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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK1 5 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK3 8 9 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK9 19 20 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 33 #pragma omp teams distribute parallel for reduction(+: t_var) 34 for (int i = 0; i < 2; ++i) { 35 t_var += (T) i; 36 } 37 return T(); 38 } 39 40 int main() { 41 static int sivar; 42 #ifdef LAMBDA 43 44 [&]() { 45 #pragma omp target 46 #pragma omp teams distribute parallel for reduction(+: sivar) 47 for (int i = 0; i < 2; ++i) { 48 49 // Skip global and bound tid vars 50 51 52 53 // Skip global and bound tid vars, and prev lb and ub vars 54 // skip loop vars 55 56 57 sivar += i; 58 59 [&]() { 60 61 sivar += 4; 62 63 }(); 64 } 65 }(); 66 return 0; 67 #else 68 #pragma omp target 69 #pragma omp teams distribute parallel for reduction(+: sivar) 70 for (int i = 0; i < 2; ++i) { 71 sivar += i; 72 } 73 return tmain<int>(); 74 #endif 75 } 76 77 78 79 80 // Skip global and bound tid vars 81 82 83 // Skip global and bound tid vars, and prev lb and ub 84 // skip loop vars 85 86 87 88 89 // Skip global and bound tid vars 90 91 92 // Skip global and bound tid vars, and prev lb and ub vars 93 // skip loop vars 94 95 #endif 96 // CHECK1-LABEL: define {{[^@]+}}@main 97 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] { 98 // CHECK1-NEXT: entry: 99 // CHECK1-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 100 // CHECK1-NEXT: [[SIVAR_CASTED:%.*]] = alloca i64, align 8 101 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8 102 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8 103 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8 104 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 105 // CHECK1-NEXT: store i32 0, i32* [[RETVAL]], align 4 106 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* @_ZZ4mainE5sivar, align 4 107 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[SIVAR_CASTED]] to i32* 108 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4 109 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[SIVAR_CASTED]], align 8 110 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 111 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64* 112 // CHECK1-NEXT: store i64 [[TMP1]], i64* [[TMP3]], align 8 113 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 114 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64* 115 // CHECK1-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8 116 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 117 // CHECK1-NEXT: store i8* null, i8** [[TMP6]], align 8 118 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 119 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 120 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB4:[0-9]+]], i64 -1, i64 2) 121 // CHECK1-NEXT: [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB4]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], 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) 122 // CHECK1-NEXT: [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0 123 // CHECK1-NEXT: br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 124 // CHECK1: omp_offload.failed: 125 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68(i64 [[TMP1]]) #[[ATTR2:[0-9]+]] 126 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] 127 // CHECK1: omp_offload.cont: 128 // CHECK1-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v() 129 // CHECK1-NEXT: ret i32 [[CALL]] 130 // 131 // 132 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68 133 // CHECK1-SAME: (i64 noundef [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] { 134 // CHECK1-NEXT: entry: 135 // CHECK1-NEXT: [[SIVAR_ADDR:%.*]] = alloca i64, align 8 136 // CHECK1-NEXT: store i64 [[SIVAR]], i64* [[SIVAR_ADDR]], align 8 137 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[SIVAR_ADDR]] to i32* 138 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB4]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]]) 139 // CHECK1-NEXT: ret void 140 // 141 // 142 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined. 143 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] { 144 // CHECK1-NEXT: entry: 145 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 146 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 147 // CHECK1-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 148 // CHECK1-NEXT: [[SIVAR1:%.*]] = alloca i32, align 4 149 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 150 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 151 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 152 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 153 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 154 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 155 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 156 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 157 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 158 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 159 // CHECK1-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 160 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 161 // CHECK1-NEXT: store i32 0, i32* [[SIVAR1]], align 4 162 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 163 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 164 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 165 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 166 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 167 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 168 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 169 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 170 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1 171 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 172 // CHECK1: cond.true: 173 // CHECK1-NEXT: br label [[COND_END:%.*]] 174 // CHECK1: cond.false: 175 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 176 // CHECK1-NEXT: br label [[COND_END]] 177 // CHECK1: cond.end: 178 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 179 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 180 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 181 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 182 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 183 // CHECK1: omp.inner.for.cond: 184 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 185 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 186 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 187 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 188 // CHECK1: omp.inner.for.body: 189 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 190 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64 191 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 192 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64 193 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], i32* [[SIVAR1]]) 194 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 195 // CHECK1: omp.inner.for.inc: 196 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 197 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 198 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 199 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 200 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 201 // CHECK1: omp.inner.for.end: 202 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 203 // CHECK1: omp.loop.exit: 204 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 205 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 206 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR1]] to i8* 207 // CHECK1-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 208 // CHECK1-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 209 // CHECK1-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var) 210 // CHECK1-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 211 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 212 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 213 // CHECK1-NEXT: ] 214 // CHECK1: .omp.reduction.case1: 215 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP0]], align 4 216 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR1]], align 4 217 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 218 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[TMP0]], align 4 219 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], [8 x i32]* @.gomp_critical_user_.reduction.var) 220 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 221 // CHECK1: .omp.reduction.case2: 222 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR1]], align 4 223 // CHECK1-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP20]] monotonic, align 4 224 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 225 // CHECK1: .omp.reduction.default: 226 // CHECK1-NEXT: ret void 227 // 228 // 229 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1 230 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] { 231 // CHECK1-NEXT: entry: 232 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 233 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 234 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 235 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 236 // CHECK1-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 237 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 238 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 239 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 240 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 241 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 242 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 243 // CHECK1-NEXT: [[SIVAR2:%.*]] = alloca i32, align 4 244 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 245 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 246 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 247 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 248 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 249 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 250 // CHECK1-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 251 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 252 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 253 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 254 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 255 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32 256 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 257 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32 258 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 259 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 260 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 261 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 262 // CHECK1-NEXT: store i32 0, i32* [[SIVAR2]], align 4 263 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 264 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 265 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 266 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 267 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 268 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 269 // CHECK1: cond.true: 270 // CHECK1-NEXT: br label [[COND_END:%.*]] 271 // CHECK1: cond.false: 272 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 273 // CHECK1-NEXT: br label [[COND_END]] 274 // CHECK1: cond.end: 275 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 276 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 277 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 278 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 279 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 280 // CHECK1: omp.inner.for.cond: 281 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 282 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 283 // CHECK1-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 284 // CHECK1-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 285 // CHECK1: omp.inner.for.body: 286 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 287 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 288 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 289 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4 290 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 291 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR2]], align 4 292 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 293 // CHECK1-NEXT: store i32 [[ADD4]], i32* [[SIVAR2]], align 4 294 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 295 // CHECK1: omp.body.continue: 296 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 297 // CHECK1: omp.inner.for.inc: 298 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 299 // CHECK1-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP13]], 1 300 // CHECK1-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4 301 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 302 // CHECK1: omp.inner.for.end: 303 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 304 // CHECK1: omp.loop.exit: 305 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 306 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 307 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR2]] to i8* 308 // CHECK1-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 309 // CHECK1-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 310 // CHECK1-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 311 // CHECK1-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 312 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 313 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 314 // CHECK1-NEXT: ] 315 // CHECK1: .omp.reduction.case1: 316 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP0]], align 4 317 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR2]], align 4 318 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 319 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[TMP0]], align 4 320 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 321 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 322 // CHECK1: .omp.reduction.case2: 323 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR2]], align 4 324 // CHECK1-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP20]] monotonic, align 4 325 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 326 // CHECK1: .omp.reduction.default: 327 // CHECK1-NEXT: ret void 328 // 329 // 330 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 331 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] { 332 // CHECK1-NEXT: entry: 333 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 334 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 335 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 336 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 337 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 338 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 339 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 340 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 341 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 342 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 343 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 344 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 345 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 346 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 347 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 348 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 349 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 350 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 351 // CHECK1-NEXT: ret void 352 // 353 // 354 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2 355 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] { 356 // CHECK1-NEXT: entry: 357 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 358 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 359 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 360 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 361 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 362 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 363 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 364 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 365 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 366 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 367 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 368 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 369 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 370 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 371 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 372 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 373 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 374 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 375 // CHECK1-NEXT: ret void 376 // 377 // 378 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 379 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] comdat { 380 // CHECK1-NEXT: entry: 381 // CHECK1-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 382 // CHECK1-NEXT: [[VEC:%.*]] = alloca [2 x i32], align 4 383 // CHECK1-NEXT: [[T_VAR_CASTED:%.*]] = alloca i64, align 8 384 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8 385 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8 386 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8 387 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 388 // CHECK1-NEXT: store i32 0, i32* [[T_VAR]], align 4 389 // CHECK1-NEXT: [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8* 390 // 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) 391 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[T_VAR]], align 4 392 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[T_VAR_CASTED]] to i32* 393 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4 394 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[T_VAR_CASTED]], align 8 395 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 396 // CHECK1-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i64* 397 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP4]], align 8 398 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 399 // CHECK1-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64* 400 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP6]], align 8 401 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0 402 // CHECK1-NEXT: store i8* null, i8** [[TMP7]], align 8 403 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 404 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 405 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB4]], i64 -1, i64 2) 406 // CHECK1-NEXT: [[TMP10:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB4]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 407 // CHECK1-NEXT: [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0 408 // CHECK1-NEXT: br i1 [[TMP11]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 409 // CHECK1: omp_offload.failed: 410 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(i64 [[TMP2]]) #[[ATTR2]] 411 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]] 412 // CHECK1: omp_offload.cont: 413 // CHECK1-NEXT: ret i32 0 414 // 415 // 416 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32 417 // CHECK1-SAME: (i64 noundef [[T_VAR:%.*]]) #[[ATTR1]] { 418 // CHECK1-NEXT: entry: 419 // CHECK1-NEXT: [[T_VAR_ADDR:%.*]] = alloca i64, align 8 420 // CHECK1-NEXT: store i64 [[T_VAR]], i64* [[T_VAR_ADDR]], align 8 421 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[T_VAR_ADDR]] to i32* 422 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB4]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[CONV]]) 423 // CHECK1-NEXT: ret void 424 // 425 // 426 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3 427 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 428 // CHECK1-NEXT: entry: 429 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 430 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 431 // CHECK1-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 8 432 // CHECK1-NEXT: [[T_VAR1:%.*]] = alloca i32, align 4 433 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 434 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 435 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 436 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 437 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 438 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 439 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 440 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 441 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 442 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 443 // CHECK1-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8 444 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8 445 // CHECK1-NEXT: store i32 0, i32* [[T_VAR1]], align 4 446 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 447 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 448 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 449 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 450 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 451 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 452 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 453 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 454 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1 455 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 456 // CHECK1: cond.true: 457 // CHECK1-NEXT: br label [[COND_END:%.*]] 458 // CHECK1: cond.false: 459 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 460 // CHECK1-NEXT: br label [[COND_END]] 461 // CHECK1: cond.end: 462 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 463 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 464 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 465 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 466 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 467 // CHECK1: omp.inner.for.cond: 468 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 469 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 470 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 471 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 472 // CHECK1: omp.inner.for.body: 473 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 474 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64 475 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 476 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64 477 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], i32* [[T_VAR1]]) 478 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 479 // CHECK1: omp.inner.for.inc: 480 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 481 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 482 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 483 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 484 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 485 // CHECK1: omp.inner.for.end: 486 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 487 // CHECK1: omp.loop.exit: 488 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 489 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 490 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i32* [[T_VAR1]] to i8* 491 // CHECK1-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 492 // CHECK1-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 493 // CHECK1-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.6, [8 x i32]* @.gomp_critical_user_.reduction.var) 494 // CHECK1-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 495 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 496 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 497 // CHECK1-NEXT: ] 498 // CHECK1: .omp.reduction.case1: 499 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP0]], align 4 500 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[T_VAR1]], align 4 501 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 502 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[TMP0]], align 4 503 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], [8 x i32]* @.gomp_critical_user_.reduction.var) 504 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 505 // CHECK1: .omp.reduction.case2: 506 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[T_VAR1]], align 4 507 // CHECK1-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP20]] monotonic, align 4 508 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 509 // CHECK1: .omp.reduction.default: 510 // CHECK1-NEXT: ret void 511 // 512 // 513 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4 514 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 515 // CHECK1-NEXT: entry: 516 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 517 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 518 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 519 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 520 // CHECK1-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 8 521 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 522 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4 523 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 524 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 525 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 526 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 527 // CHECK1-NEXT: [[T_VAR2:%.*]] = alloca i32, align 4 528 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4 529 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 530 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 531 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 532 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 533 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 534 // CHECK1-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 8 535 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 8 536 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 537 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 538 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 539 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32 540 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 541 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32 542 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 543 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 544 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 545 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 546 // CHECK1-NEXT: store i32 0, i32* [[T_VAR2]], align 4 547 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 548 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 549 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 550 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 551 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 552 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 553 // CHECK1: cond.true: 554 // CHECK1-NEXT: br label [[COND_END:%.*]] 555 // CHECK1: cond.false: 556 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 557 // CHECK1-NEXT: br label [[COND_END]] 558 // CHECK1: cond.end: 559 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 560 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 561 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 562 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 563 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 564 // CHECK1: omp.inner.for.cond: 565 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 566 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 567 // CHECK1-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 568 // CHECK1-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 569 // CHECK1: omp.inner.for.body: 570 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 571 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 572 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 573 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4 574 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 575 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[T_VAR2]], align 4 576 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 577 // CHECK1-NEXT: store i32 [[ADD4]], i32* [[T_VAR2]], align 4 578 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 579 // CHECK1: omp.body.continue: 580 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 581 // CHECK1: omp.inner.for.inc: 582 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 583 // CHECK1-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP13]], 1 584 // CHECK1-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4 585 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]] 586 // CHECK1: omp.inner.for.end: 587 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 588 // CHECK1: omp.loop.exit: 589 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 590 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 591 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i32* [[T_VAR2]] to i8* 592 // CHECK1-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 593 // CHECK1-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 594 // CHECK1-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.5, [8 x i32]* @.gomp_critical_user_.reduction.var) 595 // CHECK1-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 596 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 597 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 598 // CHECK1-NEXT: ] 599 // CHECK1: .omp.reduction.case1: 600 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP0]], align 4 601 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[T_VAR2]], align 4 602 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 603 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[TMP0]], align 4 604 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 605 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 606 // CHECK1: .omp.reduction.case2: 607 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[T_VAR2]], align 4 608 // CHECK1-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP20]] monotonic, align 4 609 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 610 // CHECK1: .omp.reduction.default: 611 // CHECK1-NEXT: ret void 612 // 613 // 614 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.5 615 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] { 616 // CHECK1-NEXT: entry: 617 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 618 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 619 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 620 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 621 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 622 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 623 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 624 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 625 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 626 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 627 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 628 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 629 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 630 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 631 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 632 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 633 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 634 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 635 // CHECK1-NEXT: ret void 636 // 637 // 638 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.6 639 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] { 640 // CHECK1-NEXT: entry: 641 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 642 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 643 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 644 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 645 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 646 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 647 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 648 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 649 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 650 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 651 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 652 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 653 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 654 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 655 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 656 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 657 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 658 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 659 // CHECK1-NEXT: ret void 660 // 661 // 662 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 663 // CHECK1-SAME: () #[[ATTR7:[0-9]+]] { 664 // CHECK1-NEXT: entry: 665 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1) 666 // CHECK1-NEXT: ret void 667 // 668 // 669 // CHECK3-LABEL: define {{[^@]+}}@main 670 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] { 671 // CHECK3-NEXT: entry: 672 // CHECK3-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 673 // CHECK3-NEXT: [[SIVAR_CASTED:%.*]] = alloca i32, align 4 674 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4 675 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4 676 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4 677 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 678 // CHECK3-NEXT: store i32 0, i32* [[RETVAL]], align 4 679 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* @_ZZ4mainE5sivar, align 4 680 // CHECK3-NEXT: store i32 [[TMP0]], i32* [[SIVAR_CASTED]], align 4 681 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[SIVAR_CASTED]], align 4 682 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 683 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32* 684 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TMP3]], align 4 685 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 686 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32* 687 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4 688 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 689 // CHECK3-NEXT: store i8* null, i8** [[TMP6]], align 4 690 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 691 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 692 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB4:[0-9]+]], i64 -1, i64 2) 693 // CHECK3-NEXT: [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB4]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], 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) 694 // CHECK3-NEXT: [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0 695 // CHECK3-NEXT: br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 696 // CHECK3: omp_offload.failed: 697 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68(i32 [[TMP1]]) #[[ATTR2:[0-9]+]] 698 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] 699 // CHECK3: omp_offload.cont: 700 // CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v() 701 // CHECK3-NEXT: ret i32 [[CALL]] 702 // 703 // 704 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l68 705 // CHECK3-SAME: (i32 noundef [[SIVAR:%.*]]) #[[ATTR1:[0-9]+]] { 706 // CHECK3-NEXT: entry: 707 // CHECK3-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32, align 4 708 // CHECK3-NEXT: store i32 [[SIVAR]], i32* [[SIVAR_ADDR]], align 4 709 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB4]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[SIVAR_ADDR]]) 710 // CHECK3-NEXT: ret void 711 // 712 // 713 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined. 714 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] { 715 // CHECK3-NEXT: entry: 716 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 717 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 718 // CHECK3-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 4 719 // CHECK3-NEXT: [[SIVAR1:%.*]] = alloca i32, align 4 720 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 721 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 722 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 723 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 724 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 725 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 726 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 727 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 728 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 729 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 730 // CHECK3-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 4 731 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 4 732 // CHECK3-NEXT: store i32 0, i32* [[SIVAR1]], align 4 733 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 734 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 735 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 736 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 737 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 738 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 739 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 740 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 741 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1 742 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 743 // CHECK3: cond.true: 744 // CHECK3-NEXT: br label [[COND_END:%.*]] 745 // CHECK3: cond.false: 746 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 747 // CHECK3-NEXT: br label [[COND_END]] 748 // CHECK3: cond.end: 749 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 750 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 751 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 752 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 753 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 754 // CHECK3: omp.inner.for.cond: 755 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 756 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 757 // CHECK3-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 758 // CHECK3-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 759 // CHECK3: omp.inner.for.body: 760 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 761 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 762 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], i32* [[SIVAR1]]) 763 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 764 // CHECK3: omp.inner.for.inc: 765 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 766 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 767 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]] 768 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 769 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 770 // CHECK3: omp.inner.for.end: 771 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 772 // CHECK3: omp.loop.exit: 773 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 774 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 775 // CHECK3-NEXT: [[TMP13:%.*]] = bitcast i32* [[SIVAR1]] to i8* 776 // CHECK3-NEXT: store i8* [[TMP13]], i8** [[TMP12]], align 4 777 // CHECK3-NEXT: [[TMP14:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 778 // CHECK3-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]], i32 1, i32 4, i8* [[TMP14]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var) 779 // CHECK3-NEXT: switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 780 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 781 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 782 // CHECK3-NEXT: ] 783 // CHECK3: .omp.reduction.case1: 784 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[TMP0]], align 4 785 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[SIVAR1]], align 4 786 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]] 787 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[TMP0]], align 4 788 // CHECK3-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], [8 x i32]* @.gomp_critical_user_.reduction.var) 789 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 790 // CHECK3: .omp.reduction.case2: 791 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[SIVAR1]], align 4 792 // CHECK3-NEXT: [[TMP19:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP18]] monotonic, align 4 793 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 794 // CHECK3: .omp.reduction.default: 795 // CHECK3-NEXT: ret void 796 // 797 // 798 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1 799 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR1]] { 800 // CHECK3-NEXT: entry: 801 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 802 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 803 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 804 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 805 // CHECK3-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 4 806 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 807 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 808 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 809 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 810 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 811 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 812 // CHECK3-NEXT: [[SIVAR1:%.*]] = alloca i32, align 4 813 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 814 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 815 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 816 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 817 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 818 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 819 // CHECK3-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 4 820 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 4 821 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 822 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 823 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 824 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 825 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4 826 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4 827 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 828 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 829 // CHECK3-NEXT: store i32 0, i32* [[SIVAR1]], align 4 830 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 831 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 832 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 833 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 834 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 835 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 836 // CHECK3: cond.true: 837 // CHECK3-NEXT: br label [[COND_END:%.*]] 838 // CHECK3: cond.false: 839 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 840 // CHECK3-NEXT: br label [[COND_END]] 841 // CHECK3: cond.end: 842 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 843 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 844 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 845 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 846 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 847 // CHECK3: omp.inner.for.cond: 848 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 849 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 850 // CHECK3-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 851 // CHECK3-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 852 // CHECK3: omp.inner.for.body: 853 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 854 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 855 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 856 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4 857 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 858 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR1]], align 4 859 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 860 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[SIVAR1]], align 4 861 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 862 // CHECK3: omp.body.continue: 863 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 864 // CHECK3: omp.inner.for.inc: 865 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 866 // CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1 867 // CHECK3-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4 868 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 869 // CHECK3: omp.inner.for.end: 870 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 871 // CHECK3: omp.loop.exit: 872 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 873 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 874 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR1]] to i8* 875 // CHECK3-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 4 876 // CHECK3-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 877 // CHECK3-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], i32 1, i32 4, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 878 // CHECK3-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 879 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 880 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 881 // CHECK3-NEXT: ] 882 // CHECK3: .omp.reduction.case1: 883 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP0]], align 4 884 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR1]], align 4 885 // CHECK3-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 886 // CHECK3-NEXT: store i32 [[ADD5]], i32* [[TMP0]], align 4 887 // CHECK3-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 888 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 889 // CHECK3: .omp.reduction.case2: 890 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR1]], align 4 891 // CHECK3-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP20]] monotonic, align 4 892 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 893 // CHECK3: .omp.reduction.default: 894 // CHECK3-NEXT: ret void 895 // 896 // 897 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 898 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] { 899 // CHECK3-NEXT: entry: 900 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 901 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 902 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 903 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 904 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 905 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 906 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 907 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 908 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 909 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 910 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 911 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 912 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 913 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 914 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 915 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 916 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 917 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 918 // CHECK3-NEXT: ret void 919 // 920 // 921 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2 922 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] { 923 // CHECK3-NEXT: entry: 924 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 925 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 926 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 927 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 928 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 929 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 930 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 931 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 932 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 933 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 934 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 935 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 936 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 937 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 938 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 939 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 940 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 941 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 942 // CHECK3-NEXT: ret void 943 // 944 // 945 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v 946 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] comdat { 947 // CHECK3-NEXT: entry: 948 // CHECK3-NEXT: [[T_VAR:%.*]] = alloca i32, align 4 949 // CHECK3-NEXT: [[VEC:%.*]] = alloca [2 x i32], align 4 950 // CHECK3-NEXT: [[T_VAR_CASTED:%.*]] = alloca i32, align 4 951 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4 952 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4 953 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4 954 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 955 // CHECK3-NEXT: store i32 0, i32* [[T_VAR]], align 4 956 // CHECK3-NEXT: [[TMP0:%.*]] = bitcast [2 x i32]* [[VEC]] to i8* 957 // 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) 958 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[T_VAR]], align 4 959 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[T_VAR_CASTED]], align 4 960 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[T_VAR_CASTED]], align 4 961 // CHECK3-NEXT: [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 962 // CHECK3-NEXT: [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32* 963 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP4]], align 4 964 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 965 // CHECK3-NEXT: [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32* 966 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP6]], align 4 967 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0 968 // CHECK3-NEXT: store i8* null, i8** [[TMP7]], align 4 969 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0 970 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0 971 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB4]], i64 -1, i64 2) 972 // CHECK3-NEXT: [[TMP10:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB4]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32.region_id, i32 1, i8** [[TMP8]], i8** [[TMP9]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0) 973 // CHECK3-NEXT: [[TMP11:%.*]] = icmp ne i32 [[TMP10]], 0 974 // CHECK3-NEXT: br i1 [[TMP11]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]] 975 // CHECK3: omp_offload.failed: 976 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32(i32 [[TMP2]]) #[[ATTR2]] 977 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]] 978 // CHECK3: omp_offload.cont: 979 // CHECK3-NEXT: ret i32 0 980 // 981 // 982 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l32 983 // CHECK3-SAME: (i32 noundef [[T_VAR:%.*]]) #[[ATTR1]] { 984 // CHECK3-NEXT: entry: 985 // CHECK3-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32, align 4 986 // CHECK3-NEXT: store i32 [[T_VAR]], i32* [[T_VAR_ADDR]], align 4 987 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB4]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32* [[T_VAR_ADDR]]) 988 // CHECK3-NEXT: ret void 989 // 990 // 991 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3 992 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 993 // CHECK3-NEXT: entry: 994 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 995 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 996 // CHECK3-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 4 997 // CHECK3-NEXT: [[T_VAR1:%.*]] = alloca i32, align 4 998 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 999 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1000 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1001 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1002 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1003 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1004 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1005 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 1006 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1007 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1008 // CHECK3-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4 1009 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4 1010 // CHECK3-NEXT: store i32 0, i32* [[T_VAR1]], align 4 1011 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1012 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 1013 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1014 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1015 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1016 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 1017 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1018 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1019 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1 1020 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1021 // CHECK3: cond.true: 1022 // CHECK3-NEXT: br label [[COND_END:%.*]] 1023 // CHECK3: cond.false: 1024 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1025 // CHECK3-NEXT: br label [[COND_END]] 1026 // CHECK3: cond.end: 1027 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 1028 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1029 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1030 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 1031 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1032 // CHECK3: omp.inner.for.cond: 1033 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1034 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1035 // CHECK3-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 1036 // CHECK3-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1037 // CHECK3: omp.inner.for.body: 1038 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1039 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1040 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], i32* [[T_VAR1]]) 1041 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1042 // CHECK3: omp.inner.for.inc: 1043 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1044 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1045 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]] 1046 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1047 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1048 // CHECK3: omp.inner.for.end: 1049 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1050 // CHECK3: omp.loop.exit: 1051 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 1052 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 1053 // CHECK3-NEXT: [[TMP13:%.*]] = bitcast i32* [[T_VAR1]] to i8* 1054 // CHECK3-NEXT: store i8* [[TMP13]], i8** [[TMP12]], align 4 1055 // CHECK3-NEXT: [[TMP14:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1056 // CHECK3-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], i32 1, i32 4, i8* [[TMP14]], void (i8*, i8*)* @.omp.reduction.reduction_func.6, [8 x i32]* @.gomp_critical_user_.reduction.var) 1057 // CHECK3-NEXT: switch i32 [[TMP15]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1058 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1059 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1060 // CHECK3-NEXT: ] 1061 // CHECK3: .omp.reduction.case1: 1062 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[TMP0]], align 4 1063 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[T_VAR1]], align 4 1064 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]] 1065 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[TMP0]], align 4 1066 // CHECK3-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1067 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1068 // CHECK3: .omp.reduction.case2: 1069 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[T_VAR1]], align 4 1070 // CHECK3-NEXT: [[TMP19:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP18]] monotonic, align 4 1071 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1072 // CHECK3: .omp.reduction.default: 1073 // CHECK3-NEXT: ret void 1074 // 1075 // 1076 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4 1077 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]]) #[[ATTR1]] { 1078 // CHECK3-NEXT: entry: 1079 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4 1080 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4 1081 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4 1082 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4 1083 // CHECK3-NEXT: [[T_VAR_ADDR:%.*]] = alloca i32*, align 4 1084 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1085 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4 1086 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1087 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1088 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1089 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1090 // CHECK3-NEXT: [[T_VAR1:%.*]] = alloca i32, align 4 1091 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4 1092 // CHECK3-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 4 1093 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4 1094 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4 1095 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1096 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1097 // CHECK3-NEXT: store i32* [[T_VAR]], i32** [[T_VAR_ADDR]], align 4 1098 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[T_VAR_ADDR]], align 4 1099 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1100 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1101 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4 1102 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4 1103 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4 1104 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4 1105 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1106 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1107 // CHECK3-NEXT: store i32 0, i32* [[T_VAR1]], align 4 1108 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4 1109 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 1110 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1111 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1112 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 1113 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1114 // CHECK3: cond.true: 1115 // CHECK3-NEXT: br label [[COND_END:%.*]] 1116 // CHECK3: cond.false: 1117 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1118 // CHECK3-NEXT: br label [[COND_END]] 1119 // CHECK3: cond.end: 1120 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 1121 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1122 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1123 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 1124 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1125 // CHECK3: omp.inner.for.cond: 1126 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1127 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1128 // CHECK3-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 1129 // CHECK3-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1130 // CHECK3: omp.inner.for.body: 1131 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1132 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 1133 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1134 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1135 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 1136 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[T_VAR1]], align 4 1137 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 1138 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[T_VAR1]], align 4 1139 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1140 // CHECK3: omp.body.continue: 1141 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1142 // CHECK3: omp.inner.for.inc: 1143 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1144 // CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP13]], 1 1145 // CHECK3-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4 1146 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]] 1147 // CHECK3: omp.inner.for.end: 1148 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1149 // CHECK3: omp.loop.exit: 1150 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 1151 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i32 0, i32 0 1152 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i32* [[T_VAR1]] to i8* 1153 // CHECK3-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 4 1154 // CHECK3-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1155 // CHECK3-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], i32 1, i32 4, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.5, [8 x i32]* @.gomp_critical_user_.reduction.var) 1156 // CHECK3-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1157 // CHECK3-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1158 // CHECK3-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1159 // CHECK3-NEXT: ] 1160 // CHECK3: .omp.reduction.case1: 1161 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP0]], align 4 1162 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[T_VAR1]], align 4 1163 // CHECK3-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 1164 // CHECK3-NEXT: store i32 [[ADD5]], i32* [[TMP0]], align 4 1165 // CHECK3-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1166 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1167 // CHECK3: .omp.reduction.case2: 1168 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[T_VAR1]], align 4 1169 // CHECK3-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP20]] monotonic, align 4 1170 // CHECK3-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1171 // CHECK3: .omp.reduction.default: 1172 // CHECK3-NEXT: ret void 1173 // 1174 // 1175 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.5 1176 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] { 1177 // CHECK3-NEXT: entry: 1178 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 1179 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 1180 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 1181 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 1182 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 1183 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1184 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 1185 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1186 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 1187 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 1188 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1189 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 1190 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 1191 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1192 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1193 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1194 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1195 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1196 // CHECK3-NEXT: ret void 1197 // 1198 // 1199 // CHECK3-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.6 1200 // CHECK3-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] { 1201 // CHECK3-NEXT: entry: 1202 // CHECK3-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 4 1203 // CHECK3-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 4 1204 // CHECK3-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 4 1205 // CHECK3-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 4 1206 // CHECK3-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 4 1207 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1208 // CHECK3-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 4 1209 // CHECK3-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1210 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i32 0, i32 0 1211 // CHECK3-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4 1212 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1213 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i32 0, i32 0 1214 // CHECK3-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 4 1215 // CHECK3-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1216 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1217 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1218 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1219 // CHECK3-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1220 // CHECK3-NEXT: ret void 1221 // 1222 // 1223 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1224 // CHECK3-SAME: () #[[ATTR7:[0-9]+]] { 1225 // CHECK3-NEXT: entry: 1226 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1) 1227 // CHECK3-NEXT: ret void 1228 // 1229 // 1230 // CHECK9-LABEL: define {{[^@]+}}@main 1231 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] { 1232 // CHECK9-NEXT: entry: 1233 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4 1234 // CHECK9-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 1 1235 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4 1236 // CHECK9-NEXT: call void @"_ZZ4mainENK3$_0clEv"(%class.anon* noundef nonnull align 1 dereferenceable(1) [[REF_TMP]]) 1237 // CHECK9-NEXT: ret i32 0 1238 // 1239 // 1240 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l45 1241 // CHECK9-SAME: (i64 noundef [[SIVAR:%.*]]) #[[ATTR2:[0-9]+]] { 1242 // CHECK9-NEXT: entry: 1243 // CHECK9-NEXT: [[SIVAR_ADDR:%.*]] = alloca i64, align 8 1244 // CHECK9-NEXT: store i64 [[SIVAR]], i64* [[SIVAR_ADDR]], align 8 1245 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[SIVAR_ADDR]] to i32* 1246 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]]) 1247 // CHECK9-NEXT: ret void 1248 // 1249 // 1250 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined. 1251 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] { 1252 // CHECK9-NEXT: entry: 1253 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1254 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1255 // CHECK9-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 1256 // CHECK9-NEXT: [[SIVAR1:%.*]] = alloca i32, align 4 1257 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1258 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1259 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4 1260 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4 1261 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1262 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1263 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 1264 // CHECK9-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 1265 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1266 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1267 // CHECK9-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 1268 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 1269 // CHECK9-NEXT: store i32 0, i32* [[SIVAR1]], align 4 1270 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4 1271 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_COMB_UB]], align 4 1272 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1273 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1274 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1275 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4 1276 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1277 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1278 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 1 1279 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1280 // CHECK9: cond.true: 1281 // CHECK9-NEXT: br label [[COND_END:%.*]] 1282 // CHECK9: cond.false: 1283 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1284 // CHECK9-NEXT: br label [[COND_END]] 1285 // CHECK9: cond.end: 1286 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ] 1287 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4 1288 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1289 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4 1290 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1291 // CHECK9: omp.inner.for.cond: 1292 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1293 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1294 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]] 1295 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1296 // CHECK9: omp.inner.for.body: 1297 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4 1298 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64 1299 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4 1300 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64 1301 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], i32* [[SIVAR1]]) 1302 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1303 // CHECK9: omp.inner.for.inc: 1304 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1305 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4 1306 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1307 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4 1308 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]] 1309 // CHECK9: omp.inner.for.end: 1310 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1311 // CHECK9: omp.loop.exit: 1312 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]]) 1313 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 1314 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i32* [[SIVAR1]] to i8* 1315 // CHECK9-NEXT: store i8* [[TMP15]], i8** [[TMP14]], align 8 1316 // CHECK9-NEXT: [[TMP16:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1317 // CHECK9-NEXT: [[TMP17:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]], i32 1, i64 8, i8* [[TMP16]], void (i8*, i8*)* @.omp.reduction.reduction_func.2, [8 x i32]* @.gomp_critical_user_.reduction.var) 1318 // CHECK9-NEXT: switch i32 [[TMP17]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1319 // CHECK9-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1320 // CHECK9-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1321 // CHECK9-NEXT: ] 1322 // CHECK9: .omp.reduction.case1: 1323 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[TMP0]], align 4 1324 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[SIVAR1]], align 4 1325 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP18]], [[TMP19]] 1326 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[TMP0]], align 4 1327 // CHECK9-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1328 // CHECK9-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1329 // CHECK9: .omp.reduction.case2: 1330 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR1]], align 4 1331 // CHECK9-NEXT: [[TMP21:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP20]] monotonic, align 4 1332 // CHECK9-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1333 // CHECK9: .omp.reduction.default: 1334 // CHECK9-NEXT: ret void 1335 // 1336 // 1337 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1 1338 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[SIVAR:%.*]]) #[[ATTR2]] { 1339 // CHECK9-NEXT: entry: 1340 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8 1341 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8 1342 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8 1343 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8 1344 // CHECK9-NEXT: [[SIVAR_ADDR:%.*]] = alloca i32*, align 8 1345 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4 1346 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4 1347 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4 1348 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4 1349 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4 1350 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4 1351 // CHECK9-NEXT: [[SIVAR2:%.*]] = alloca i32, align 4 1352 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4 1353 // CHECK9-NEXT: [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8 1354 // CHECK9-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8 1355 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8 1356 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8 1357 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8 1358 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8 1359 // CHECK9-NEXT: store i32* [[SIVAR]], i32** [[SIVAR_ADDR]], align 8 1360 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[SIVAR_ADDR]], align 8 1361 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4 1362 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_UB]], align 4 1363 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8 1364 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32 1365 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8 1366 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32 1367 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4 1368 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4 1369 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4 1370 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4 1371 // CHECK9-NEXT: store i32 0, i32* [[SIVAR2]], align 4 1372 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8 1373 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4 1374 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1) 1375 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1376 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 1 1377 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 1378 // CHECK9: cond.true: 1379 // CHECK9-NEXT: br label [[COND_END:%.*]] 1380 // CHECK9: cond.false: 1381 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1382 // CHECK9-NEXT: br label [[COND_END]] 1383 // CHECK9: cond.end: 1384 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ] 1385 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4 1386 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4 1387 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4 1388 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]] 1389 // CHECK9: omp.inner.for.cond: 1390 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1391 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4 1392 // CHECK9-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]] 1393 // CHECK9-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]] 1394 // CHECK9: omp.inner.for.body: 1395 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1396 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1 1397 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]] 1398 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4 1399 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4 1400 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[SIVAR2]], align 4 1401 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP12]], [[TMP11]] 1402 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[SIVAR2]], align 4 1403 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0 1404 // CHECK9-NEXT: store i32* [[SIVAR2]], i32** [[TMP13]], align 8 1405 // CHECK9-NEXT: call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* noundef nonnull align 8 dereferenceable(8) [[REF_TMP]]) 1406 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]] 1407 // CHECK9: omp.body.continue: 1408 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]] 1409 // CHECK9: omp.inner.for.inc: 1410 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4 1411 // CHECK9-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP14]], 1 1412 // CHECK9-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4 1413 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]] 1414 // CHECK9: omp.inner.for.end: 1415 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]] 1416 // CHECK9: omp.loop.exit: 1417 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]]) 1418 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0 1419 // CHECK9-NEXT: [[TMP16:%.*]] = bitcast i32* [[SIVAR2]] to i8* 1420 // CHECK9-NEXT: store i8* [[TMP16]], i8** [[TMP15]], align 8 1421 // CHECK9-NEXT: [[TMP17:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8* 1422 // CHECK9-NEXT: [[TMP18:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], i32 1, i64 8, i8* [[TMP17]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var) 1423 // CHECK9-NEXT: switch i32 [[TMP18]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [ 1424 // CHECK9-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]] 1425 // CHECK9-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]] 1426 // CHECK9-NEXT: ] 1427 // CHECK9: .omp.reduction.case1: 1428 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP0]], align 4 1429 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[SIVAR2]], align 4 1430 // CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] 1431 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[TMP0]], align 4 1432 // CHECK9-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_.reduction.var) 1433 // CHECK9-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1434 // CHECK9: .omp.reduction.case2: 1435 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[SIVAR2]], align 4 1436 // CHECK9-NEXT: [[TMP22:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP21]] monotonic, align 4 1437 // CHECK9-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]] 1438 // CHECK9: .omp.reduction.default: 1439 // CHECK9-NEXT: ret void 1440 // 1441 // 1442 // CHECK9-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func 1443 // CHECK9-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR4:[0-9]+]] { 1444 // CHECK9-NEXT: entry: 1445 // CHECK9-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 1446 // CHECK9-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 1447 // CHECK9-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 1448 // CHECK9-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 1449 // CHECK9-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 1450 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1451 // CHECK9-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 1452 // CHECK9-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1453 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 1454 // CHECK9-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 1455 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1456 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 1457 // CHECK9-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 1458 // CHECK9-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1459 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1460 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1461 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1462 // CHECK9-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1463 // CHECK9-NEXT: ret void 1464 // 1465 // 1466 // CHECK9-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.2 1467 // CHECK9-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR4]] { 1468 // CHECK9-NEXT: entry: 1469 // CHECK9-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8 1470 // CHECK9-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8 1471 // CHECK9-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8 1472 // CHECK9-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8 1473 // CHECK9-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8 1474 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]* 1475 // CHECK9-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8 1476 // CHECK9-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]* 1477 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0 1478 // CHECK9-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8 1479 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32* 1480 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0 1481 // CHECK9-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8 1482 // CHECK9-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32* 1483 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4 1484 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4 1485 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]] 1486 // CHECK9-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4 1487 // CHECK9-NEXT: ret void 1488 // 1489 // 1490 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg 1491 // CHECK9-SAME: () #[[ATTR6:[0-9]+]] { 1492 // CHECK9-NEXT: entry: 1493 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1) 1494 // CHECK9-NEXT: ret void 1495 // 1496