1 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck %s 2 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s 3 // RUN: %clang_cc1 -fopenmp -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s 4 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s 5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s 6 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DARRAY -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=ARRAY %s 7 // expected-no-diagnostics 8 #ifndef ARRAY 9 #ifndef HEADER 10 #define HEADER 11 12 struct St { 13 int a, b; 14 St() : a(0), b(0) {} 15 St(const St &st) : a(st.a + st.b), b(0) {} 16 ~St() {} 17 }; 18 19 volatile int g __attribute__((aligned(128))) = 1212; 20 21 struct SS { 22 int a; 23 int b : 4; 24 int &c; 25 int e[4]; 26 SS(int &d) : a(0), b(0), c(d) { 27 #pragma omp parallel firstprivate(a, b, c, e) 28 #ifdef LAMBDA 29 [&]() { 30 ++this->a, --b, (this)->c /= 1; 31 #pragma omp parallel firstprivate(a, b, c) 32 ++(this)->a, --b, this->c /= 1; 33 }(); 34 #elif defined(BLOCKS) 35 ^{ 36 ++a; 37 --this->b; 38 (this)->c /= 1; 39 #pragma omp parallel firstprivate(a, b, c) 40 ++(this)->a, --b, this->c /= 1; 41 }(); 42 #else 43 ++this->a, --b, c /= 1, e[2] = 1111; 44 #endif 45 } 46 }; 47 48 template<typename T> 49 struct SST { 50 T a; 51 SST() : a(T()) { 52 #pragma omp parallel firstprivate(a) 53 #ifdef LAMBDA 54 [&]() { 55 [&]() { 56 ++this->a; 57 #pragma omp parallel firstprivate(a) 58 ++(this)->a; 59 }(); 60 }(); 61 #elif defined(BLOCKS) 62 ^{ 63 ^{ 64 ++a; 65 #pragma omp parallel firstprivate(a) 66 ++(this)->a; 67 }(); 68 }(); 69 #else 70 ++(this)->a; 71 #endif 72 } 73 }; 74 75 template <class T> 76 struct S { 77 T f; 78 S(T a) : f(a + g) {} 79 S() : f(g) {} 80 S(const S &s, St t = St()) : f(s.f + t.a) {} 81 operator T() { return T(); } 82 ~S() {} 83 }; 84 85 // CHECK: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8 86 // LAMBDA: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8 87 // BLOCKS: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8 88 // CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } 89 // CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } 90 // CHECK-DAG: [[ST_TY:%.+]] = type { i{{[0-9]+}}, i{{[0-9]+}} } 91 92 template <typename T> 93 T tmain() { 94 S<T> test; 95 SST<T> sst; 96 T t_var __attribute__((aligned(128))) = T(); 97 T vec[] __attribute__((aligned(128))) = {1, 2}; 98 S<T> s_arr[] __attribute__((aligned(128))) = {1, 2}; 99 S<T> var __attribute__((aligned(128))) (3); 100 #pragma omp parallel firstprivate(t_var, vec, s_arr, var) 101 { 102 vec[0] = t_var; 103 s_arr[0] = var; 104 } 105 #pragma omp parallel firstprivate(t_var) 106 {} 107 return T(); 108 } 109 110 int main() { 111 static int sivar; 112 SS ss(sivar); 113 #ifdef LAMBDA 114 // LAMBDA: [[G:@.+]] = global i{{[0-9]+}} 1212, 115 // LAMBDA-LABEL: @main 116 // LAMBDA: alloca [[SS_TY]], 117 // LAMBDA: alloca [[CAP_TY:%.+]], 118 // LAMBDA: call{{.*}} void [[OUTER_LAMBDA:@[^(]+]]([[CAP_TY]]* 119 [&]() { 120 // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( 121 // LAMBDA: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 2, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i32* [[G]], {{.+}}) 122 #pragma omp parallel firstprivate(g, sivar) 123 { 124 // LAMBDA: define {{.+}} @{{.+}}([[SS_TY]]* 125 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 126 // LAMBDA: store i{{[0-9]+}} 0, i{{[0-9]+}}* % 127 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 128 // LAMBDA: store i8 129 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 130 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 131 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 132 // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 133 // LAMBDA: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 5, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[SS_TY]]*, i32, i32, i32, [4 x i{{[0-9]+}}]*)* [[SS_MICROTASK:@.+]] to void 134 // LAMBDA: ret 135 136 // LAMBDA: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32 {{.+}}, i32 {{.+}}, i32 {{.+}}, [4 x i{{[0-9]+}}]* {{.+}}) 137 // LAMBDA-NOT: getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* % 138 // LAMBDA: call{{.*}} void 139 // LAMBDA: ret void 140 141 // LAMBDA: define internal void @{{.+}}(i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32 {{.+}}, i32 {{.+}}, i32 {{.+}}) 142 // LAMBDA: [[A_PRIV:%.+]] = alloca i{{[0-9]+}}, 143 // LAMBDA: [[B_PRIV:%.+]] = alloca i{{[0-9]+}}, 144 // LAMBDA: [[C_PRIV:%.+]] = alloca i{{[0-9]+}}, 145 // LAMBDA: store i{{[0-9]+}}* [[A_PRIV]], i{{[0-9]+}}** [[REFA:%.+]], 146 // LAMBDA: store i{{[0-9]+}}* [[C_PRIV]], i{{[0-9]+}}** [[REFC:%.+]], 147 // LAMBDA-NEXT: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]], 148 // LAMBDA-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]], 149 // LAMBDA-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1 150 // LAMBDA-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]], 151 // LAMBDA-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]], 152 // LAMBDA-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1 153 // LAMBDA-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]], 154 // LAMBDA-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]], 155 // LAMBDA-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]], 156 // LAMBDA-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1 157 // LAMBDA-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]], 158 // LAMBDA-NEXT: ret void 159 160 // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, i32 {{.*}}%{{.+}}) 161 // LAMBDA: [[SIVAR_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, 162 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128 163 // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]] 164 // LAMBDA: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G_REF]], align 128 165 // LAMBDA: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128 166 // LAMBDA-NOT: call {{.*}}void @__kmpc_barrier( 167 g = 1; 168 sivar = 2; 169 // LAMBDA: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], 170 // LAMBDA: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]], 171 // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 172 // LAMBDA: store i{{[0-9]+}}* [[G_PRIVATE_ADDR]], i{{[0-9]+}}** [[G_PRIVATE_ADDR_REF]] 173 // LAMBDA: [[SIVAR_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 174 // LAMBDA: store i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]], i{{[0-9]+}}** [[SIVAR_PRIVATE_ADDR_REF]] 175 // LAMBDA: call{{.*}} void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) 176 [&]() { 177 // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) 178 // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], 179 g = 2; 180 sivar = 4; 181 // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] 182 // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 183 // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_PTR_REF]] 184 // LAMBDA: [[SIVAR_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 1 185 // LAMBDA: [[SIVAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[SIVAR_PTR_REF]] 186 // LAMBDA: store i{{[0-9]+}} 4, i{{[0-9]+}}* [[SIVAR_REF]] 187 }(); 188 } 189 }(); 190 return 0; 191 #elif defined(BLOCKS) 192 // BLOCKS: [[G:@.+]] = global i{{[0-9]+}} 1212, 193 // BLOCKS-LABEL: @main 194 // BLOCKS: call 195 // BLOCKS: call {{.*}}void {{%.+}}(i8 196 ^{ 197 // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* 198 // BLOCKS: call {{.*}}void {{.+}} @__kmpc_fork_call({{.+}}, i32 2, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i32* [[G]], {{.+}}) 199 #pragma omp parallel firstprivate(g, sivar) 200 { 201 // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}}, i32 {{.*}}%{{.+}}) 202 // BLOCKS: [[SIVAR_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, 203 // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128 204 // BLOCKS: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]] 205 // BLOCKS: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G_REF]], align 128 206 // BLOCKS: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128 207 // BLOCKS-NOT: call {{.*}}void @__kmpc_barrier( 208 g = 1; 209 sivar = 2; 210 // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], 211 // BLOCKS: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]], 212 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 213 // BLOCKS: i{{[0-9]+}}* [[G_PRIVATE_ADDR]] 214 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 215 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 216 // BLOCKS: i{{[0-9]+}}* [[SIVAR_PRIVATE_ADDR]] 217 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 218 // BLOCKS: call {{.*}}void {{%.+}}(i8 219 ^{ 220 // BLOCKS: define {{.+}} void {{@.+}}(i8* 221 g = 2; 222 sivar = 4; 223 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 224 // BLOCKS: store i{{[0-9]+}} 2, i{{[0-9]+}}* 225 // BLOCKS-NOT: [[G]]{{[[^:word:]]}} 226 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 227 // BLOCKS: store i{{[0-9]+}} 4, i{{[0-9]+}}* 228 // BLOCKS-NOT: [[SIVAR]]{{[[^:word:]]}} 229 // BLOCKS: ret 230 }(); 231 } 232 }(); 233 return 0; 234 // BLOCKS: define {{.+}} @{{.+}}([[SS_TY]]* 235 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 236 // BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* % 237 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 238 // BLOCKS: store i8 239 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 240 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 241 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 242 // BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 243 // BLOCKS: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 5, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[SS_TY]]*, i32, i32, i32, [4 x i{{[0-9]+}}]*)* [[SS_MICROTASK:@.+]] to void 244 // BLOCKS: ret 245 246 // BLOCKS: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32 {{.+}}, i32 {{.+}}, i32 {{.+}}, [4 x i{{[0-9]+}}]* {{.+}}) 247 // BLOCKS-NOT: getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* % 248 // BLOCKS: call{{.*}} void 249 // BLOCKS: ret void 250 251 // BLOCKS: define internal void @{{.+}}(i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32 {{.+}}, i32 {{.+}}, i32 {{.+}}) 252 // BLOCKS: [[A_PRIV:%.+]] = alloca i{{[0-9]+}}, 253 // BLOCKS: [[B_PRIV:%.+]] = alloca i{{[0-9]+}}, 254 // BLOCKS: [[C_PRIV:%.+]] = alloca i{{[0-9]+}}, 255 // BLOCKS: store i{{[0-9]+}}* [[A_PRIV]], i{{[0-9]+}}** [[REFA:%.+]], 256 // BLOCKS: store i{{[0-9]+}}* [[C_PRIV]], i{{[0-9]+}}** [[REFC:%.+]], 257 // BLOCKS-NEXT: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]], 258 // BLOCKS-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]], 259 // BLOCKS-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1 260 // BLOCKS-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]], 261 // BLOCKS-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]], 262 // BLOCKS-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1 263 // BLOCKS-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]], 264 // BLOCKS-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]], 265 // BLOCKS-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]], 266 // BLOCKS-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1 267 // BLOCKS-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]], 268 // BLOCKS-NEXT: ret void 269 #else 270 S<float> test; 271 int t_var = 0; 272 int vec[] = {1, 2}; 273 S<float> s_arr[] = {1, 2}; 274 S<float> var(3); 275 #pragma omp parallel firstprivate(t_var, vec, s_arr, var, sivar) 276 { 277 vec[0] = t_var; 278 s_arr[0] = var; 279 sivar = 2; 280 } 281 #pragma omp parallel firstprivate(t_var) 282 {} 283 return tmain<int>(); 284 #endif 285 } 286 287 // CHECK: define {{.*}}i{{[0-9]+}} @main() 288 // CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], 289 // CHECK: call {{.*}} [[S_FLOAT_TY_DEF_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) 290 // CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 5, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x i32]*, i32, [2 x [[S_FLOAT_TY]]]*, [[S_FLOAT_TY]]*, i{{[0-9]+}})* [[MAIN_MICROTASK:@.+]] to void 291 // CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]() 292 // CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* 293 // CHECK: ret 294 // 295 // CHECK: define internal {{.*}}void [[MAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, i32 {{.*}}%{{.+}}, [2 x [[S_FLOAT_TY]]]* dereferenceable(8) %{{.+}}, [[S_FLOAT_TY]]* dereferenceable(4) %{{.+}}, i32 {{.*}}[[SIVAR:%.+]]) 296 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, 297 // CHECK: [[SIVAR7_PRIV:%.+]] = alloca i{{[0-9]+}}, 298 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], 299 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]], 300 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]], 301 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 302 303 // CHECK: [[VEC_REF:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** % 304 // CHECK-NOT: load i{{[0-9]+}}*, i{{[0-9]+}}** % 305 // CHECK: [[S_ARR_REF:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** % 306 // CHECK: [[VAR_REF:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** % 307 // CHECK-NOT: load i{{[0-9]+}}*, i{{[0-9]+}}** % 308 // CHECK: [[VEC_DEST:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8* 309 // CHECK: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_REF]] to i8* 310 // CHECK: call void @llvm.memcpy.{{.+}}(i8* [[VEC_DEST]], i8* [[VEC_SRC]], 311 // CHECK: [[S_ARR_PRIV_BEGIN:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[S_ARR_PRIV]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 312 // CHECK: [[S_ARR_BEGIN:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[S_ARR_REF]] to [[S_FLOAT_TY]]* 313 // CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2 314 // CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]] 315 // CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]] 316 // CHECK: [[S_ARR_BODY]] 317 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR:@.+]]([[ST_TY]]* [[ST_TY_TEMP:%.+]]) 318 // CHECK: call {{.*}} [[S_FLOAT_TY_COPY_CONSTR:@.+]]([[S_FLOAT_TY]]* {{.+}}, [[S_FLOAT_TY]]* {{.+}}, [[ST_TY]]* [[ST_TY_TEMP]]) 319 // CHECK: call {{.*}} [[ST_TY_DESTR:@.+]]([[ST_TY]]* [[ST_TY_TEMP]]) 320 // CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]] 321 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* [[ST_TY_TEMP:%.+]]) 322 // CHECK: call {{.*}} [[S_FLOAT_TY_COPY_CONSTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]], [[S_FLOAT_TY]]* {{.*}} [[VAR_REF]], [[ST_TY]]* [[ST_TY_TEMP]]) 323 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* [[ST_TY_TEMP]]) 324 325 // CHECK: store i{{[0-9]+}} 2, i{{[0-9]+}}* [[SIVAR7_PRIV]], 326 327 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) 328 // CHECK-DAG: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* 329 // CHECK: ret void 330 // CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() 331 // CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], 332 // CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) 333 // CHECK: call {{.*}}void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x i32]*, i32*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]*)* [[TMAIN_MICROTASK:@.+]] to void 334 // CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* 335 // CHECK: ret 336 // 337 // CHECK: define {{.+}} @{{.+}}([[SS_TY]]* 338 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 339 // CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* % 340 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 341 // CHECK: store i8 342 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 343 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0 344 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1 345 // CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2 346 // CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 5, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[SS_TY]]*, i32, i32, i32, [4 x i32]*)* [[SS_MICROTASK:@.+]] to void 347 // CHECK: ret 348 349 // CHECK: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32 {{.+}}, i32 {{.+}}, i32 {{.+}}, [4 x i{{[0-9]+}}]* {{.+}}) 350 // CHECK: [[A_PRIV:%.+]] = alloca i{{[0-9]+}}, 351 // CHECK: [[B_PRIV:%.+]] = alloca i{{[0-9]+}}, 352 // CHECK: [[C_PRIV:%.+]] = alloca i{{[0-9]+}}, 353 // CHECK: [[E_PRIV:%.+]] = alloca [4 x i{{[0-9]+}}], 354 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[A_PRIV]] 355 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[B_PRIV]] 356 // CHECK: store i{{[0-9]+}} {{.+}}, i{{[0-9]+}}* [[C_PRIV]] 357 // CHECK: store i{{[0-9]+}}* [[A_PRIV]], i{{[0-9]+}}** [[REFA:%.+]], 358 // CHECK: store i{{[0-9]+}}* [[C_PRIV]], i{{[0-9]+}}** [[REFC:%.+]], 359 // CHECK: bitcast [4 x i{{[0-9]+}}]* [[E_PRIV]] to i8* 360 // CHECK: bitcast [4 x i{{[0-9]+}}]* %{{.+}} to i8* 361 // CHECK: call void @llvm.memcpy 362 // CHECK: store [4 x i{{[0-9]+}}]* [[E_PRIV]], [4 x i{{[0-9]+}}]** [[REFE:%.+]], 363 // CHECK-NEXT: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]], 364 // CHECK-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]], 365 // CHECK-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1 366 // CHECK-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]], 367 // CHECK-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]], 368 // CHECK-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1 369 // CHECK-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]], 370 // CHECK-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]], 371 // CHECK-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]], 372 // CHECK-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1 373 // CHECK-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]], 374 // CHECK-NEXT: [[E_PRIV:%.+]] = load [4 x i{{[0-9]+}}]*, [4 x i{{[0-9]+}}]** [[REFE]], 375 // CHECK-NEXT: [[E_PRIV_2:%.+]] = getelementptr inbounds [4 x i{{[0-9]+}}], [4 x i{{[0-9]+}}]* [[E_PRIV]], i{{[0-9]+}} 0, i{{[0-9]+}} 2 376 // CHECK-NEXT: store i32 1111, i32* [[E_PRIV_2]], 377 // CHECK-NEXT: ret void 378 379 // CHECK: define internal {{.*}}void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, i32* dereferenceable(4) %{{.+}}, [2 x [[S_INT_TY]]]* dereferenceable(8) %{{.+}}, [[S_INT_TY]]* dereferenceable(4) %{{.+}}) 380 // CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128 381 // CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], align 128 382 // CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], align 128 383 // CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], align 128 384 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], 385 386 // CHECK: [[VEC_REF:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** % 387 // CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** % 388 // CHECK: [[S_ARR_REF:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** % 389 // CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** % 390 391 // CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], align 128 392 // CHECK: store i{{[0-9]+}} [[T_VAR_VAL]], i{{[0-9]+}}* [[T_VAR_PRIV]], align 128 393 // CHECK: [[VEC_DEST:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8* 394 // CHECK: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_REF]] to i8* 395 // CHECK: call void @llvm.memcpy.{{.+}}(i8* [[VEC_DEST]], i8* [[VEC_SRC]], i{{[0-9]+}} {{[0-9]+}}, i{{[0-9]+}} 128, 396 // CHECK: [[S_ARR_PRIV_BEGIN:%.+]] = getelementptr inbounds [2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* [[S_ARR_PRIV]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 397 // CHECK: [[S_ARR_BEGIN:%.+]] = bitcast [2 x [[S_INT_TY]]]* [[S_ARR_REF]] to [[S_INT_TY]]* 398 // CHECK: [[S_ARR_PRIV_END:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], i{{[0-9]+}} 2 399 // CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_INT_TY]]* [[S_ARR_PRIV_BEGIN]], [[S_ARR_PRIV_END]] 400 // CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]] 401 // CHECK: [[S_ARR_BODY]] 402 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* [[ST_TY_TEMP:%.+]]) 403 // CHECK: call {{.*}} [[S_INT_TY_COPY_CONSTR:@.+]]([[S_INT_TY]]* {{.+}}, [[S_INT_TY]]* {{.+}}, [[ST_TY]]* [[ST_TY_TEMP]]) 404 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* [[ST_TY_TEMP]]) 405 // CHECK: br i1 {{.+}}, label %{{.+}}, label %[[S_ARR_BODY]] 406 // CHECK: call {{.*}} [[ST_TY_DEFAULT_CONSTR]]([[ST_TY]]* [[ST_TY_TEMP:%.+]]) 407 // CHECK: call {{.*}} [[S_INT_TY_COPY_CONSTR]]([[S_INT_TY]]* [[VAR_PRIV]], [[S_INT_TY]]* {{.*}} [[VAR_REF]], [[ST_TY]]* [[ST_TY_TEMP]]) 408 // CHECK: call {{.*}} [[ST_TY_DESTR]]([[ST_TY]]* [[ST_TY_TEMP]]) 409 // CHECK-NOT: call {{.*}}void @__kmpc_barrier( 410 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) 411 // CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* 412 // CHECK: ret void 413 414 #endif 415 #else 416 struct St { 417 int a, b; 418 St() : a(0), b(0) {} 419 St(const St &) { } 420 ~St() {} 421 void St_func(St s[2], int n, long double vla1[n]) { 422 double vla2[n][n] __attribute__((aligned(128))); 423 a = b; 424 #pragma omp parallel firstprivate(s, vla1, vla2) 425 vla1[b] = vla2[1][n - 1] = a = b; 426 } 427 }; 428 429 // ARRAY-LABEL: array_func 430 void array_func(float a[3], St s[2], int n, long double vla1[n]) { 431 double vla2[n][n] __attribute__((aligned(128))); 432 // ARRAY: @__kmpc_fork_call( 433 // ARRAY-DAG: [[PRIV_S:%.+]] = alloca %struct.St*, 434 // ARRAY-DAG: [[PRIV_VLA1:%.+]] = alloca x86_fp80*, 435 // ARRAY-DAG: [[PRIV_A:%.+]] = alloca float*, 436 // ARRAY-DAG: [[PRIV_VLA2:%.+]] = alloca double*, 437 // ARRAY-DAG: store %struct.St* %{{.+}}, %struct.St** [[PRIV_S]], 438 // ARRAY-DAG: store x86_fp80* %{{.+}}, x86_fp80** [[PRIV_VLA1]], 439 // ARRAY-DAG: store float* %{{.+}}, float** [[PRIV_A]], 440 // ARRAY-DAG: store double* %{{.+}}, double** [[PRIV_VLA2]], 441 // ARRAY: call i8* @llvm.stacksave() 442 // ARRAY: [[SIZE:%.+]] = mul nuw i64 %{{.+}}, 8 443 // ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* %{{.+}}, i64 [[SIZE]], i32 128, i1 false) 444 #pragma omp parallel firstprivate(a, s, vla1, vla2) 445 s[0].St_func(s, n, vla1); 446 ; 447 } 448 449 // ARRAY-LABEL: St_func 450 // ARRAY: @__kmpc_fork_call( 451 // ARRAY-DAG: [[PRIV_VLA1:%.+]] = alloca x86_fp80*, 452 // ARRAY-DAG: [[PRIV_S:%.+]] = alloca %struct.St*, 453 // ARRAY-DAG: [[PRIV_VLA2:%.+]] = alloca double*, 454 // ARRAY-DAG: store %struct.St* %{{.+}}, %struct.St** [[PRIV_S]], 455 // ARRAY-DAG: store x86_fp80* %{{.+}}, x86_fp80** [[PRIV_VLA1]], 456 // ARRAY-DAG: store double* %{{.+}}, double** [[PRIV_VLA2]], 457 // ARRAY: call i8* @llvm.stacksave() 458 // ARRAY: [[SIZE:%.+]] = mul nuw i64 %{{.+}}, 8 459 // ARRAY: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.+}}, i8* %{{.+}}, i64 [[SIZE]], i32 128, i1 false) 460 #endif 461 462 463