 0, i32* [[RETVAL]], align 4 106 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 107 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32* 108 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4 109 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[X_CASTED]], align 8 110 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 111 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64* 112 // CHECK1-NEXT: store i64 [[TMP1]], i64* [[TMP3]], align 8 113 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 114 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64* 115 // CHECK1-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8 116 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 117 // CHECK1-NEXT: store i8* null, i8** [[TMP6]], align 8 118 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 119 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 120 // CHECK1-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 8 121 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 122 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 123 // CHECK1-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 8 124 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 125 // CHECK1-NEXT: store i8* null, i8** [[TMP11]], align 8 126 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 127 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 128 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 129 // CHECK1-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 130 // CHECK1-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 131 // CHECK1-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 132 // CHECK1: omp_offload.failed: 133 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i64 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2:[0-9]+]] 134 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] 135 // CHECK1: omp_offload.cont: 136 // CHECK1-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v() 137 // CHECK1-NEXT: ret i32 [[CALL]] 138 // 139 // 140 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64 141 // CHECK1-SAME: (i64 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1:[0-9]+]] { 142 // CHECK1-NEXT: entry: 143 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8 144 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8 145 // CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 146 // CHECK1-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8 147 // CHECK1-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8 148 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32* 149 // CHECK1-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8 150 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 151 // CHECK1-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 8 152 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 153 // CHECK1-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 154 // CHECK1-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.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) 155 // CHECK1-NEXT: ret void 156 // 157 // 158 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined. 159 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 160 // CHECK1-NEXT: entry: 161 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 162 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 163 // CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 164 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 165 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 166 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 167 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 168 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 169 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 170 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 171 // CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 172 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 173 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 174 // CHECK1-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 175 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 176 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 177 // CHECK1-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 178 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 179 // CHECK1-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 180 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 181 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 182 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 183 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 184 // CHECK1-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 185 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 186 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 187 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 188 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 189 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 190 // CHECK1: cond.true: 191 // CHECK1-NEXT: br label [[COND_END:%.*]] 192 // CHECK1: cond.false: 193 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 194 // CHECK1-NEXT: br label [[COND_END]] 195 // CHECK1: cond.end: 196 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 197 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 198 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 199 // CHECK1-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 200 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 201 // CHECK1: omp.inner.for.cond: 202 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 203 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 204 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 205 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 206 // CHECK1: omp.inner.for.body: 207 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 208 // CHECK1-NEXT: [[TMP13:%.*]] = zext i32 [[TMP12]] to i64 209 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 210 // CHECK1-NEXT: [[TMP15:%.*]] = zext i32 [[TMP14]] to i64 211 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 212 // CHECK1-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP16]], align 8 213 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 214 // CHECK1-NEXT: store i32* [[TMP4]], i32** [[TMP17]], align 8 215 // CHECK1-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.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP13]], i64 [[TMP15]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 216 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 217 // CHECK1: omp.inner.for.inc: 218 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 219 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 220 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 221 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 222 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 223 // CHECK1: omp.inner.for.end: 224 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 225 // CHECK1: omp.loop.exit: 226 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 227 // CHECK1-NEXT: ret void 228 // 229 // 230 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1 231 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 232 // CHECK1-NEXT: entry: 233 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 234 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 235 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 236 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 237 // CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 238 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 239 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 240 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 241 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 242 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 243 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 244 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 245 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 246 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 247 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 248 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 249 // CHECK1-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 250 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 251 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 252 // CHECK1-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 253 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 254 // CHECK1-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 255 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 256 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 257 // CHECK1-NEXT: [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 258 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP5]] to i32 259 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 260 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP6]] to i32 261 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 262 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 263 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 264 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 265 // CHECK1-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 266 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 267 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 268 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 269 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 270 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 271 // CHECK1: cond.true: 272 // CHECK1-NEXT: br label [[COND_END:%.*]] 273 // CHECK1: cond.false: 274 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 275 // CHECK1-NEXT: br label [[COND_END]] 276 // CHECK1: cond.end: 277 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 278 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 279 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 280 // CHECK1-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 281 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 282 // CHECK1: omp.inner.for.cond: 283 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 284 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 285 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 286 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 287 // CHECK1: omp.inner.for.body: 288 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 289 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 290 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 291 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4 292 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 293 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 294 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64 295 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] 296 // CHECK1-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 297 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 298 // CHECK1: omp.body.continue: 299 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 300 // CHECK1: omp.inner.for.inc: 301 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 302 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1 303 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 304 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 305 // CHECK1: omp.inner.for.end: 306 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 307 // CHECK1: omp.loop.exit: 308 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 309 // CHECK1-NEXT: ret void 310 // 311 // 312 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 313 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat { 314 // CHECK1-NEXT: entry: 315 // CHECK1-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 316 // CHECK1-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8 317 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8 318 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8 319 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8 320 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 321 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 322 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32* 323 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4 324 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[X_CASTED]], align 8 325 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 326 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64* 327 // CHECK1-NEXT: store i64 [[TMP1]], i64* [[TMP3]], align 8 328 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 329 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64* 330 // CHECK1-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8 331 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 332 // CHECK1-NEXT: store i8* null, i8** [[TMP6]], align 8 333 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 334 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 335 // CHECK1-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 8 336 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 337 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 338 // CHECK1-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 8 339 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 340 // CHECK1-NEXT: store i8* null, i8** [[TMP11]], align 8 341 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 342 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 343 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 344 // CHECK1-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 345 // CHECK1-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 346 // CHECK1-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 347 // CHECK1: omp_offload.failed: 348 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34(i64 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2]] 349 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] 350 // CHECK1: omp_offload.cont: 351 // CHECK1-NEXT: ret i32 0 352 // 353 // 354 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34 355 // CHECK1-SAME: (i64 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] { 356 // CHECK1-NEXT: entry: 357 // CHECK1-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8 358 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8 359 // CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8 360 // CHECK1-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8 361 // CHECK1-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8 362 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32* 363 // CHECK1-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8 364 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 365 // CHECK1-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 8 366 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 367 // CHECK1-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 368 // CHECK1-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.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]]) 369 // CHECK1-NEXT: ret void 370 // 371 // 372 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2 373 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 374 // CHECK1-NEXT: entry: 375 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 376 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 377 // CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8 378 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 379 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 380 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 381 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 382 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 383 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 384 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 385 // CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8 386 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 387 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 388 // CHECK1-NEXT: store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8 389 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8 390 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0 391 // CHECK1-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 392 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1 393 // CHECK1-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 394 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 395 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 396 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 397 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 398 // CHECK1-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 399 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 400 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 401 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 402 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 403 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 404 // CHECK1: cond.true: 405 // CHECK1-NEXT: br label [[COND_END:%.*]] 406 // CHECK1: cond.false: 407 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 408 // CHECK1-NEXT: br label [[COND_END]] 409 // CHECK1: cond.end: 410 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 411 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 412 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 413 // CHECK1-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 414 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 415 // CHECK1: omp.inner.for.cond: 416 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 417 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 418 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 419 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 420 // CHECK1: omp.inner.for.body: 421 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 422 // CHECK1-NEXT: [[TMP13:%.*]] = zext i32 [[TMP12]] to i64 423 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 424 // CHECK1-NEXT: [[TMP15:%.*]] = zext i32 [[TMP14]] to i64 425 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 426 // CHECK1-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP16]], align 8 427 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 428 // CHECK1-NEXT: store i32* [[TMP4]], i32** [[TMP17]], align 8 429 // CHECK1-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.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP13]], i64 [[TMP15]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]]) 430 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 431 // CHECK1: omp.inner.for.inc: 432 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 433 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 434 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 435 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 436 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 437 // CHECK1: omp.inner.for.end: 438 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 439 // CHECK1: omp.loop.exit: 440 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 441 // CHECK1-NEXT: ret void 442 // 443 // 444 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3 445 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 446 // CHECK1-NEXT: entry: 447 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 448 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 449 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 450 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 451 // CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8 452 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 453 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 454 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 455 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 456 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 457 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 458 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 459 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 460 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 461 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 462 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 463 // CHECK1-NEXT: store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8 464 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8 465 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0 466 // CHECK1-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 467 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1 468 // CHECK1-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 469 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 470 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 471 // CHECK1-NEXT: [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 472 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP5]] to i32 473 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 474 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP6]] to i32 475 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 476 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 477 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 478 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 479 // CHECK1-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 480 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 481 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 482 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 483 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 484 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 485 // CHECK1: cond.true: 486 // CHECK1-NEXT: br label [[COND_END:%.*]] 487 // CHECK1: cond.false: 488 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 489 // CHECK1-NEXT: br label [[COND_END]] 490 // CHECK1: cond.end: 491 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 492 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 493 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 494 // CHECK1-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 495 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 496 // CHECK1: omp.inner.for.cond: 497 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 498 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 499 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 500 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 501 // CHECK1: omp.inner.for.body: 502 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 503 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 504 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 505 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4 506 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 507 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 508 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64 509 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] 510 // CHECK1-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 511 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 512 // CHECK1: omp.body.continue: 513 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 514 // CHECK1: omp.inner.for.inc: 515 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 516 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1 517 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 518 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 519 // CHECK1: omp.inner.for.end: 520 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 521 // CHECK1: omp.loop.exit: 522 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 523 // CHECK1-NEXT: ret void 524 // 525 // 526 // CHECK1-LABEL: define {{[^@]+}}@_ZTW1x 527 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] comdat { 528 // CHECK1-NEXT: ret i32* @x 529 // 530 // 531 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 532 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] { 533 // CHECK1-NEXT: entry: 534 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1) 535 // CHECK1-NEXT: ret void 536 // 537 // 538 // CHECK2-LABEL: define {{[^@]+}}@main 539 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] { 540 // CHECK2-NEXT: entry: 541 // CHECK2-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 542 // CHECK2-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 543 // CHECK2-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8 544 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8 545 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8 546 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8 547 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 548 // CHECK2-NEXT: store i32 0, i32* [[RETVAL]], align 4 549 // CHECK2-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 550 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32* 551 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4 552 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[X_CASTED]], align 8 553 // CHECK2-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 554 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64* 555 // CHECK2-NEXT: store i64 [[TMP1]], i64* [[TMP3]], align 8 556 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 557 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64* 558 // CHECK2-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8 559 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 560 // CHECK2-NEXT: store i8* null, i8** [[TMP6]], align 8 561 // CHECK2-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 562 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 563 // CHECK2-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 8 564 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 565 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 566 // CHECK2-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 8 567 // CHECK2-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 568 // CHECK2-NEXT: store i8* null, i8** [[TMP11]], align 8 569 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 570 // CHECK2-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 571 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 572 // CHECK2-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 573 // CHECK2-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 574 // CHECK2-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 575 // CHECK2: omp_offload.failed: 576 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i64 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2:[0-9]+]] 577 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]] 578 // CHECK2: omp_offload.cont: 579 // CHECK2-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v() 580 // CHECK2-NEXT: ret i32 [[CALL]] 581 // 582 // 583 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64 584 // CHECK2-SAME: (i64 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1:[0-9]+]] { 585 // CHECK2-NEXT: entry: 586 // CHECK2-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8 587 // CHECK2-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8 588 // CHECK2-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 589 // CHECK2-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8 590 // CHECK2-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8 591 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32* 592 // CHECK2-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8 593 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 594 // CHECK2-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 8 595 // CHECK2-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 596 // CHECK2-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 597 // CHECK2-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.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) 598 // CHECK2-NEXT: ret void 599 // 600 // 601 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined. 602 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 603 // CHECK2-NEXT: entry: 604 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 605 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 606 // CHECK2-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 607 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 608 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 609 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 610 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 611 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 612 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 613 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4 614 // CHECK2-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 615 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 616 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 617 // CHECK2-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 618 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 619 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 620 // CHECK2-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 621 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 622 // CHECK2-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 623 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 624 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 625 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 626 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 627 // CHECK2-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 628 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 629 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 630 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 631 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 632 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 633 // CHECK2: cond.true: 634 // CHECK2-NEXT: br label [[COND_END:%.*]] 635 // CHECK2: cond.false: 636 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 637 // CHECK2-NEXT: br label [[COND_END]] 638 // CHECK2: cond.end: 639 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 640 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 641 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 642 // CHECK2-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 643 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 644 // CHECK2: omp.inner.for.cond: 645 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 646 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 647 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 648 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 649 // CHECK2: omp.inner.for.body: 650 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 651 // CHECK2-NEXT: [[TMP13:%.*]] = zext i32 [[TMP12]] to i64 652 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 653 // CHECK2-NEXT: [[TMP15:%.*]] = zext i32 [[TMP14]] to i64 654 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 655 // CHECK2-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP16]], align 8 656 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 657 // CHECK2-NEXT: store i32* [[TMP4]], i32** [[TMP17]], align 8 658 // CHECK2-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.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP13]], i64 [[TMP15]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 659 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 660 // CHECK2: omp.inner.for.inc: 661 // CHECK2-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 662 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 663 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 664 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 665 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]] 666 // CHECK2: omp.inner.for.end: 667 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 668 // CHECK2: omp.loop.exit: 669 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 670 // CHECK2-NEXT: ret void 671 // 672 // 673 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1 674 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 675 // CHECK2-NEXT: entry: 676 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 677 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 678 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 679 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 680 // CHECK2-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 681 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 682 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 683 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 684 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 685 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 686 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 687 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4 688 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 689 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 690 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 691 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 692 // CHECK2-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 693 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 694 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 695 // CHECK2-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 696 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 697 // CHECK2-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 698 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 699 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 700 // CHECK2-NEXT: [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 701 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP5]] to i32 702 // CHECK2-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 703 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP6]] to i32 704 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 705 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 706 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 707 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 708 // CHECK2-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 709 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 710 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 711 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 712 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 713 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 714 // CHECK2: cond.true: 715 // CHECK2-NEXT: br label [[COND_END:%.*]] 716 // CHECK2: cond.false: 717 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 718 // CHECK2-NEXT: br label [[COND_END]] 719 // CHECK2: cond.end: 720 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 721 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 722 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 723 // CHECK2-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 724 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 725 // CHECK2: omp.inner.for.cond: 726 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 727 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 728 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 729 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 730 // CHECK2: omp.inner.for.body: 731 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 732 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 733 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 734 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4 735 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 736 // CHECK2-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 737 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64 738 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] 739 // CHECK2-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 740 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 741 // CHECK2: omp.body.continue: 742 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 743 // CHECK2: omp.inner.for.inc: 744 // CHECK2-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 745 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1 746 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 747 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]] 748 // CHECK2: omp.inner.for.end: 749 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 750 // CHECK2: omp.loop.exit: 751 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 752 // CHECK2-NEXT: ret void 753 // 754 // 755 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 756 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] comdat { 757 // CHECK2-NEXT: entry: 758 // CHECK2-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 759 // CHECK2-NEXT: [[X_CASTED:%.*]] = alloca i64, align 8 760 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8 761 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8 762 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8 763 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 764 // CHECK2-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 765 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32* 766 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4 767 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[X_CASTED]], align 8 768 // CHECK2-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 769 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64* 770 // CHECK2-NEXT: store i64 [[TMP1]], i64* [[TMP3]], align 8 771 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 772 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64* 773 // CHECK2-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8 774 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 775 // CHECK2-NEXT: store i8* null, i8** [[TMP6]], align 8 776 // CHECK2-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 777 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 778 // CHECK2-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 8 779 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 780 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 781 // CHECK2-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 8 782 // CHECK2-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1 783 // CHECK2-NEXT: store i8* null, i8** [[TMP11]], align 8 784 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 785 // CHECK2-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 786 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 787 // CHECK2-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 788 // CHECK2-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 789 // CHECK2-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 790 // CHECK2: omp_offload.failed: 791 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34(i64 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2]] 792 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]] 793 // CHECK2: omp_offload.cont: 794 // CHECK2-NEXT: ret i32 0 795 // 796 // 797 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34 798 // CHECK2-SAME: (i64 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] { 799 // CHECK2-NEXT: entry: 800 // CHECK2-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8 801 // CHECK2-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8 802 // CHECK2-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8 803 // CHECK2-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8 804 // CHECK2-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8 805 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32* 806 // CHECK2-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8 807 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 808 // CHECK2-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 8 809 // CHECK2-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 810 // CHECK2-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 811 // CHECK2-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.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]]) 812 // CHECK2-NEXT: ret void 813 // 814 // 815 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2 816 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 817 // CHECK2-NEXT: entry: 818 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 819 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 820 // CHECK2-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8 821 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 822 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 823 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 824 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 825 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 826 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 827 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4 828 // CHECK2-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8 829 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 830 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 831 // CHECK2-NEXT: store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8 832 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8 833 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0 834 // CHECK2-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 835 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1 836 // CHECK2-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 837 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 838 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 839 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 840 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 841 // CHECK2-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 842 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 843 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 844 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 845 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 846 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 847 // CHECK2: cond.true: 848 // CHECK2-NEXT: br label [[COND_END:%.*]] 849 // CHECK2: cond.false: 850 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 851 // CHECK2-NEXT: br label [[COND_END]] 852 // CHECK2: cond.end: 853 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 854 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 855 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 856 // CHECK2-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 857 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 858 // CHECK2: omp.inner.for.cond: 859 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 860 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 861 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 862 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 863 // CHECK2: omp.inner.for.body: 864 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 865 // CHECK2-NEXT: [[TMP13:%.*]] = zext i32 [[TMP12]] to i64 866 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 867 // CHECK2-NEXT: [[TMP15:%.*]] = zext i32 [[TMP14]] to i64 868 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 869 // CHECK2-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP16]], align 8 870 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 871 // CHECK2-NEXT: store i32* [[TMP4]], i32** [[TMP17]], align 8 872 // CHECK2-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.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP13]], i64 [[TMP15]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]]) 873 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 874 // CHECK2: omp.inner.for.inc: 875 // CHECK2-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 876 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 877 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 878 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 879 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]] 880 // CHECK2: omp.inner.for.end: 881 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 882 // CHECK2: omp.loop.exit: 883 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 884 // CHECK2-NEXT: ret void 885 // 886 // 887 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3 888 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 889 // CHECK2-NEXT: entry: 890 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 891 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 892 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 893 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 894 // CHECK2-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8 895 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 896 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 897 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 898 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 899 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 900 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 901 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4 902 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 903 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 904 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 905 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 906 // CHECK2-NEXT: store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8 907 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8 908 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0 909 // CHECK2-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 910 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1 911 // CHECK2-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 912 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 913 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 914 // CHECK2-NEXT: [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 915 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP5]] to i32 916 // CHECK2-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 917 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP6]] to i32 918 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 919 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 920 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 921 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 922 // CHECK2-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 923 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 924 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 925 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 926 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 927 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 928 // CHECK2: cond.true: 929 // CHECK2-NEXT: br label [[COND_END:%.*]] 930 // CHECK2: cond.false: 931 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 932 // CHECK2-NEXT: br label [[COND_END]] 933 // CHECK2: cond.end: 934 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 935 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 936 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 937 // CHECK2-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 938 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 939 // CHECK2: omp.inner.for.cond: 940 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 941 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 942 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 943 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 944 // CHECK2: omp.inner.for.body: 945 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 946 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 947 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 948 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4 949 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 950 // CHECK2-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 951 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64 952 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] 953 // CHECK2-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 954 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 955 // CHECK2: omp.body.continue: 956 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 957 // CHECK2: omp.inner.for.inc: 958 // CHECK2-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 959 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1 960 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 961 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]] 962 // CHECK2: omp.inner.for.end: 963 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 964 // CHECK2: omp.loop.exit: 965 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 966 // CHECK2-NEXT: ret void 967 // 968 // 969 // CHECK2-LABEL: define {{[^@]+}}@_ZTW1x 970 // CHECK2-SAME: () #[[ATTR4:[0-9]+]] comdat { 971 // CHECK2-NEXT: ret i32* @x 972 // 973 // 974 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 975 // CHECK2-SAME: () #[[ATTR5:[0-9]+]] { 976 // CHECK2-NEXT: entry: 977 // CHECK2-NEXT: call void @__tgt_register_requires(i64 1) 978 // CHECK2-NEXT: ret void 979 // 980 // 981 // CHECK3-LABEL: define {{[^@]+}}@main 982 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] { 983 // CHECK3-NEXT: entry: 984 // CHECK3-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 985 // CHECK3-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 986 // CHECK3-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4 987 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4 988 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4 989 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4 990 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 991 // CHECK3-NEXT: store i32 0, i32* [[RETVAL]], align 4 992 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 993 // CHECK3-NEXT: store i32 [[TMP0]], i32* [[X_CASTED]], align 4 994 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[X_CASTED]], align 4 995 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 996 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32* 997 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TMP3]], align 4 998 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 999 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32* 1000 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4 1001 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 1002 // CHECK3-NEXT: store i8* null, i8** [[TMP6]], align 4 1003 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 1004 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 1005 // CHECK3-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 4 1006 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 1007 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 1008 // CHECK3-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 4 1009 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 1010 // CHECK3-NEXT: store i8* null, i8** [[TMP11]], align 4 1011 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1012 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1013 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 1014 // CHECK3-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 1015 // CHECK3-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 1016 // CHECK3-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1017 // CHECK3: omp_offload.failed: 1018 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i32 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2:[0-9]+]] 1019 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] 1020 // CHECK3: omp_offload.cont: 1021 // CHECK3-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiET_v() 1022 // CHECK3-NEXT: ret i32 [[CALL]] 1023 // 1024 // 1025 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64 1026 // CHECK3-SAME: (i32 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1:[0-9]+]] { 1027 // CHECK3-NEXT: entry: 1028 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4 1029 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4 1030 // CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 1031 // CHECK3-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4 1032 // CHECK3-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4 1033 // CHECK3-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4 1034 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1035 // CHECK3-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 4 1036 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1037 // CHECK3-NEXT: store i32* [[X_ADDR]], i32** [[TMP2]], align 4 1038 // CHECK3-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.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) 1039 // CHECK3-NEXT: ret void 1040 // 1041 // 1042 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined. 1043 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1044 // CHECK3-NEXT: entry: 1045 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1046 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1047 // CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 1048 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1049 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1050 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1051 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1052 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1053 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1054 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1055 // CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 1056 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1057 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1058 // CHECK3-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 1059 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 1060 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 1061 // CHECK3-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1062 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 1063 // CHECK3-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1064 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1065 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 1066 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1067 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1068 // CHECK3-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1069 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 1070 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1071 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1072 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 1073 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1074 // CHECK3: cond.true: 1075 // CHECK3-NEXT: br label [[COND_END:%.*]] 1076 // CHECK3: cond.false: 1077 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1078 // CHECK3-NEXT: br label [[COND_END]] 1079 // CHECK3: cond.end: 1080 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 1081 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1082 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1083 // CHECK3-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 1084 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1085 // CHECK3: omp.inner.for.cond: 1086 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1087 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1088 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 1089 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1090 // CHECK3: omp.inner.for.body: 1091 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1092 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1093 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1094 // CHECK3-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP14]], align 4 1095 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1096 // CHECK3-NEXT: store i32* [[TMP4]], i32** [[TMP15]], align 4 1097 // CHECK3-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.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP12]], i32 [[TMP13]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 1098 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1099 // CHECK3: omp.inner.for.inc: 1100 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1101 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1102 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]] 1103 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1104 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1105 // CHECK3: omp.inner.for.end: 1106 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1107 // CHECK3: omp.loop.exit: 1108 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 1109 // CHECK3-NEXT: ret void 1110 // 1111 // 1112 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1 1113 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1114 // CHECK3-NEXT: entry: 1115 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1116 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1117 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1118 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1119 // CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4 1120 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1121 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1122 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1123 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1124 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1125 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1126 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1127 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1128 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1129 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1130 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1131 // CHECK3-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4 1132 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4 1133 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 1134 // CHECK3-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1135 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 1136 // CHECK3-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1137 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1138 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1139 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1140 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1141 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4 1142 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 1143 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1144 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1145 // CHECK3-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1146 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 1147 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1148 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1149 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 1150 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1151 // CHECK3: cond.true: 1152 // CHECK3-NEXT: br label [[COND_END:%.*]] 1153 // CHECK3: cond.false: 1154 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1155 // CHECK3-NEXT: br label [[COND_END]] 1156 // CHECK3: cond.end: 1157 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 1158 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1159 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1160 // CHECK3-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 1161 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1162 // CHECK3: omp.inner.for.cond: 1163 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1164 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1165 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 1166 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1167 // CHECK3: omp.inner.for.body: 1168 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1169 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 1170 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1171 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1172 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 1173 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 1174 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i32 0, i32 [[TMP16]] 1175 // CHECK3-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 1176 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1177 // CHECK3: omp.body.continue: 1178 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1179 // CHECK3: omp.inner.for.inc: 1180 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1181 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP17]], 1 1182 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4 1183 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1184 // CHECK3: omp.inner.for.end: 1185 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1186 // CHECK3: omp.loop.exit: 1187 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 1188 // CHECK3-NEXT: ret void 1189 // 1190 // 1191 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 1192 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] comdat { 1193 // CHECK3-NEXT: entry: 1194 // CHECK3-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 1195 // CHECK3-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4 1196 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4 1197 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4 1198 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4 1199 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1200 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 1201 // CHECK3-NEXT: store i32 [[TMP0]], i32* [[X_CASTED]], align 4 1202 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[X_CASTED]], align 4 1203 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1204 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32* 1205 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TMP3]], align 4 1206 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1207 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32* 1208 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4 1209 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 1210 // CHECK3-NEXT: store i8* null, i8** [[TMP6]], align 4 1211 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 1212 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 1213 // CHECK3-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 4 1214 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 1215 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 1216 // CHECK3-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 4 1217 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 1218 // CHECK3-NEXT: store i8* null, i8** [[TMP11]], align 4 1219 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1220 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1221 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 1222 // CHECK3-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 1223 // CHECK3-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 1224 // CHECK3-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1225 // CHECK3: omp_offload.failed: 1226 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34(i32 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2]] 1227 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] 1228 // CHECK3: omp_offload.cont: 1229 // CHECK3-NEXT: ret i32 0 1230 // 1231 // 1232 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34 1233 // CHECK3-SAME: (i32 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] { 1234 // CHECK3-NEXT: entry: 1235 // CHECK3-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4 1236 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4 1237 // CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4 1238 // CHECK3-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4 1239 // CHECK3-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4 1240 // CHECK3-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4 1241 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1242 // CHECK3-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 4 1243 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1244 // CHECK3-NEXT: store i32* [[X_ADDR]], i32** [[TMP2]], align 4 1245 // CHECK3-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.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]]) 1246 // CHECK3-NEXT: ret void 1247 // 1248 // 1249 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2 1250 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1251 // CHECK3-NEXT: entry: 1252 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1253 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1254 // CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4 1255 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1256 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1257 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1258 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1259 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1260 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1261 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1262 // CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4 1263 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1264 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1265 // CHECK3-NEXT: store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4 1266 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4 1267 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0 1268 // CHECK3-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1269 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1 1270 // CHECK3-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1271 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1272 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 1273 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1274 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1275 // CHECK3-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1276 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 1277 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1278 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1279 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 1280 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1281 // CHECK3: cond.true: 1282 // CHECK3-NEXT: br label [[COND_END:%.*]] 1283 // CHECK3: cond.false: 1284 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1285 // CHECK3-NEXT: br label [[COND_END]] 1286 // CHECK3: cond.end: 1287 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 1288 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1289 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1290 // CHECK3-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 1291 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1292 // CHECK3: omp.inner.for.cond: 1293 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1294 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1295 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 1296 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1297 // CHECK3: omp.inner.for.body: 1298 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1299 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1300 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1301 // CHECK3-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP14]], align 4 1302 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1303 // CHECK3-NEXT: store i32* [[TMP4]], i32** [[TMP15]], align 4 1304 // CHECK3-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.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP12]], i32 [[TMP13]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]]) 1305 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1306 // CHECK3: omp.inner.for.inc: 1307 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1308 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1309 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]] 1310 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1311 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1312 // CHECK3: omp.inner.for.end: 1313 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1314 // CHECK3: omp.loop.exit: 1315 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 1316 // CHECK3-NEXT: ret void 1317 // 1318 // 1319 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3 1320 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1321 // CHECK3-NEXT: entry: 1322 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1323 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1324 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1325 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1326 // CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4 1327 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1328 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1329 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1330 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1331 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1332 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1333 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1334 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1335 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1336 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1337 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1338 // CHECK3-NEXT: store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4 1339 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4 1340 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0 1341 // CHECK3-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1342 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1 1343 // CHECK3-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1344 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1345 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1346 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1347 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1348 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4 1349 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 1350 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1351 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1352 // CHECK3-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1353 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 1354 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1355 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1356 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 1357 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1358 // CHECK3: cond.true: 1359 // CHECK3-NEXT: br label [[COND_END:%.*]] 1360 // CHECK3: cond.false: 1361 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1362 // CHECK3-NEXT: br label [[COND_END]] 1363 // CHECK3: cond.end: 1364 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 1365 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1366 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1367 // CHECK3-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 1368 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1369 // CHECK3: omp.inner.for.cond: 1370 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1371 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1372 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 1373 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1374 // CHECK3: omp.inner.for.body: 1375 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1376 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 1377 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1378 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1379 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 1380 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 1381 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i32 0, i32 [[TMP16]] 1382 // CHECK3-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 1383 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1384 // CHECK3: omp.body.continue: 1385 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1386 // CHECK3: omp.inner.for.inc: 1387 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1388 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP17]], 1 1389 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4 1390 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1391 // CHECK3: omp.inner.for.end: 1392 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1393 // CHECK3: omp.loop.exit: 1394 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 1395 // CHECK3-NEXT: ret void 1396 // 1397 // 1398 // CHECK3-LABEL: define {{[^@]+}}@_ZTW1x 1399 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] comdat { 1400 // CHECK3-NEXT: ret i32* @x 1401 // 1402 // 1403 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1404 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] { 1405 // CHECK3-NEXT: entry: 1406 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1) 1407 // CHECK3-NEXT: ret void 1408 // 1409 // 1410 // CHECK4-LABEL: define {{[^@]+}}@main 1411 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] { 1412 // CHECK4-NEXT: entry: 1413 // CHECK4-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 1414 // CHECK4-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 1415 // CHECK4-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4 1416 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4 1417 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4 1418 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4 1419 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1420 // CHECK4-NEXT: store i32 0, i32* [[RETVAL]], align 4 1421 // CHECK4-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 1422 // CHECK4-NEXT: store i32 [[TMP0]], i32* [[X_CASTED]], align 4 1423 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[X_CASTED]], align 4 1424 // CHECK4-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1425 // CHECK4-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32* 1426 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[TMP3]], align 4 1427 // CHECK4-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1428 // CHECK4-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32* 1429 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4 1430 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 1431 // CHECK4-NEXT: store i8* null, i8** [[TMP6]], align 4 1432 // CHECK4-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 1433 // CHECK4-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 1434 // CHECK4-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 4 1435 // CHECK4-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 1436 // CHECK4-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 1437 // CHECK4-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 4 1438 // CHECK4-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 1439 // CHECK4-NEXT: store i8* null, i8** [[TMP11]], align 4 1440 // CHECK4-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1441 // CHECK4-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1442 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 1443 // CHECK4-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 1444 // CHECK4-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 1445 // CHECK4-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1446 // CHECK4: omp_offload.failed: 1447 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i32 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2:[0-9]+]] 1448 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT]] 1449 // CHECK4: omp_offload.cont: 1450 // CHECK4-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiET_v() 1451 // CHECK4-NEXT: ret i32 [[CALL]] 1452 // 1453 // 1454 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64 1455 // CHECK4-SAME: (i32 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1:[0-9]+]] { 1456 // CHECK4-NEXT: entry: 1457 // CHECK4-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4 1458 // CHECK4-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4 1459 // CHECK4-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 1460 // CHECK4-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4 1461 // CHECK4-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4 1462 // CHECK4-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4 1463 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1464 // CHECK4-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 4 1465 // CHECK4-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1466 // CHECK4-NEXT: store i32* [[X_ADDR]], i32** [[TMP2]], align 4 1467 // CHECK4-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.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) 1468 // CHECK4-NEXT: ret void 1469 // 1470 // 1471 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined. 1472 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1473 // CHECK4-NEXT: entry: 1474 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1475 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1476 // CHECK4-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 1477 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1478 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1479 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1480 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1481 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1482 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1483 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4 1484 // CHECK4-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 1485 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1486 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1487 // CHECK4-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 1488 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 1489 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 1490 // CHECK4-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1491 // CHECK4-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 1492 // CHECK4-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1493 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1494 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 1495 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1496 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1497 // CHECK4-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1498 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 1499 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1500 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1501 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 1502 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1503 // CHECK4: cond.true: 1504 // CHECK4-NEXT: br label [[COND_END:%.*]] 1505 // CHECK4: cond.false: 1506 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1507 // CHECK4-NEXT: br label [[COND_END]] 1508 // CHECK4: cond.end: 1509 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 1510 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1511 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1512 // CHECK4-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 1513 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1514 // CHECK4: omp.inner.for.cond: 1515 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1516 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1517 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 1518 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1519 // CHECK4: omp.inner.for.body: 1520 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1521 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1522 // CHECK4-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1523 // CHECK4-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP14]], align 4 1524 // CHECK4-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1525 // CHECK4-NEXT: store i32* [[TMP4]], i32** [[TMP15]], align 4 1526 // CHECK4-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.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP12]], i32 [[TMP13]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 1527 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1528 // CHECK4: omp.inner.for.inc: 1529 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1530 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1531 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]] 1532 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1533 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]] 1534 // CHECK4: omp.inner.for.end: 1535 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1536 // CHECK4: omp.loop.exit: 1537 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 1538 // CHECK4-NEXT: ret void 1539 // 1540 // 1541 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1 1542 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1543 // CHECK4-NEXT: entry: 1544 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1545 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1546 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1547 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1548 // CHECK4-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4 1549 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1550 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1551 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1552 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1553 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1554 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1555 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4 1556 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1557 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1558 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1559 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1560 // CHECK4-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4 1561 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4 1562 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 1563 // CHECK4-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1564 // CHECK4-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 1565 // CHECK4-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1566 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1567 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1568 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1569 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1570 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4 1571 // CHECK4-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 1572 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1573 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1574 // CHECK4-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1575 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 1576 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1577 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1578 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 1579 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1580 // CHECK4: cond.true: 1581 // CHECK4-NEXT: br label [[COND_END:%.*]] 1582 // CHECK4: cond.false: 1583 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1584 // CHECK4-NEXT: br label [[COND_END]] 1585 // CHECK4: cond.end: 1586 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 1587 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1588 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1589 // CHECK4-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 1590 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1591 // CHECK4: omp.inner.for.cond: 1592 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1593 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1594 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 1595 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1596 // CHECK4: omp.inner.for.body: 1597 // CHECK4-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1598 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 1599 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1600 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1601 // CHECK4-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 1602 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 1603 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i32 0, i32 [[TMP16]] 1604 // CHECK4-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 1605 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1606 // CHECK4: omp.body.continue: 1607 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1608 // CHECK4: omp.inner.for.inc: 1609 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1610 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP17]], 1 1611 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4 1612 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]] 1613 // CHECK4: omp.inner.for.end: 1614 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1615 // CHECK4: omp.loop.exit: 1616 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 1617 // CHECK4-NEXT: ret void 1618 // 1619 // 1620 // CHECK4-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 1621 // CHECK4-SAME: () #[[ATTR3:[0-9]+]] comdat { 1622 // CHECK4-NEXT: entry: 1623 // CHECK4-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 1624 // CHECK4-NEXT: [[X_CASTED:%.*]] = alloca i32, align 4 1625 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4 1626 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4 1627 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4 1628 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1629 // CHECK4-NEXT: [[TMP0:%.*]] = load i32, i32* @x, align 4 1630 // CHECK4-NEXT: store i32 [[TMP0]], i32* [[X_CASTED]], align 4 1631 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[X_CASTED]], align 4 1632 // CHECK4-NEXT: [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1633 // CHECK4-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32* 1634 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[TMP3]], align 4 1635 // CHECK4-NEXT: [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1636 // CHECK4-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32* 1637 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4 1638 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 1639 // CHECK4-NEXT: store i8* null, i8** [[TMP6]], align 4 1640 // CHECK4-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1 1641 // CHECK4-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]** 1642 // CHECK4-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 4 1643 // CHECK4-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1 1644 // CHECK4-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]** 1645 // CHECK4-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 4 1646 // CHECK4-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1 1647 // CHECK4-NEXT: store i8* null, i8** [[TMP11]], align 4 1648 // CHECK4-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1649 // CHECK4-NEXT: [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1650 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 1651 // CHECK4-NEXT: [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 1652 // CHECK4-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0 1653 // CHECK4-NEXT: br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1654 // CHECK4: omp_offload.failed: 1655 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34(i32 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2]] 1656 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT]] 1657 // CHECK4: omp_offload.cont: 1658 // CHECK4-NEXT: ret i32 0 1659 // 1660 // 1661 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34 1662 // CHECK4-SAME: (i32 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] { 1663 // CHECK4-NEXT: entry: 1664 // CHECK4-NEXT: [[X_ADDR:%.*]] = alloca i32, align 4 1665 // CHECK4-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4 1666 // CHECK4-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 4 1667 // CHECK4-NEXT: store i32 [[X]], i32* [[X_ADDR]], align 4 1668 // CHECK4-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4 1669 // CHECK4-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4 1670 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1671 // CHECK4-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 4 1672 // CHECK4-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1673 // CHECK4-NEXT: store i32* [[X_ADDR]], i32** [[TMP2]], align 4 1674 // CHECK4-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.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]]) 1675 // CHECK4-NEXT: ret void 1676 // 1677 // 1678 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2 1679 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1680 // CHECK4-NEXT: entry: 1681 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1682 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1683 // CHECK4-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 4 1684 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1685 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1686 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1687 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1688 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1689 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1690 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4 1691 // CHECK4-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 4 1692 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1693 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1694 // CHECK4-NEXT: store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 4 1695 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 4 1696 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0 1697 // CHECK4-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1698 // CHECK4-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1 1699 // CHECK4-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1700 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1701 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 1702 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1703 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1704 // CHECK4-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1705 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 1706 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1707 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1708 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 1709 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1710 // CHECK4: cond.true: 1711 // CHECK4-NEXT: br label [[COND_END:%.*]] 1712 // CHECK4: cond.false: 1713 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1714 // CHECK4-NEXT: br label [[COND_END]] 1715 // CHECK4: cond.end: 1716 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 1717 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1718 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1719 // CHECK4-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 1720 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1721 // CHECK4: omp.inner.for.cond: 1722 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1723 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1724 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 1725 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1726 // CHECK4: omp.inner.for.body: 1727 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1728 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1729 // CHECK4-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1730 // CHECK4-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP14]], align 4 1731 // CHECK4-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1732 // CHECK4-NEXT: store i32* [[TMP4]], i32** [[TMP15]], align 4 1733 // CHECK4-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.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP12]], i32 [[TMP13]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]]) 1734 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1735 // CHECK4: omp.inner.for.inc: 1736 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1737 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1738 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], [[TMP17]] 1739 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1740 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]] 1741 // CHECK4: omp.inner.for.end: 1742 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1743 // CHECK4: omp.loop.exit: 1744 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 1745 // CHECK4-NEXT: ret void 1746 // 1747 // 1748 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3 1749 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1750 // CHECK4-NEXT: entry: 1751 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1752 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1753 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1754 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1755 // CHECK4-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 4 1756 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1757 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1758 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1759 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1760 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1761 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1762 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4 1763 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1764 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1765 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1766 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1767 // CHECK4-NEXT: store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 4 1768 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 4 1769 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0 1770 // CHECK4-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 4 1771 // CHECK4-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1 1772 // CHECK4-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 4 1773 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1774 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1775 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1776 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1777 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_LB]], align 4 1778 // CHECK4-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4 1779 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1780 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1781 // CHECK4-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1782 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 1783 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1784 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1785 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 1786 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1787 // CHECK4: cond.true: 1788 // CHECK4-NEXT: br label [[COND_END:%.*]] 1789 // CHECK4: cond.false: 1790 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1791 // CHECK4-NEXT: br label [[COND_END]] 1792 // CHECK4: cond.end: 1793 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 1794 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1795 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1796 // CHECK4-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 1797 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1798 // CHECK4: omp.inner.for.cond: 1799 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1800 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1801 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 1802 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1803 // CHECK4: omp.inner.for.body: 1804 // CHECK4-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1805 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 1806 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1807 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1808 // CHECK4-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 1809 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 1810 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i32 0, i32 [[TMP16]] 1811 // CHECK4-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 1812 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1813 // CHECK4: omp.body.continue: 1814 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1815 // CHECK4: omp.inner.for.inc: 1816 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1817 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP17]], 1 1818 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4 1819 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]] 1820 // CHECK4: omp.inner.for.end: 1821 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1822 // CHECK4: omp.loop.exit: 1823 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 1824 // CHECK4-NEXT: ret void 1825 // 1826 // 1827 // CHECK4-LABEL: define {{[^@]+}}@_ZTW1x 1828 // CHECK4-SAME: () #[[ATTR4:[0-9]+]] comdat { 1829 // CHECK4-NEXT: ret i32* @x 1830 // 1831 // 1832 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1833 // CHECK4-SAME: () #[[ATTR5:[0-9]+]] { 1834 // CHECK4-NEXT: entry: 1835 // CHECK4-NEXT: call void @__tgt_register_requires(i64 1) 1836 // CHECK4-NEXT: ret void 1837 // 1838 // 1839 // CHECK9-LABEL: define {{[^@]+}}@main 1840 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] { 1841 // CHECK9-NEXT: entry: 1842 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 1843 // CHECK9-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 1844 // CHECK9-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8 1845 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4 1846 // CHECK9-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 0 1847 // CHECK9-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP0]], align 8 1848 // CHECK9-NEXT: call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 8 dereferenceable(8) [[REF_TMP]]) 1849 // CHECK9-NEXT: ret i32 0 1850 // 1851 // 1852 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46 1853 // CHECK9-SAME: (i64 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR2:[0-9]+]] { 1854 // CHECK9-NEXT: entry: 1855 // CHECK9-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8 1856 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8 1857 // CHECK9-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 1858 // CHECK9-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8 1859 // CHECK9-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8 1860 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32* 1861 // CHECK9-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8 1862 // CHECK9-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1863 // CHECK9-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 8 1864 // CHECK9-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1865 // CHECK9-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 1866 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) 1867 // CHECK9-NEXT: ret void 1868 // 1869 // 1870 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined. 1871 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] { 1872 // CHECK9-NEXT: entry: 1873 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1874 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1875 // CHECK9-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 1876 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1877 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1878 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1879 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1880 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1881 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1882 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 1883 // CHECK9-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 1884 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1885 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1886 // CHECK9-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 1887 // CHECK9-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 1888 // CHECK9-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 1889 // CHECK9-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 1890 // CHECK9-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 1891 // CHECK9-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 1892 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1893 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 1894 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1895 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1896 // CHECK9-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1897 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 1898 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1899 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1900 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 1901 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1902 // CHECK9: cond.true: 1903 // CHECK9-NEXT: br label [[COND_END:%.*]] 1904 // CHECK9: cond.false: 1905 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1906 // CHECK9-NEXT: br label [[COND_END]] 1907 // CHECK9: cond.end: 1908 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 1909 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1910 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1911 // CHECK9-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 1912 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1913 // CHECK9: omp.inner.for.cond: 1914 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1915 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1916 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 1917 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1918 // CHECK9: omp.inner.for.body: 1919 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1920 // CHECK9-NEXT: [[TMP13:%.*]] = zext i32 [[TMP12]] to i64 1921 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1922 // CHECK9-NEXT: [[TMP15:%.*]] = zext i32 [[TMP14]] to i64 1923 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1924 // CHECK9-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP16]], align 8 1925 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 1926 // CHECK9-NEXT: store i32* [[TMP4]], i32** [[TMP17]], align 8 1927 // CHECK9-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.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP13]], i64 [[TMP15]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 1928 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1929 // CHECK9: omp.inner.for.inc: 1930 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1931 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1932 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 1933 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1934 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]] 1935 // CHECK9: omp.inner.for.end: 1936 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1937 // CHECK9: omp.loop.exit: 1938 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 1939 // CHECK9-NEXT: ret void 1940 // 1941 // 1942 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1 1943 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] { 1944 // CHECK9-NEXT: entry: 1945 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1946 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1947 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 1948 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 1949 // CHECK9-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 1950 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1951 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1952 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1953 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1954 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1955 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1956 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 1957 // CHECK9-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON_1:%.*]], align 8 1958 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1959 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1960 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 1961 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 1962 // CHECK9-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 1963 // CHECK9-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 1964 // CHECK9-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 1965 // CHECK9-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 1966 // CHECK9-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 1967 // CHECK9-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 1968 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1969 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1970 // CHECK9-NEXT: [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 1971 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP5]] to i32 1972 // CHECK9-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 1973 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP6]] to i32 1974 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 1975 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 1976 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1977 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1978 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1979 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 1980 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1981 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1982 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 1983 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1984 // CHECK9: cond.true: 1985 // CHECK9-NEXT: br label [[COND_END:%.*]] 1986 // CHECK9: cond.false: 1987 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1988 // CHECK9-NEXT: br label [[COND_END]] 1989 // CHECK9: cond.end: 1990 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 1991 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1992 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1993 // CHECK9-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 1994 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1995 // CHECK9: omp.inner.for.cond: 1996 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1997 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1998 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 1999 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2000 // CHECK9: omp.inner.for.body: 2001 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2002 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 2003 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 2004 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4 2005 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 2006 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 2007 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64 2008 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] 2009 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 2010 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], %class.anon.1* [[REF_TMP]], i32 0, i32 0 2011 // CHECK9-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP17]], align 8 2012 // CHECK9-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], %class.anon.1* [[REF_TMP]], i32 0, i32 1 2013 // CHECK9-NEXT: store i32* [[I]], i32** [[TMP18]], align 8 2014 // CHECK9-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], %class.anon.1* [[REF_TMP]], i32 0, i32 2 2015 // CHECK9-NEXT: store i32* [[TMP4]], i32** [[TMP19]], align 8 2016 // CHECK9-NEXT: call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.1* nonnull align 8 dereferenceable(24) [[REF_TMP]]) 2017 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 2018 // CHECK9: omp.body.continue: 2019 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2020 // CHECK9: omp.inner.for.inc: 2021 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2022 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP20]], 1 2023 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 2024 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]] 2025 // CHECK9: omp.inner.for.end: 2026 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2027 // CHECK9: omp.loop.exit: 2028 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 2029 // CHECK9-NEXT: ret void 2030 // 2031 // 2032 // CHECK9-LABEL: define {{[^@]+}}@_ZTW1x 2033 // CHECK9-SAME: () #[[ATTR4:[0-9]+]] comdat { 2034 // CHECK9-NEXT: ret i32* @x 2035 // 2036 // 2037 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 2038 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] { 2039 // CHECK9-NEXT: entry: 2040 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1) 2041 // CHECK9-NEXT: ret void 2042 // 2043 // 2044 // CHECK10-LABEL: define {{[^@]+}}@main 2045 // CHECK10-SAME: () #[[ATTR0:[0-9]+]] { 2046 // CHECK10-NEXT: entry: 2047 // CHECK10-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 2048 // CHECK10-NEXT: [[A:%.*]] = alloca [2 x i32], align 4 2049 // CHECK10-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8 2050 // CHECK10-NEXT: store i32 0, i32* [[RETVAL]], align 4 2051 // CHECK10-NEXT: [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 0 2052 // CHECK10-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[TMP0]], align 8 2053 // CHECK10-NEXT: call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 8 dereferenceable(8) [[REF_TMP]]) 2054 // CHECK10-NEXT: ret i32 0 2055 // 2056 // 2057 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46 2058 // CHECK10-SAME: (i64 [[X:%.*]], [2 x i32]* nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR2:[0-9]+]] { 2059 // CHECK10-NEXT: entry: 2060 // CHECK10-NEXT: [[X_ADDR:%.*]] = alloca i64, align 8 2061 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8 2062 // CHECK10-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 2063 // CHECK10-NEXT: store i64 [[X]], i64* [[X_ADDR]], align 8 2064 // CHECK10-NEXT: store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8 2065 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32* 2066 // CHECK10-NEXT: [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8 2067 // CHECK10-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 2068 // CHECK10-NEXT: store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP1]], align 8 2069 // CHECK10-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 2070 // CHECK10-NEXT: store i32* [[CONV]], i32** [[TMP2]], align 8 2071 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]) 2072 // CHECK10-NEXT: ret void 2073 // 2074 // 2075 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined. 2076 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] { 2077 // CHECK10-NEXT: entry: 2078 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2079 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2080 // CHECK10-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 2081 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2082 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4 2083 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 2084 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 2085 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2086 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2087 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4 2088 // CHECK10-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 2089 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2090 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2091 // CHECK10-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 2092 // CHECK10-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 2093 // CHECK10-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 2094 // CHECK10-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 2095 // CHECK10-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1 2096 // CHECK10-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 2097 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 2098 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 2099 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2100 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2101 // CHECK10-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2102 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4 2103 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP6]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2104 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2105 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 1 2106 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2107 // CHECK10: cond.true: 2108 // CHECK10-NEXT: br label [[COND_END:%.*]] 2109 // CHECK10: cond.false: 2110 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2111 // CHECK10-NEXT: br label [[COND_END]] 2112 // CHECK10: cond.end: 2113 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ] 2114 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 2115 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2116 // CHECK10-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4 2117 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2118 // CHECK10: omp.inner.for.cond: 2119 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2120 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2121 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]] 2122 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2123 // CHECK10: omp.inner.for.body: 2124 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 2125 // CHECK10-NEXT: [[TMP13:%.*]] = zext i32 [[TMP12]] to i64 2126 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 2127 // CHECK10-NEXT: [[TMP15:%.*]] = zext i32 [[TMP14]] to i64 2128 // CHECK10-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 2129 // CHECK10-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP16]], align 8 2130 // CHECK10-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1 2131 // CHECK10-NEXT: store i32* [[TMP4]], i32** [[TMP17]], align 8 2132 // CHECK10-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.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP13]], i64 [[TMP15]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 2133 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2134 // CHECK10: omp.inner.for.inc: 2135 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2136 // CHECK10-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 2137 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 2138 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 2139 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]] 2140 // CHECK10: omp.inner.for.end: 2141 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2142 // CHECK10: omp.loop.exit: 2143 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]]) 2144 // CHECK10-NEXT: ret void 2145 // 2146 // 2147 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1 2148 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] { 2149 // CHECK10-NEXT: entry: 2150 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 2151 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 2152 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 2153 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 2154 // CHECK10-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 2155 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 2156 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4 2157 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 2158 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 2159 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 2160 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 2161 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4 2162 // CHECK10-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON_1:%.*]], align 8 2163 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 2164 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 2165 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2166 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2167 // CHECK10-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 2168 // CHECK10-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 2169 // CHECK10-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 2170 // CHECK10-NEXT: [[TMP2:%.*]] = load [2 x i32]*, [2 x i32]** [[TMP1]], align 8 2171 // CHECK10-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1 2172 // CHECK10-NEXT: [[TMP4:%.*]] = load i32*, i32** [[TMP3]], align 8 2173 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 2174 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 2175 // CHECK10-NEXT: [[TMP5:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 2176 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP5]] to i32 2177 // CHECK10-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 2178 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP6]] to i32 2179 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 2180 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 2181 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 2182 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 2183 // CHECK10-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 2184 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4 2185 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 2186 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2187 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1 2188 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 2189 // CHECK10: cond.true: 2190 // CHECK10-NEXT: br label [[COND_END:%.*]] 2191 // CHECK10: cond.false: 2192 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2193 // CHECK10-NEXT: br label [[COND_END]] 2194 // CHECK10: cond.end: 2195 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ] 2196 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 2197 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 2198 // CHECK10-NEXT: store i32 [[TMP11]], i32* [[DOTOMP_IV]], align 4 2199 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 2200 // CHECK10: omp.inner.for.cond: 2201 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2202 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 2203 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]] 2204 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 2205 // CHECK10: omp.inner.for.body: 2206 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2207 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1 2208 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 2209 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4 2210 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP4]], align 4 2211 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4 2212 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64 2213 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]] 2214 // CHECK10-NEXT: store i32 [[TMP15]], i32* [[ARRAYIDX]], align 4 2215 // CHECK10-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], %class.anon.1* [[REF_TMP]], i32 0, i32 0 2216 // CHECK10-NEXT: store [2 x i32]* [[TMP2]], [2 x i32]** [[TMP17]], align 8 2217 // CHECK10-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], %class.anon.1* [[REF_TMP]], i32 0, i32 1 2218 // CHECK10-NEXT: store i32* [[I]], i32** [[TMP18]], align 8 2219 // CHECK10-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[CLASS_ANON_1]], %class.anon.1* [[REF_TMP]], i32 0, i32 2 2220 // CHECK10-NEXT: store i32* [[TMP4]], i32** [[TMP19]], align 8 2221 // CHECK10-NEXT: call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.1* nonnull align 8 dereferenceable(24) [[REF_TMP]]) 2222 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 2223 // CHECK10: omp.body.continue: 2224 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 2225 // CHECK10: omp.inner.for.inc: 2226 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 2227 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP20]], 1 2228 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 2229 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]] 2230 // CHECK10: omp.inner.for.end: 2231 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 2232 // CHECK10: omp.loop.exit: 2233 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]]) 2234 // CHECK10-NEXT: ret void 2235 // 2236 // 2237 // CHECK10-LABEL: define {{[^@]+}}@_ZTW1x 2238 // CHECK10-SAME: () #[[ATTR4:[0-9]+]] comdat { 2239 // CHECK10-NEXT: ret i32* @x 2240 // 2241 // 2242 // CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 2243 // CHECK10-SAME: () #[[ATTR5:[0-9]+]] { 2244 // CHECK10-NEXT: entry: 2245 // CHECK10-NEXT: call void @__tgt_register_requires(i64 1) 2246 // CHECK10-NEXT: ret void 2247 // 2248