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 template <typename T> 29 T tmain() { 30 T t_var = T(); 31 T vec[] = {1, 2}; 32 #pragma omp target teams distribute reduction(+: t_var) 33 for (int i = 0; i < 2; ++i) { 34 t_var += (T) i; 35 } 36 return T(); 37 } 38 39 int main() { 40 static int sivar; 41 #ifdef LAMBDA 42 43 [&]() { 44 #pragma omp target teams distribute reduction(+: sivar) 45 for (int i = 0; i < 2; ++i) { 46 47 // Skip global and bound tid vars 48 49 sivar += i; 50 51 [&]() { 52 53 sivar += 4; 54 55 }(); 56 } 57 }(); 58 return 0; 59 #else 60 #pragma omp target teams distribute reduction(+: sivar) 61 for (int i = 0; i < 2; ++i) { 62 sivar += i; 63 } 64 return tmain<int>(); 65 #endif 66 } 67 68 69 70 71 // Skip global and bound tid vars 72 73 74 75 76 77 // Skip global and bound tid vars 78 79 80 #endif 81 // CHECK1-LABEL: define {{[^@]+}}@main 82 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] { 83 // CHECK1-NEXT: entry: 84 // CHECK1-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 85 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8 86 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8 87 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8 88 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 89 // CHECK1-NEXT: store i32 0, i32* [[RETVAL]], align 4 90 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 91 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32** 92 // CHECK1-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP1]], align 8 93 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 94 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32** 95 // CHECK1-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP3]], align 8 96 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 97 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8 98 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 99 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 100 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 101 // CHECK1-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 102 // CHECK1-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0 103 // CHECK1-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 104 // CHECK1: omp_offload.failed: 105 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60(i32* @_ZZ4mainE5sivar) #[[ATTR2:[0-9]+]] 106 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] 107 // CHECK1: omp_offload.cont: 108 // CHECK1-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v() 109 // CHECK1-NEXT: ret i32 [[CALL]] 110 // 111 // 112 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60 113 // CHECK1-SAME: (i32* nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] { 114 // CHECK1-NEXT: entry: 115 // CHECK1-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 116 // CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 117 // CHECK1-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 118 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 119 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 120 // CHECK1-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 8 121 // 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_]]) 122 // CHECK1-NEXT: ret void 123 // 124 // 125 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined. 126 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 127 // CHECK1-NEXT: entry: 128 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 129 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 130 // CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 131 // CHECK1-NEXT: [[SIVAR:%.*]] = alloca i32, align 4 132 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 133 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 134 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 135 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 136 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 137 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 138 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 139 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 140 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 141 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 142 // CHECK1-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 143 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 144 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 145 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 146 // CHECK1-NEXT: store i32 0, i32* [[SIVAR]], align 4 147 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 148 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 149 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 150 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 151 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 152 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 153 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 154 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 155 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 156 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 157 // CHECK1: cond.true: 158 // CHECK1-NEXT: br label [[COND_END:%.*]] 159 // CHECK1: cond.false: 160 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 161 // CHECK1-NEXT: br label [[COND_END]] 162 // CHECK1: cond.end: 163 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 164 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 165 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 166 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 167 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 168 // CHECK1: omp.inner.for.cond: 169 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 170 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 171 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 172 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 173 // CHECK1: omp.inner.for.body: 174 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 175 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 176 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 177 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4 178 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 179 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR]], align 4 180 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 181 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[SIVAR]], align 4 182 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 183 // CHECK1: omp.body.continue: 184 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 185 // CHECK1: omp.inner.for.inc: 186 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 187 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 188 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 189 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 190 // CHECK1: omp.inner.for.end: 191 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 192 // CHECK1: omp.loop.exit: 193 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 194 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 195 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR]] to i8* 196 // CHECK1-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 197 // CHECK1-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 198 // CHECK1-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 199 // CHECK1-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 200 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 201 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 202 // CHECK1-NEXT: ] 203 // CHECK1: .omp.reduction.case1: 204 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 205 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR]], align 4 206 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 207 // CHECK1-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 208 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 209 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 210 // CHECK1: .omp.reduction.case2: 211 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR]], align 4 212 // CHECK1-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 213 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 214 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 215 // CHECK1: .omp.reduction.default: 216 // CHECK1-NEXT: ret void 217 // 218 // 219 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 220 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] { 221 // CHECK1-NEXT: entry: 222 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 223 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 224 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 225 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 226 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 227 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 228 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 229 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 230 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 231 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 232 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 233 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 234 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 235 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 236 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 237 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 238 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 239 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 240 // CHECK1-NEXT: ret void 241 // 242 // 243 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 244 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] comdat { 245 // CHECK1-NEXT: entry: 246 // CHECK1-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 247 // CHECK1-NEXT: [[VEC:%.*]] = alloca [2 x i32], align 4 248 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8 249 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8 250 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8 251 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 252 // CHECK1-NEXT: store i32 0, i32* [[T_VAR]], align 4 253 // CHECK1-NEXT: [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8* 254 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i64 8, i1 false) 255 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 256 // CHECK1-NEXT: [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32** 257 // CHECK1-NEXT: store i32* [[T_VAR]], i32** [[TMP2]], align 8 258 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 259 // CHECK1-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32** 260 // CHECK1-NEXT: store i32* [[T_VAR]], i32** [[TMP4]], align 8 261 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 262 // CHECK1-NEXT: store i8* null, i8** [[TMP5]], align 8 263 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 264 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 265 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 266 // CHECK1-NEXT: [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 267 // CHECK1-NEXT: [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0 268 // CHECK1-NEXT: br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 269 // CHECK1: omp_offload.failed: 270 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(i32* [[T_VAR]]) #[[ATTR2]] 271 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] 272 // CHECK1: omp_offload.cont: 273 // CHECK1-NEXT: ret i32 0 274 // 275 // 276 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32 277 // CHECK1-SAME: (i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 278 // CHECK1-NEXT: entry: 279 // CHECK1-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 8 280 // CHECK1-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 281 // CHECK1-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8 282 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8 283 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 284 // CHECK1-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 8 285 // 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.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 286 // CHECK1-NEXT: ret void 287 // 288 // 289 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1 290 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 291 // CHECK1-NEXT: entry: 292 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 293 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 294 // CHECK1-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 295 // CHECK1-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 296 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 297 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 298 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 299 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 300 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 301 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 302 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 303 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 304 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 305 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 306 // CHECK1-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 307 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 308 // CHECK1-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 309 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 310 // CHECK1-NEXT: store i32 0, i32* [[T_VAR]], align 4 311 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 312 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 313 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 314 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 315 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 316 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 317 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 318 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 319 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 320 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 321 // CHECK1: cond.true: 322 // CHECK1-NEXT: br label [[COND_END:%.*]] 323 // CHECK1: cond.false: 324 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 325 // CHECK1-NEXT: br label [[COND_END]] 326 // CHECK1: cond.end: 327 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 328 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 329 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 330 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 331 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 332 // CHECK1: omp.inner.for.cond: 333 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 334 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 335 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 336 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 337 // CHECK1: omp.inner.for.body: 338 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 339 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 340 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 341 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4 342 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 343 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[T_VAR]], align 4 344 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 345 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[T_VAR]], align 4 346 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 347 // CHECK1: omp.body.continue: 348 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 349 // CHECK1: omp.inner.for.inc: 350 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 351 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 352 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 353 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 354 // CHECK1: omp.inner.for.end: 355 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 356 // CHECK1: omp.loop.exit: 357 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 358 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 359 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i32* [[T_VAR]] to i8* 360 // CHECK1-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 361 // CHECK1-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 362 // CHECK1-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var) 363 // CHECK1-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 364 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 365 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 366 // CHECK1-NEXT: ] 367 // CHECK1: .omp.reduction.case1: 368 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 369 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[T_VAR]], align 4 370 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 371 // CHECK1-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 372 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 373 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 374 // CHECK1: .omp.reduction.case2: 375 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[T_VAR]], align 4 376 // CHECK1-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 377 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 378 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 379 // CHECK1: .omp.reduction.default: 380 // CHECK1-NEXT: ret void 381 // 382 // 383 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2 384 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] { 385 // CHECK1-NEXT: entry: 386 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 387 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 388 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 389 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 390 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 391 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 392 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 393 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 394 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 395 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 396 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 397 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 398 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 399 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 400 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 401 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 402 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 403 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 404 // CHECK1-NEXT: ret void 405 // 406 // 407 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 408 // CHECK1-SAME: () #[[ATTR7:[0-9]+]] { 409 // CHECK1-NEXT: entry: 410 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1) 411 // CHECK1-NEXT: ret void 412 // 413 // 414 // CHECK2-LABEL: define {{[^@]+}}@main 415 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] { 416 // CHECK2-NEXT: entry: 417 // CHECK2-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 418 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8 419 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8 420 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8 421 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 422 // CHECK2-NEXT: store i32 0, i32* [[RETVAL]], align 4 423 // CHECK2-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 424 // CHECK2-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32** 425 // CHECK2-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP1]], align 8 426 // CHECK2-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 427 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32** 428 // CHECK2-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP3]], align 8 429 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 430 // CHECK2-NEXT: store i8* null, i8** [[TMP4]], align 8 431 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 432 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 433 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 434 // CHECK2-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 435 // CHECK2-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0 436 // CHECK2-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 437 // CHECK2: omp_offload.failed: 438 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60(i32* @_ZZ4mainE5sivar) #[[ATTR2:[0-9]+]] 439 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]] 440 // CHECK2: omp_offload.cont: 441 // CHECK2-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v() 442 // CHECK2-NEXT: ret i32 [[CALL]] 443 // 444 // 445 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60 446 // CHECK2-SAME: (i32* nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] { 447 // CHECK2-NEXT: entry: 448 // CHECK2-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 449 // CHECK2-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 450 // CHECK2-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 451 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 452 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 453 // CHECK2-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 8 454 // 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_]]) 455 // CHECK2-NEXT: ret void 456 // 457 // 458 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined. 459 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 460 // CHECK2-NEXT: entry: 461 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 462 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 463 // CHECK2-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 464 // CHECK2-NEXT: [[SIVAR:%.*]] = alloca i32, align 4 465 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 466 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 467 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 468 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 469 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 470 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 471 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4 472 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 473 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 474 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 475 // CHECK2-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 476 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 477 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 478 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 479 // CHECK2-NEXT: store i32 0, i32* [[SIVAR]], align 4 480 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 481 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 482 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 483 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 484 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 485 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 486 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 487 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 488 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 489 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 490 // CHECK2: cond.true: 491 // CHECK2-NEXT: br label [[COND_END:%.*]] 492 // CHECK2: cond.false: 493 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 494 // CHECK2-NEXT: br label [[COND_END]] 495 // CHECK2: cond.end: 496 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 497 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 498 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 499 // CHECK2-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 500 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 501 // CHECK2: omp.inner.for.cond: 502 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 503 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 504 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 505 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 506 // CHECK2: omp.inner.for.body: 507 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 508 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 509 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 510 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4 511 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 512 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR]], align 4 513 // CHECK2-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 514 // CHECK2-NEXT: store i32 [[ADD2]], i32* [[SIVAR]], align 4 515 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 516 // CHECK2: omp.body.continue: 517 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 518 // CHECK2: omp.inner.for.inc: 519 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 520 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 521 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 522 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]] 523 // CHECK2: omp.inner.for.end: 524 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 525 // CHECK2: omp.loop.exit: 526 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 527 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 528 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR]] to i8* 529 // CHECK2-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 530 // CHECK2-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 531 // CHECK2-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 532 // CHECK2-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 533 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 534 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 535 // CHECK2-NEXT: ] 536 // CHECK2: .omp.reduction.case1: 537 // CHECK2-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 538 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR]], align 4 539 // CHECK2-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 540 // CHECK2-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 541 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 542 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 543 // CHECK2: .omp.reduction.case2: 544 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR]], align 4 545 // CHECK2-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 546 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 547 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 548 // CHECK2: .omp.reduction.default: 549 // CHECK2-NEXT: ret void 550 // 551 // 552 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 553 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] { 554 // CHECK2-NEXT: entry: 555 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 556 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 557 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 558 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 559 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 560 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 561 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 562 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 563 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 564 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 565 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 566 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 567 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 568 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 569 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 570 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 571 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 572 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 573 // CHECK2-NEXT: ret void 574 // 575 // 576 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 577 // CHECK2-SAME: () #[[ATTR5:[0-9]+]] comdat { 578 // CHECK2-NEXT: entry: 579 // CHECK2-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 580 // CHECK2-NEXT: [[VEC:%.*]] = alloca [2 x i32], align 4 581 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8 582 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8 583 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8 584 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 585 // CHECK2-NEXT: store i32 0, i32* [[T_VAR]], align 4 586 // CHECK2-NEXT: [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8* 587 // CHECK2-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i64 8, i1 false) 588 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 589 // CHECK2-NEXT: [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32** 590 // CHECK2-NEXT: store i32* [[T_VAR]], i32** [[TMP2]], align 8 591 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 592 // CHECK2-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32** 593 // CHECK2-NEXT: store i32* [[T_VAR]], i32** [[TMP4]], align 8 594 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 595 // CHECK2-NEXT: store i8* null, i8** [[TMP5]], align 8 596 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 597 // CHECK2-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 598 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 599 // CHECK2-NEXT: [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 600 // CHECK2-NEXT: [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0 601 // CHECK2-NEXT: br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 602 // CHECK2: omp_offload.failed: 603 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(i32* [[T_VAR]]) #[[ATTR2]] 604 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]] 605 // CHECK2: omp_offload.cont: 606 // CHECK2-NEXT: ret i32 0 607 // 608 // 609 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32 610 // CHECK2-SAME: (i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 611 // CHECK2-NEXT: entry: 612 // CHECK2-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 8 613 // CHECK2-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8 614 // CHECK2-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8 615 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8 616 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 617 // CHECK2-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 8 618 // 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.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 619 // CHECK2-NEXT: ret void 620 // 621 // 622 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1 623 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 624 // CHECK2-NEXT: entry: 625 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 626 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 627 // CHECK2-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8 628 // CHECK2-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 629 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 630 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4 631 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 632 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 633 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 634 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 635 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4 636 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 637 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 638 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 639 // CHECK2-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8 640 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8 641 // CHECK2-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 642 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 643 // CHECK2-NEXT: store i32 0, i32* [[T_VAR]], align 4 644 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 645 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 646 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 647 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 648 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 649 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 650 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 651 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 652 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 653 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 654 // CHECK2: cond.true: 655 // CHECK2-NEXT: br label [[COND_END:%.*]] 656 // CHECK2: cond.false: 657 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 658 // CHECK2-NEXT: br label [[COND_END]] 659 // CHECK2: cond.end: 660 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 661 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 662 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 663 // CHECK2-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 664 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 665 // CHECK2: omp.inner.for.cond: 666 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 667 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 668 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 669 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 670 // CHECK2: omp.inner.for.body: 671 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 672 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 673 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 674 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4 675 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 676 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[T_VAR]], align 4 677 // CHECK2-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 678 // CHECK2-NEXT: store i32 [[ADD2]], i32* [[T_VAR]], align 4 679 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 680 // CHECK2: omp.body.continue: 681 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 682 // CHECK2: omp.inner.for.inc: 683 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 684 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 685 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 686 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]] 687 // CHECK2: omp.inner.for.end: 688 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 689 // CHECK2: omp.loop.exit: 690 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 691 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 692 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast i32* [[T_VAR]] to i8* 693 // CHECK2-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 694 // CHECK2-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 695 // CHECK2-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var) 696 // CHECK2-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 697 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 698 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 699 // CHECK2-NEXT: ] 700 // CHECK2: .omp.reduction.case1: 701 // CHECK2-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 702 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[T_VAR]], align 4 703 // CHECK2-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 704 // CHECK2-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 705 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 706 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 707 // CHECK2: .omp.reduction.case2: 708 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[T_VAR]], align 4 709 // CHECK2-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 710 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 711 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 712 // CHECK2: .omp.reduction.default: 713 // CHECK2-NEXT: ret void 714 // 715 // 716 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2 717 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] { 718 // CHECK2-NEXT: entry: 719 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 720 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 721 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 722 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 723 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 724 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 725 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 726 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 727 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 728 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 729 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 730 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 731 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 732 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 733 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 734 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 735 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 736 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 737 // CHECK2-NEXT: ret void 738 // 739 // 740 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 741 // CHECK2-SAME: () #[[ATTR7:[0-9]+]] { 742 // CHECK2-NEXT: entry: 743 // CHECK2-NEXT: call void @__tgt_register_requires(i64 1) 744 // CHECK2-NEXT: ret void 745 // 746 // 747 // CHECK3-LABEL: define {{[^@]+}}@main 748 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] { 749 // CHECK3-NEXT: entry: 750 // CHECK3-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 751 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4 752 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4 753 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4 754 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 755 // CHECK3-NEXT: store i32 0, i32* [[RETVAL]], align 4 756 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 757 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32** 758 // CHECK3-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP1]], align 4 759 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 760 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32** 761 // CHECK3-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP3]], align 4 762 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 763 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4 764 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 765 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 766 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 767 // CHECK3-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 768 // CHECK3-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0 769 // CHECK3-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 770 // CHECK3: omp_offload.failed: 771 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60(i32* @_ZZ4mainE5sivar) #[[ATTR2:[0-9]+]] 772 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] 773 // CHECK3: omp_offload.cont: 774 // CHECK3-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiET_v() 775 // CHECK3-NEXT: ret i32 [[CALL]] 776 // 777 // 778 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60 779 // CHECK3-SAME: (i32* nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] { 780 // CHECK3-NEXT: entry: 781 // CHECK3-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 4 782 // CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 783 // CHECK3-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 4 784 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 4 785 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 786 // CHECK3-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 4 787 // 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_]]) 788 // CHECK3-NEXT: ret void 789 // 790 // 791 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined. 792 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 793 // CHECK3-NEXT: entry: 794 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 795 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 796 // CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 797 // CHECK3-NEXT: [[SIVAR:%.*]] = alloca i32, align 4 798 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 799 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 800 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 801 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 802 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 803 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 804 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 805 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 806 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 807 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 808 // CHECK3-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 809 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 810 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 811 // CHECK3-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 812 // CHECK3-NEXT: store i32 0, i32* [[SIVAR]], align 4 813 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 814 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 815 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 816 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 817 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 818 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 819 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 820 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 821 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 822 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 823 // CHECK3: cond.true: 824 // CHECK3-NEXT: br label [[COND_END:%.*]] 825 // CHECK3: cond.false: 826 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 827 // CHECK3-NEXT: br label [[COND_END]] 828 // CHECK3: cond.end: 829 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 830 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 831 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 832 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 833 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 834 // CHECK3: omp.inner.for.cond: 835 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 836 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 837 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 838 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 839 // CHECK3: omp.inner.for.body: 840 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 841 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 842 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 843 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4 844 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 845 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR]], align 4 846 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 847 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[SIVAR]], align 4 848 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 849 // CHECK3: omp.body.continue: 850 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 851 // CHECK3: omp.inner.for.inc: 852 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 853 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 854 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 855 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 856 // CHECK3: omp.inner.for.end: 857 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 858 // CHECK3: omp.loop.exit: 859 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 860 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 861 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR]] to i8* 862 // CHECK3-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 4 863 // CHECK3-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 864 // CHECK3-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 1, i32 4, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 865 // CHECK3-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 866 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 867 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 868 // CHECK3-NEXT: ] 869 // CHECK3: .omp.reduction.case1: 870 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 871 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR]], align 4 872 // CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 873 // CHECK3-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 874 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 875 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 876 // CHECK3: .omp.reduction.case2: 877 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR]], align 4 878 // CHECK3-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 879 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 880 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 881 // CHECK3: .omp.reduction.default: 882 // CHECK3-NEXT: ret void 883 // 884 // 885 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 886 // CHECK3-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] { 887 // CHECK3-NEXT: entry: 888 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 889 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 890 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 891 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 892 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 893 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 894 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 895 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 896 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 897 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 898 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 899 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 900 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 901 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 902 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 903 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 904 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 905 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 906 // CHECK3-NEXT: ret void 907 // 908 // 909 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 910 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] comdat { 911 // CHECK3-NEXT: entry: 912 // CHECK3-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 913 // CHECK3-NEXT: [[VEC:%.*]] = alloca [2 x i32], align 4 914 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4 915 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4 916 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4 917 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 918 // CHECK3-NEXT: store i32 0, i32* [[T_VAR]], align 4 919 // CHECK3-NEXT: [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8* 920 // CHECK3-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i32 8, i1 false) 921 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 922 // CHECK3-NEXT: [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32** 923 // CHECK3-NEXT: store i32* [[T_VAR]], i32** [[TMP2]], align 4 924 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 925 // CHECK3-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32** 926 // CHECK3-NEXT: store i32* [[T_VAR]], i32** [[TMP4]], align 4 927 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 928 // CHECK3-NEXT: store i8* null, i8** [[TMP5]], align 4 929 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 930 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 931 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 932 // CHECK3-NEXT: [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 933 // CHECK3-NEXT: [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0 934 // CHECK3-NEXT: br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 935 // CHECK3: omp_offload.failed: 936 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(i32* [[T_VAR]]) #[[ATTR2]] 937 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] 938 // CHECK3: omp_offload.cont: 939 // CHECK3-NEXT: ret i32 0 940 // 941 // 942 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32 943 // CHECK3-SAME: (i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 944 // CHECK3-NEXT: entry: 945 // CHECK3-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 4 946 // CHECK3-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 947 // CHECK3-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4 948 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4 949 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 950 // CHECK3-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 4 951 // 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.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 952 // CHECK3-NEXT: ret void 953 // 954 // 955 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1 956 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 957 // CHECK3-NEXT: entry: 958 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 959 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 960 // CHECK3-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4 961 // CHECK3-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 962 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 963 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 964 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 965 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 966 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 967 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 968 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 969 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 970 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 971 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 972 // CHECK3-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4 973 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4 974 // CHECK3-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 975 // CHECK3-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 976 // CHECK3-NEXT: store i32 0, i32* [[T_VAR]], align 4 977 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 978 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 979 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 980 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 981 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 982 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 983 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 984 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 985 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 986 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 987 // CHECK3: cond.true: 988 // CHECK3-NEXT: br label [[COND_END:%.*]] 989 // CHECK3: cond.false: 990 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 991 // CHECK3-NEXT: br label [[COND_END]] 992 // CHECK3: cond.end: 993 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 994 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 995 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 996 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 997 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 998 // CHECK3: omp.inner.for.cond: 999 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1000 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1001 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 1002 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1003 // CHECK3: omp.inner.for.body: 1004 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1005 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 1006 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1007 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1008 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 1009 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[T_VAR]], align 4 1010 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 1011 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[T_VAR]], align 4 1012 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1013 // CHECK3: omp.body.continue: 1014 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1015 // CHECK3: omp.inner.for.inc: 1016 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1017 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 1018 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 1019 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1020 // CHECK3: omp.inner.for.end: 1021 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1022 // CHECK3: omp.loop.exit: 1023 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 1024 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 1025 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i32* [[T_VAR]] to i8* 1026 // CHECK3-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 4 1027 // CHECK3-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1028 // CHECK3-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 4, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var) 1029 // CHECK3-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1030 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1031 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1032 // CHECK3-NEXT: ] 1033 // CHECK3: .omp.reduction.case1: 1034 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 1035 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[T_VAR]], align 4 1036 // CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 1037 // CHECK3-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 1038 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1039 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1040 // CHECK3: .omp.reduction.case2: 1041 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[T_VAR]], align 4 1042 // CHECK3-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 1043 // CHECK3-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1044 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1045 // CHECK3: .omp.reduction.default: 1046 // CHECK3-NEXT: ret void 1047 // 1048 // 1049 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2 1050 // CHECK3-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] { 1051 // CHECK3-NEXT: entry: 1052 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 1053 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 1054 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 1055 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 1056 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 1057 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1058 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 1059 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1060 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 1061 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 1062 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1063 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 1064 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 1065 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1066 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1067 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1068 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1069 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1070 // CHECK3-NEXT: ret void 1071 // 1072 // 1073 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1074 // CHECK3-SAME: () #[[ATTR7:[0-9]+]] { 1075 // CHECK3-NEXT: entry: 1076 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1) 1077 // CHECK3-NEXT: ret void 1078 // 1079 // 1080 // CHECK4-LABEL: define {{[^@]+}}@main 1081 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] { 1082 // CHECK4-NEXT: entry: 1083 // CHECK4-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 1084 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4 1085 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4 1086 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4 1087 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1088 // CHECK4-NEXT: store i32 0, i32* [[RETVAL]], align 4 1089 // CHECK4-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1090 // CHECK4-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to i32** 1091 // CHECK4-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP1]], align 4 1092 // CHECK4-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1093 // CHECK4-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32** 1094 // CHECK4-NEXT: store i32* @_ZZ4mainE5sivar, i32** [[TMP3]], align 4 1095 // CHECK4-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 1096 // CHECK4-NEXT: store i8* null, i8** [[TMP4]], align 4 1097 // CHECK4-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1098 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1099 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2) 1100 // CHECK4-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 1101 // CHECK4-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0 1102 // CHECK4-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1103 // CHECK4: omp_offload.failed: 1104 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60(i32* @_ZZ4mainE5sivar) #[[ATTR2:[0-9]+]] 1105 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT]] 1106 // CHECK4: omp_offload.cont: 1107 // CHECK4-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiET_v() 1108 // CHECK4-NEXT: ret i32 [[CALL]] 1109 // 1110 // 1111 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l60 1112 // CHECK4-SAME: (i32* nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] { 1113 // CHECK4-NEXT: entry: 1114 // CHECK4-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 4 1115 // CHECK4-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4 1116 // CHECK4-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 4 1117 // CHECK4-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 4 1118 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1119 // CHECK4-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 4 1120 // 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_]]) 1121 // CHECK4-NEXT: ret void 1122 // 1123 // 1124 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined. 1125 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1126 // CHECK4-NEXT: entry: 1127 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1128 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1129 // CHECK4-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 4 1130 // CHECK4-NEXT: [[SIVAR:%.*]] = alloca i32, align 4 1131 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1132 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1133 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1134 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1135 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1136 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1137 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4 1138 // CHECK4-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 1139 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1140 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1141 // CHECK4-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 4 1142 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 4 1143 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 1144 // CHECK4-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 1145 // CHECK4-NEXT: store i32 0, i32* [[SIVAR]], align 4 1146 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1147 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1148 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1149 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1150 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1151 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 1152 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1153 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1154 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 1155 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1156 // CHECK4: cond.true: 1157 // CHECK4-NEXT: br label [[COND_END:%.*]] 1158 // CHECK4: cond.false: 1159 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1160 // CHECK4-NEXT: br label [[COND_END]] 1161 // CHECK4: cond.end: 1162 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 1163 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1164 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1165 // CHECK4-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 1166 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1167 // CHECK4: omp.inner.for.cond: 1168 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1169 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1170 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 1171 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1172 // CHECK4: omp.inner.for.body: 1173 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1174 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 1175 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1176 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1177 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 1178 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR]], align 4 1179 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 1180 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[SIVAR]], align 4 1181 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1182 // CHECK4: omp.body.continue: 1183 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1184 // CHECK4: omp.inner.for.inc: 1185 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1186 // CHECK4-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 1187 // CHECK4-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 1188 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]] 1189 // CHECK4: omp.inner.for.end: 1190 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1191 // CHECK4: omp.loop.exit: 1192 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 1193 // CHECK4-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 1194 // CHECK4-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR]] to i8* 1195 // CHECK4-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 4 1196 // CHECK4-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1197 // CHECK4-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 1, i32 4, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 1198 // CHECK4-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1199 // CHECK4-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1200 // CHECK4-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1201 // CHECK4-NEXT: ] 1202 // CHECK4: .omp.reduction.case1: 1203 // CHECK4-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 1204 // CHECK4-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR]], align 4 1205 // CHECK4-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 1206 // CHECK4-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 1207 // CHECK4-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1208 // CHECK4-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1209 // CHECK4: .omp.reduction.case2: 1210 // CHECK4-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR]], align 4 1211 // CHECK4-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 1212 // CHECK4-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1213 // CHECK4-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1214 // CHECK4: .omp.reduction.default: 1215 // CHECK4-NEXT: ret void 1216 // 1217 // 1218 // CHECK4-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 1219 // CHECK4-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] { 1220 // CHECK4-NEXT: entry: 1221 // CHECK4-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 1222 // CHECK4-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 1223 // CHECK4-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 1224 // CHECK4-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 1225 // CHECK4-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 1226 // CHECK4-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1227 // CHECK4-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 1228 // CHECK4-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1229 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 1230 // CHECK4-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 1231 // CHECK4-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1232 // CHECK4-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 1233 // CHECK4-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 1234 // CHECK4-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1235 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1236 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1237 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1238 // CHECK4-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1239 // CHECK4-NEXT: ret void 1240 // 1241 // 1242 // CHECK4-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 1243 // CHECK4-SAME: () #[[ATTR5:[0-9]+]] comdat { 1244 // CHECK4-NEXT: entry: 1245 // CHECK4-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 1246 // CHECK4-NEXT: [[VEC:%.*]] = alloca [2 x i32], align 4 1247 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4 1248 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4 1249 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4 1250 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1251 // CHECK4-NEXT: store i32 0, i32* [[T_VAR]], align 4 1252 // CHECK4-NEXT: [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8* 1253 // CHECK4-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP0]], i8* align 4 bitcast ([2 x i32]* @__const._Z5tmainIiET_v.vec to i8*), i32 8, i1 false) 1254 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1255 // CHECK4-NEXT: [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32** 1256 // CHECK4-NEXT: store i32* [[T_VAR]], i32** [[TMP2]], align 4 1257 // CHECK4-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1258 // CHECK4-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32** 1259 // CHECK4-NEXT: store i32* [[T_VAR]], i32** [[TMP4]], align 4 1260 // CHECK4-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 1261 // CHECK4-NEXT: store i8* null, i8** [[TMP5]], align 4 1262 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 1263 // CHECK4-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 1264 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2) 1265 // CHECK4-NEXT: [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.3, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 1266 // CHECK4-NEXT: [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0 1267 // CHECK4-NEXT: br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 1268 // CHECK4: omp_offload.failed: 1269 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(i32* [[T_VAR]]) #[[ATTR2]] 1270 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT]] 1271 // CHECK4: omp_offload.cont: 1272 // CHECK4-NEXT: ret i32 0 1273 // 1274 // 1275 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32 1276 // CHECK4-SAME: (i32* nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 1277 // CHECK4-NEXT: entry: 1278 // CHECK4-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 4 1279 // CHECK4-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 1280 // CHECK4-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4 1281 // CHECK4-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4 1282 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1283 // CHECK4-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 4 1284 // 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.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]) 1285 // CHECK4-NEXT: ret void 1286 // 1287 // 1288 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1 1289 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] { 1290 // CHECK4-NEXT: entry: 1291 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1292 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1293 // CHECK4-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 4 1294 // CHECK4-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 1295 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1296 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4 1297 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1298 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1299 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1300 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1301 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4 1302 // CHECK4-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 1303 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1304 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1305 // CHECK4-NEXT: store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 4 1306 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 4 1307 // CHECK4-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0 1308 // CHECK4-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 4 1309 // CHECK4-NEXT: store i32 0, i32* [[T_VAR]], align 4 1310 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1311 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1312 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1313 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1314 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1315 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 1316 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1317 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1318 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 1319 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1320 // CHECK4: cond.true: 1321 // CHECK4-NEXT: br label [[COND_END:%.*]] 1322 // CHECK4: cond.false: 1323 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1324 // CHECK4-NEXT: br label [[COND_END]] 1325 // CHECK4: cond.end: 1326 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 1327 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1328 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1329 // CHECK4-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 1330 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1331 // CHECK4: omp.inner.for.cond: 1332 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1333 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1334 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 1335 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1336 // CHECK4: omp.inner.for.body: 1337 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1338 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 1339 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1340 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1341 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 1342 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[T_VAR]], align 4 1343 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 1344 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[T_VAR]], align 4 1345 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1346 // CHECK4: omp.body.continue: 1347 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1348 // CHECK4: omp.inner.for.inc: 1349 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1350 // CHECK4-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], 1 1351 // CHECK4-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 1352 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]] 1353 // CHECK4: omp.inner.for.end: 1354 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1355 // CHECK4: omp.loop.exit: 1356 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 1357 // CHECK4-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 1358 // CHECK4-NEXT: [[TMP15:%.*]] = bitcast i32* [[T_VAR]] to i8* 1359 // CHECK4-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 4 1360 // CHECK4-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1361 // CHECK4-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 1, i32 4, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var) 1362 // CHECK4-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1363 // CHECK4-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1364 // CHECK4-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1365 // CHECK4-NEXT: ] 1366 // CHECK4: .omp.reduction.case1: 1367 // CHECK4-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP2]], align 4 1368 // CHECK4-NEXT: [[TMP19:%.*]] = load i32, i32* [[T_VAR]], align 4 1369 // CHECK4-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 1370 // CHECK4-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 1371 // CHECK4-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1372 // CHECK4-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1373 // CHECK4: .omp.reduction.case2: 1374 // CHECK4-NEXT: [[TMP20:%.*]] = load i32, i32* [[T_VAR]], align 4 1375 // CHECK4-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4 1376 // CHECK4-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1377 // CHECK4-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1378 // CHECK4: .omp.reduction.default: 1379 // CHECK4-NEXT: ret void 1380 // 1381 // 1382 // CHECK4-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2 1383 // CHECK4-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] { 1384 // CHECK4-NEXT: entry: 1385 // CHECK4-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 1386 // CHECK4-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 1387 // CHECK4-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 1388 // CHECK4-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 1389 // CHECK4-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 1390 // CHECK4-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1391 // CHECK4-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 1392 // CHECK4-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1393 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 1394 // CHECK4-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 1395 // CHECK4-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1396 // CHECK4-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 1397 // CHECK4-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 1398 // CHECK4-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1399 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1400 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1401 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1402 // CHECK4-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1403 // CHECK4-NEXT: ret void 1404 // 1405 // 1406 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1407 // CHECK4-SAME: () #[[ATTR7:[0-9]+]] { 1408 // CHECK4-NEXT: entry: 1409 // CHECK4-NEXT: call void @__tgt_register_requires(i64 1) 1410 // CHECK4-NEXT: ret void 1411 // 1412 // 1413 // CHECK9-LABEL: define {{[^@]+}}@main 1414 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] { 1415 // CHECK9-NEXT: entry: 1416 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 1417 // CHECK9-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1 1418 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4 1419 // CHECK9-NEXT: call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 1 dereferenceable(1) [[REF_TMP]]) 1420 // CHECK9-NEXT: ret i32 0 1421 // 1422 // 1423 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l44 1424 // CHECK9-SAME: (i32* nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2:[0-9]+]] { 1425 // CHECK9-NEXT: entry: 1426 // CHECK9-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 1427 // CHECK9-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 1428 // CHECK9-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 1429 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 1430 // CHECK9-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1431 // CHECK9-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 8 1432 // 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_]]) 1433 // CHECK9-NEXT: ret void 1434 // 1435 // 1436 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined. 1437 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] { 1438 // CHECK9-NEXT: entry: 1439 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1440 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1441 // CHECK9-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 1442 // CHECK9-NEXT: [[SIVAR:%.*]] = alloca i32, align 4 1443 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1444 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1445 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1446 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1447 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1448 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1449 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 1450 // CHECK9-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8 1451 // CHECK9-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 1452 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1453 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1454 // CHECK9-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 1455 // CHECK9-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 1456 // CHECK9-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 1457 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 1458 // CHECK9-NEXT: store i32 0, i32* [[SIVAR]], align 4 1459 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1460 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1461 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1462 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1463 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1464 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 1465 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1466 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1467 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 1468 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1469 // CHECK9: cond.true: 1470 // CHECK9-NEXT: br label [[COND_END:%.*]] 1471 // CHECK9: cond.false: 1472 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1473 // CHECK9-NEXT: br label [[COND_END]] 1474 // CHECK9: cond.end: 1475 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 1476 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1477 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1478 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 1479 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1480 // CHECK9: omp.inner.for.cond: 1481 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1482 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1483 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 1484 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1485 // CHECK9: omp.inner.for.body: 1486 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1487 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 1488 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1489 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1490 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 1491 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR]], align 4 1492 // CHECK9-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 1493 // CHECK9-NEXT: store i32 [[ADD2]], i32* [[SIVAR]], align 4 1494 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0 1495 // CHECK9-NEXT: store i32* [[SIVAR]], i32** [[TMP13]], align 8 1496 // CHECK9-NEXT: call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* nonnull align 8 dereferenceable(8) [[REF_TMP]]) 1497 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1498 // CHECK9: omp.body.continue: 1499 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1500 // CHECK9: omp.inner.for.inc: 1501 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1502 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1 1503 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 1504 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]] 1505 // CHECK9: omp.inner.for.end: 1506 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1507 // CHECK9: omp.loop.exit: 1508 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 1509 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 1510 // CHECK9-NEXT: [[TMP16:%.*]] = bitcast i32* [[SIVAR]] to i8* 1511 // CHECK9-NEXT: store i8* [[TMP16]], i8** [[TMP15]], align 8 1512 // CHECK9-NEXT: [[TMP17:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1513 // CHECK9-NEXT: [[TMP18:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP17]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 1514 // CHECK9-NEXT: switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1515 // CHECK9-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1516 // CHECK9-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1517 // CHECK9-NEXT: ] 1518 // CHECK9: .omp.reduction.case1: 1519 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP2]], align 4 1520 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR]], align 4 1521 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] 1522 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 1523 // CHECK9-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1524 // CHECK9-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1525 // CHECK9: .omp.reduction.case2: 1526 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[SIVAR]], align 4 1527 // CHECK9-NEXT: [[TMP22:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP21]] monotonic, align 4 1528 // CHECK9-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1529 // CHECK9-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1530 // CHECK9: .omp.reduction.default: 1531 // CHECK9-NEXT: ret void 1532 // 1533 // 1534 // CHECK9-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 1535 // CHECK9-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] { 1536 // CHECK9-NEXT: entry: 1537 // CHECK9-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 1538 // CHECK9-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 1539 // CHECK9-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 1540 // CHECK9-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 1541 // CHECK9-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 1542 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1543 // CHECK9-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 1544 // CHECK9-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1545 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 1546 // CHECK9-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 1547 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1548 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 1549 // CHECK9-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 1550 // CHECK9-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1551 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1552 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1553 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1554 // CHECK9-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1555 // CHECK9-NEXT: ret void 1556 // 1557 // 1558 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1559 // CHECK9-SAME: () #[[ATTR6:[0-9]+]] { 1560 // CHECK9-NEXT: entry: 1561 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1) 1562 // CHECK9-NEXT: ret void 1563 // 1564 // 1565 // CHECK10-LABEL: define {{[^@]+}}@main 1566 // CHECK10-SAME: () #[[ATTR0:[0-9]+]] { 1567 // CHECK10-NEXT: entry: 1568 // CHECK10-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 1569 // CHECK10-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1 1570 // CHECK10-NEXT: store i32 0, i32* [[RETVAL]], align 4 1571 // CHECK10-NEXT: call void @"_ZZ4mainENK3$_0clEv"(%class.anon* nonnull align 1 dereferenceable(1) [[REF_TMP]]) 1572 // CHECK10-NEXT: ret i32 0 1573 // 1574 // 1575 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l44 1576 // CHECK10-SAME: (i32* nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2:[0-9]+]] { 1577 // CHECK10-NEXT: entry: 1578 // CHECK10-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 1579 // CHECK10-NEXT: [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 1580 // CHECK10-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 1581 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 1582 // CHECK10-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0 1583 // CHECK10-NEXT: store i32* [[TMP0]], i32** [[TMP1]], align 8 1584 // 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_]]) 1585 // CHECK10-NEXT: ret void 1586 // 1587 // 1588 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined. 1589 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR2]] { 1590 // CHECK10-NEXT: entry: 1591 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1592 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1593 // CHECK10-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8 1594 // CHECK10-NEXT: [[SIVAR:%.*]] = alloca i32, align 4 1595 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1596 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4 1597 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1598 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1599 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1600 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1601 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4 1602 // CHECK10-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8 1603 // CHECK10-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 1604 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1605 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1606 // CHECK10-NEXT: store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8 1607 // CHECK10-NEXT: [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8 1608 // CHECK10-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0 1609 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8 1610 // CHECK10-NEXT: store i32 0, i32* [[SIVAR]], align 4 1611 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1612 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1613 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1614 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1615 // CHECK10-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1616 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 1617 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1618 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1619 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 1620 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1621 // CHECK10: cond.true: 1622 // CHECK10-NEXT: br label [[COND_END:%.*]] 1623 // CHECK10: cond.false: 1624 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1625 // CHECK10-NEXT: br label [[COND_END]] 1626 // CHECK10: cond.end: 1627 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 1628 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1629 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1630 // CHECK10-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 1631 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1632 // CHECK10: omp.inner.for.cond: 1633 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1634 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1635 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 1636 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1637 // CHECK10: omp.inner.for.body: 1638 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1639 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 1640 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1641 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1642 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 1643 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR]], align 4 1644 // CHECK10-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 1645 // CHECK10-NEXT: store i32 [[ADD2]], i32* [[SIVAR]], align 4 1646 // CHECK10-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0 1647 // CHECK10-NEXT: store i32* [[SIVAR]], i32** [[TMP13]], align 8 1648 // CHECK10-NEXT: call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* nonnull align 8 dereferenceable(8) [[REF_TMP]]) 1649 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1650 // CHECK10: omp.body.continue: 1651 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1652 // CHECK10: omp.inner.for.inc: 1653 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1654 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1 1655 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4 1656 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]] 1657 // CHECK10: omp.inner.for.end: 1658 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1659 // CHECK10: omp.loop.exit: 1660 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 1661 // CHECK10-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 1662 // CHECK10-NEXT: [[TMP16:%.*]] = bitcast i32* [[SIVAR]] to i8* 1663 // CHECK10-NEXT: store i8* [[TMP16]], i8** [[TMP15]], align 8 1664 // CHECK10-NEXT: [[TMP17:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1665 // CHECK10-NEXT: [[TMP18:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP17]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 1666 // CHECK10-NEXT: switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1667 // CHECK10-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1668 // CHECK10-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1669 // CHECK10-NEXT: ] 1670 // CHECK10: .omp.reduction.case1: 1671 // CHECK10-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP2]], align 4 1672 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR]], align 4 1673 // CHECK10-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] 1674 // CHECK10-NEXT: store i32 [[ADD4]], i32* [[TMP2]], align 4 1675 // CHECK10-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1676 // CHECK10-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1677 // CHECK10: .omp.reduction.case2: 1678 // CHECK10-NEXT: [[TMP21:%.*]] = load i32, i32* [[SIVAR]], align 4 1679 // CHECK10-NEXT: [[TMP22:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP21]] monotonic, align 4 1680 // CHECK10-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1681 // CHECK10-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1682 // CHECK10: .omp.reduction.default: 1683 // CHECK10-NEXT: ret void 1684 // 1685 // 1686 // CHECK10-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 1687 // CHECK10-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] { 1688 // CHECK10-NEXT: entry: 1689 // CHECK10-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 1690 // CHECK10-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 1691 // CHECK10-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 1692 // CHECK10-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 1693 // CHECK10-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 1694 // CHECK10-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1695 // CHECK10-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 1696 // CHECK10-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1697 // CHECK10-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 1698 // CHECK10-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 1699 // CHECK10-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1700 // CHECK10-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 1701 // CHECK10-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 1702 // CHECK10-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1703 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1704 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1705 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1706 // CHECK10-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1707 // CHECK10-NEXT: ret void 1708 // 1709 // 1710 // CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1711 // CHECK10-SAME: () #[[ATTR6:[0-9]+]] { 1712 // CHECK10-NEXT: entry: 1713 // CHECK10-NEXT: call void @__tgt_register_requires(i64 1) 1714 // CHECK10-NEXT: ret void 1715 // 1716